I actually found that with all the advancements of technology, I was able to get 10-15 more average. Which May not seem like much, but with the higher resolutions and stuff 10fps can be huge. ( In my experience anyway )
@@garedroessel4315 10fps gains today is noticeable for sure. My comment was in reference to the P4 and ATI 9800 AIW I was running at the time. I should have elaborated a bit more in my original comment. My monitor is a 27" 1080p @ 75mhz. I vsynch at 75 and just play. Occasionally I'll unlock it to see what my frame rate can be. I always put it back though. I don't play demanding games however.
-40 on the curve is way too aggressive in my experience, Jay. -30 might work (typically only the best 7000 series hit that all core) and -20 should work fine
-40 might only work for a test (might even not work) and even -30 might only work for a few dozen minutes before crashing in a game or just going on the web.
I could only get -10 on my 7950X3d, but someday I'd like to do a per core undervolt. My cinebench scores went up after the undervolt, so imo it's definitely worth the time if you're into optimizing things
Yeah you cant just crank PBO and expect it to work so high especially all core. Needs to tone it down a bit, find the sweet spot then start going core by core. It has nothing to do with Fanboyism like Jay said. The crashes are due to pushing PBO without tuning it manually.
I went from a 1600 to a 5700x. Might maybe get a friend's 7600x and push it hard. But otherwise, I am in no rush. Probably just wait for the Zen 6 X3d stuff
That is why AMD didnt do the raw power thing. It didnt make sense to crank power so much for such little gains. At 90% of full performance, you get a huge efficiency buff.
I think what AMD did made sense. If people didnt care about poor efficiency then turn on PBO. But by default at 65W it is very efficient. The disapointing thing is Zen5 is very small upgrade over 7xxx series.
A -30 curve optimizer offset is not -30 millivolts. It is actually somewhere between -90 and -150 millivolts depending on the part of the frequency curve you’re looking at.
The board literally had the unit on screen right there. -40mV! Ryzen Master claimed -33 counts, he did NOT set this up manually. It was the result of Ryzen Master's algorithm which seems to be utter crap. Might just be an early adopter issue.
I mean, why could he be bothered to research from someone like Scatterbencher to better understand how to overclock 9 Series. Just toggle button with garbage RAM.
@@kyoudaiken "The board literally had the unit on screen right there. -40mV" yeah, becasue everything inside the UEFI is a 100% correct? No, the UEFI programmers simply put mV behind the number, because they did not understand the issue either. There's a good explanation from AMD what the values mean. You do not shift the voltage-frequency curve in terms of voltage, but you shift the curve in terms of frequency at the same voltage. And if you don't need the higher frequency, the CPU will downclock and with that reduces the voltage.
This was how chip designers used to do, they default ship you the most balanced settings and anything above you tune by your own, even the corrupted Intel used to be like this. Jay is no longer the Jay we loved, he is just biased to Intel and pick every issues that are non issue against AMD. He said he wanted the first video as close to stock as possible, but he didn’t even mention PBO one time in the video as this feature did not exist. What a big disappointment.
@@Raghav-gd3fn its not. 9700x is better than previous gen. It just works differently. I have it now for 24hrs. Difference is power consumption. If you need CPU which runs all cores at maximum frequency, then dont even look at Ryzen 9000. Low TDP is enough to boost few cores to maximum performance. How often do you really need all cores to run at maximum power? If you play older games then they can use only few cores, and for this type of workload 65W TDP is more than enough. And you can see those 5,4-5,5GHz frequencies on few cores. This is super easy to cool. Compare to my previous i7-7700k I decrease my fan curve by 20%. PC is very silent. I dont need more performance out of it. For all my tasks I do -> my PC draw less Watts and run super fast.
I do see most people talking about offset as mV. This is not correct for AMD. It is steps. Each step=3-5mV. So a -30 offset is 30 steps of 3-5mV each = -90mV to -150mV.
yeah... this video was painfull to watch. I love how people pretend to know something without study it. 2024 I guss... they want everything in not time. thanks un-social networks... thanks tik tok videos.
I'm glad you explained this here. It's not at all obvious looking at the software or in the bios. It very much gives the impression that it is actual mV, and not steps.
That restart into 95c feels like some old recovery fallback/defaults not getting updated to 85c, though it is weird that most of the other settings stayed.
The default temperature limit on the 9700X is still 95°C, and there was no change to the spec from the 7000 series. The AMD product page also shows this. Not sure where he even got that idea. (X3D 7000 series parts were 89°C).
@@AK-Brian I would assume these are old Asrock preset settings that stayed in the bios from the times when Motherboards and x3d chips were burning 1 year ago, and they were probably good for x3d chips and probably still are, for manual tuning you press PBO enabled and you do PBO manually, which is not rocket science, but Jay doesn't care to do any research or preparation for his videos
why is 7800x3d missing from the charts I get the new chips are not x3d but you have intel on the charts it would be nice to include the best selling cpu of this generation for reference
Because he just established these charts just now and hasn't tested a whole lot of CPUs yet, because he has health issues and is in the middle of moving his whole studio...
i am run an intel but was wondering the same thing because when the 50s come out i want to build a AMD system never had one and my intel is not giving me any problems its a 14700k/ 4080 first system in 20 years
he is always shooting his AMD systems in the foot with the RAM then complains it doesnt run well or whatever, when people been running 6000+mhz on AM5 for almost 1 year at this point, its all around the internet, not to mention all of his Hardware channel colleagues always have their AM5s running at least 6000mhz with semi-decent timings, not to mention Buildzoid gave settings for decent 6000mhz with a bit tighter timings that were compatible with almost everything shortley after launch, I would assume it works on 99.9% of the systems right now, Jay never cared to put them in even once
Well, this does seem to be a 9000 series related issue. 7000 series runs 6000MHz fine, but a lot of tech channels seem to be having issues getting the 9000 series to run stable with current BIOS revisions available on some boards. 🤔
You got curve optimizer completely wrong. The number you enter in the curve optimizer is called count and 1 count correspond to 3 to 5 mv. So -33 CO = -99mv to -165mv which would be great if that worked but most peeps won't even get that on there best cores lol Start at -15 CO(-45mv to -75mv)
Either you or I got curve optimizer completely wrong. I was under the impression that your best cores ran closer to zero, as they are the cores doing the very heavy and very light workloads. They require more volts to improve stability and to hit higher boost. For example, my 58x3d has -15 and -17 on the 2 best cores, and the other cores all in the mid to high -20's. But this... "-99mv to -165mv which would be great if that worked but most peeps won't even get that on there best cores lol" ...reads like you believe the opposite to me?
never gonna see this in this channel, he is reluctant to deal with any kind of Ram set up, if it doesn't work he just puts something lower until it works blindly, its super dissapointing, like every other channel and person who has a bit of hardware knowledge is running Ryzens past 6000mhz easily, this is not 2022 anymore
The difference in price between DDR5 5200 and DDR5 6000 is negligible at best. High quality price vs. whatever anyone else buys is almost non-existent in that speed range. The fact that Jay selected 5200 instead of the memory's rated speed of 6000 seems odd.
@@ZackSNetworkthis is an odd argument. Given Ryzen 7, 8, and 9000 series CPUs involve a sweet spot clock speed of 6000 MHz, why not push for 6000 MHz low latency? Is there something wrong with spending around $120 USD on a 2x16GB kit? We aren't buying 8000 MHz RAM.
Random black screen restarts happen because of light load instability imo. There is always a side effect of curve optimization for higher loads which is that it moves the entire curve including for low loads. It’s why their curve shaper feature might actually be pretty useful. It happens more often when you are doing ECLK overclocking.
Yup after playing around with CO on my 5950x for the last year, low load is almost always what shows instability for at least this gen of chip. I could game at -30 CO and +75mhz offset ALL DAY, but it'll crash within an hour of browsing the internet. At -30 CO and +50 offset, its rock solid.
I agree with this too, when I'm on the limit with my 7950X3d using curve optimizer (-20 iirc) it would only crash at low load. Someday I want to do a per core undervolt
Yep. I remember when setting up my 5900x that I often thought that I've got it right now, only for it to randomly crash while on desktop doing nothing much like a day or even a week later. Took me a while to get this one set right, but was very glad I did it per core because I've had some cores only take as little as -13, while others were fine with up to -27. Was a bit of work, though.
i'm willing to bet that the 9800x3d will be massive improvement of performance. these zen5 are bandwith starved. i'm also willing to bet that the reason they went with the 65W is to shift the attention to the power limits and not the bandwith limits. Waiting for Wendell or phoronix to test how much faster the Epyc version of zen5 with 8 channel bandwith is going to be.
To be fair, anyone who wants to move that fast from ANY X3D chip soo fast, just have to much money and nothing else cuz on performance point of view, it js just pointless to move from it if you have it UNLESS you really want to have ddr5 machine
@8:00 Check your ram multiplier when over clocking it. Good Tip. In the past, ive gotten new speeds to populate in the option table by changing the bin interval.
Thanks for the walk through. Don't understand though why you expect that undervolting the CPU, esp. with an aggressive offset of -35mV shouldn't lead to crashes. Finding a value where it runs stable is the point of tuning that value in, no?
One change made was location of internal temperature probe points...the result is a 7-9 degree drop in "reported" temp (under like for like temps) so dropping max temp is needed.
@@RavTokomiSteve explained this in the gamers nexus review. They moved some probes around and it just happens to cause it to report lower. It's not a big thing but just know that's why the tj max has been lowered.
@@RavTokomi That's NOT why they did it, they did it to get a better safer boost curve algorithm and increase the overall efficiency Enough with the Conspiracy Theory nonsense, we get enough of that BS from politics
@@RavTokomi Would that explain the TDP drops? Unless AMD has also changed how they calculated them, it would still require the same level of cooling if the temperature is actually the same, and only differs because they're measuring it differently
You could have done a power parity comparison with the 7700X (PPT: 142, TDC: 110, EDC: 170 on the 9700X). Another thing is you went too aggressive on curve optimizer, typically to get -30 to -45 CO you need to bump the LLC up one, unless you have a golden sample. You would have been better off test -20 which is more realistic. Also test with DDR-6400 since this will probably be the new standard on these chips, it seems like every one of them does it just fine with a 2133.33 FCLK.
OCCT was a lifesaver for me with my 7800x3D cpu I was able to realize that in the first 7 cores it supported -22 and only the eighth -20 What I like most is that you can configure it to stop the WHEA errors and tell you which thread failed and with CPUZ you can get the thread order list to which core it belongs to and know which one does not support the offset and lower it a bit. real lifesaver
i had a falcon NW system back in the early 2000's, it was a BEAST at the time with its overclocked q6600 and 4 gb ram and 8800gts (512 gb) and it lasted for almost a decade before i eventually went to a LGA 1366 setup, funny thing is i still have that 1366 system but but i really wish i had kept my old falcon instead these days
@@HesterDW well if you had the headroom you could of done a little overclocking on those, but i feel yeah the dual cores were nice compared to P4's but the q6600's felt like melted butter compared to the duos frozen margarine XD
Jay, I mine with my 7000 series and pbo plus CO is popular to keep clocks as high as possible while remaining efficient. All core PBO usually leaves performance on the table since theres usually a few weak cores. Best to do per core core offset which will require a lot more tweaking. For 7000 series, CO -30 was rare and usually -20 or worse was the norm depending on silicon lottery.
Yeah, it was too ambitious to go for -30 but to be fair finding the optimal undervolt pbo can take MONTHS. I have had a 5800x undervolt around -25 -28(per core) for 8 months and I still get some crashes sometimes but it's quite rare, so I have to tweak my curve optimizer by time to time
@@dagnisnierlins188 I did it, got some values (no more crashes using occt or core cycler) A few months later my PC just crashed and reboot, so it wasn't stable. It's just my personal experience, core cycler was really good to find the base per-core values to use. Now I just use my computer if it crashes, check the logs and change the values in the BIOS. Nowadays I rarely get any crashes but at least for me it took a lot of time and actually using my PC (gaming, watching videos, editing videos, etc...) to get per core values where it doesn't/rarely crash.
The manner by which you talk about and comment upon data makes this video vastly richer and more informative than others.' Regardless of whether your testing methodology is or is not ideal by everybody else's standards. Chapeau! 👍
@@kyoudaiken My experience trying out Ryzen master is that it's overly optimistic. If all cores are 100% stable with a -30 offset Ryzen master will try to go beyond. Whilst it may pass it's own tests you have a very high chance that it'll just turn off in other stress tests/games. My 5900x was a brilliant undervolter. 200+ MHz and -30 offset all core in PBO, with a further -0.075v offset in the base mbo overclocking menu. Rick solid stable, scores over base pbo and runs very cool (I'm a silence freak so want fans going as slow as possible)
I'm still rocking an AMD 8000 series. FX-8350 that is, lol. I also have a Xeon E3-1271 V3 (equivalent to a i7 4790). I don't do a ton of gaming, so it's not a big issue for me. My hope is that I'll be able to pick up a decent AM4 system soon when everyone around me upgrades to AM5.
why are people mad that the cpu isn't pulling 1.21gw out of the box and isn't turning their pc into a goddamn fusion reactor? or do people really want their future phones to have a lead acid car battery built in for it to last only 5h all while reaching temperatures that can melt steel in seconds? it's both cooler/more efficient and better performing, why all the "ryzen 9000 sux" videos? also weren't we laughing at intel/amd when their cpus were impossible to cool because they were drawing so much power? did we just do a 180 in that regard in just a few days for some reason?
The temperatures between 7000 series and 9000 series are not directly comparable because they moved the temperature sensor locations. All the boost algorithms have been adjusted for a new temperature curve. People obsess with temperature but what really matters is the Wattage which is the true test of how much energy is being used and the efficiency of the device. 100Watts @ 70C = 100 Watts @ 90C Yes temperature (and voltage) is still import because if either is too high it can cause electromigration (Intel's current problem) and chip degradation but that has been accounted for in the new temperature curves
@11:15 - "This is kinda like intel territory on all-core by the way" - completely ignores that all of 7000-series boosted all core almost exactly as high as single core, and higher than the result here, 7700X hit 5.4 GHz all core all day. Also, (Auto) Curve Optimizer is undervolting the CPU / increasing frequency during the setup / testing - of course it is going to crash, why would you say it shouldn't? It's supposed to find the limit, even though the super quick testing can't catch every possible error and you should be doing this yourself. If you want to push single core with PBO on or Curve Optimizer on, you must set a boost clock override / "Auto OC" up to +200 MHz. A lot of CPUs are going to have some headroom there.
@@tim3172 It depends on the workload, but something like Cinebench R23 would be above 5.3 GHz with PBO (~20K score) even on a bad CPU, assuming you have the thermal headroom (ie. 360mm AIO). Even if we assume 5.2 GHz (pretty much all 7700X at stock) you're missing the point of the comment - "This is kinda like intel territory on all-core by the way" is what he said when PBO was enabled on the 9700X and hitting 5.2 GHz in Cinebench.
Jay, what I’m more interested in is the comparison of 65w parts… why aren’t you comparing the 65w 9700x to the 65w 7700(non-x)? Let’s compare it to its own weight class please.
That still doesn't negate the fact that it's not working correctly. Ryzen Master is supposed to write the values to the BIOS. Unless he turned that off.
Ryzen Master is not now and has really never been stable. Make your adjustments in the BIOS after figuring out your two best cores using Ryzen Master to test. If you want AMD to work, make the adjustments in the BIOS. And don't start with -40 or even -30 on Curve Optimizer. That works with some chips .. it does not with most.
Its a 65 watts part coz there was no peformance ulpift with more power and it wasnt an uplift from previous gen so they just said atleast make it 65 watts so that people dont get that dissapointed Also people saying that this cpu is efficient by comparing 105watt 7700x to 65watt 9700x but unstead they should compare to 65watt 7700
I'm with you, but on that last part, I blame AMD - why name the successor of the 7700 as though it were the successor of 7700X, unless they wanted reviewers to compare the 9700X with the latter? They did the same thing when the 2700 successor became the 3700X rather than the 3700, but the generational jump was large enough then that it was still an upgrade from the 2700X.
In my opinion the efficiency is nice for SFF builds, but having the option to power it up will be nice for other users too. Options are always a good thing
@@blainehock754 Also consider that for multi thread workloads this new CPUs do not seem to offer a real improvement over intel's 13th and 14th gens (due to the core count disadvantage). For instance if I lock my 12600k at 90W (and stock voltage) I get 16'211pt in R23 which is already quite close to the 16'900pt of the 9600x, thus 13th and 14th gen are gonna be more efficient at that TDP. Besides all the recent issues with intel cpus, I believe their bad reputation for efficiency is only Intel's fault, the architecture is actually decent, but they decided to push it to the brink of catching fire just for winning benchmarks, what a shame (and massife f***up).
Now we only need some decent SFF GPUs. And intel 14thgen can be really efficient - not even hard to achieve, but the default settings were just atrocious and it finally boiled over.
I feel like amd expects users to have PBO enabled. But its still good to test both since one is technically overclocking. Also if you want to mess with curve optimizer again. Try up to -10max then add negative curve shaper on medium/high frequency. This will get the performance boost without messing with stability. Or at least its working so far for me.
C'omon do a research before, and learn how to do the things on bios side. I have a ASRock MB and they have a lot of options for manually adjusts. And yes, here the memory profiles don't work a content. I have to set the timmings and frequency manually. PBO -33? It's too much... Starts with -10,-20 and test stability.
@@tsvetanbliznakov7694 i've been saying this since the launch, none of them know what they are doing with PBO, let alone even notice their chips are only boosting to 4.4 ghz on Asus boards
*There is a 3-8% uplift in fps gains when using 8,000MHz ram, plus an additional 4% when tightening the timing, plus PBO, etc. The 9700X is actually a whole league ahead of the 7700X. It keeps up perfectly with the 7800X3D. And with using less electricity, and keeping cooler.* *But only a couple of people went against AMD's stupid rule book.* *7000 series support SPD 5,200MHz. 9000 series support SPD 5,600MHz. It is designed to run 6,800MHz to 8,000MHz ram and get better results.*
Hey there... I am still only on AM4 with a Ryzen 9 3900X with 2x32GB ram (tried 4x32GB but behaved weird)... Waiting on the test results on the new AM4 cpu's... I do not experiment with overclocking, because i want a rock solid system that just works... Keep up your great work, because it is very interesting... The best part is the tongue twisters that happens sometimes, because it happens for all of us...
It reverted to 95 because it detected the crash and reapplied defaults. Look at 9:50 in the video where it is also set to 95 after the crashes when you tried to set the voltage offset. You then lowered it from 95 to 90. I don't know why it was 85 at the start of the video. I can only assume it got set there while you were trying it before recording. No, the platform isn't perfect, but this was user error my friend. ;) You even mention in the video how it will go back to defaults on crash.
@@Jayztwocents It's literally in the video, before you load profile 1 it displays 95. Applying a preset or auto OC then expecting perfect stability isn't always going to happen. Lastly, cinebench is a terrible test and has nearly zero bearing on real world productivity.
@@Jayztwocents Perhaps nobody told Ryzen Master yet. 😅 Also weird it would choose an Expo profile as a "safe" post crash default. I'm curious how much of this is BIOS vs Master. My MSI X570 will go straight to JEDEC defaults when it detects crashes.
@@BigMan7o0 As someone who just built my AM4 system last year long after AMD said the TJmax was 95C, the default set in the bios is 85. I noticed that when I was just running tests out of the box and I had to change it myself to get it above 85 to avoid throttling. This is the exact same things that AM5 will do, even if you update the bios.
Appreciate the video! I much prefer separate videos for OOB and "overclocked" performance. This way you can do a bit of a deeper dive into both subjects. Again, great video!
I find it interesting that all the benchmarks showing big leads for Intel include a 14700k and 14900k while not including the 7900x and 7950x, which are both much closer in MT workloads and both cheaper. At the launch prices for the 9600x and 9700x, they certainly aren't good value though.
I have been running AMD since I bought my Ryzen 5 3600, it have never restarted for me. Didn't know that was a "thing" for AMD, currently I'm running a Ryzen 9 3950x because I found it for a good price and I like to play with VM's. But still havn't had any random restarts. I do however had some random strange crashing. Removed one stick of RAM to run only 16Gb and it worked fine for month, regardless of which of the sticks of RAM I was running it worked fine. But as soon as I used both at the same time (no matter the slot) it randomly crashes again.
same ive had phenom, fx, ryzen 2000,ryzen 3000 and since 2022 i7 12700kf. Never had an AMD restart but all AMD i've had default voltage by MB was too high and could be ajusted for more efficiency and less heating
I can go back as far AM386SX and that thing never crash on me … unless DOS/Windows wanted to. Doesn’t mean the newer ones won’t crash. I’m currently using 3700x, still solid all year around.
the problem is people are expecting an overall improvement but mainly the biggest improvements are within the AVX instruction sets, especially AVX-512. I'd be interested to see a comparison of CPUs using the PS3 emulator RPCS3
I'm not completely in the loop but from my use of RPCS3, I don't have a single-issue running any game at a locked 60 that I've tried. The only issue I seen people ever had performance-wise was God of War at 4K due to a memory bandwidth limitation on VRAM of some kind. So from my understanding AVX-512 will help convert the PowerPC code to x86 more efficiently?
@@gex581990 Yeah after doing some research AVX512 helps a lot depending on where the bottleneck is for each game.(SPU or RSX) Once you start pushing more than 60fps the bottleneck switches too. Seems like the 12900k with AVX is currently the best but the chart did not have Ryzen 9000 series yet.
I've seen several different reviews now and its all over the place. But not everyone is using the same memory or mb, so that makes it interesting. I'd like to see a little bit of a do over in say 3 weeks to a month because to be sure amd is gonna release an updated bios. Perhaps that will make memory more stable and we can get a better feel for these altogether as a group. How all 4 cpus stack together is what I curious about.
To AMD; Thank you for listening to the enthusiast community. Fun overclocking has been missing from the high end desktops due to factory setting pushing chips for all they're worth. Thanks for putting it back in our han... fingers.
I was really happy to see you include Handbrake in your testing suite. That is what I really look at when considering a CPU upgrade. (I do a lot of video encoding)
I agree, for the out of the box experience they need to start testing these with the ram specs listed. I am pretty sure they can use the DDR5 6000 to void your warranty.
@@TheJmac82 Agree there. Not worth risking the RAM speed with Intel now they've increased Warranty period from 2 to 5 years. THAT alone brings them back into the equation again. The other 2 AMD CPUs with higher core count should be more appealling.
That's what AMD guarantee that will works out of the box. They say that 6000 MT/s (CL30) is the sweet spot. Intel's 13th and 14th gens are on paper also supported up to 5600MT/s, but you can enable XMP to get higher speed (your mileage may vary).
@@N3v3r_S3ttl3 That just further pushes my point though, so based on that if for what ever reason I happen to get one that cant run over 5600, i cannot RMA that but i will never get the speeds that are being reviewed. I really think it is deceptive. And where do they say that? I see tons of documents to the press saying that but no where have I seen AMD state to the public to up their ram speeds to 6000... why? because they dont want to be sued when there are issues. So with that its at minimum deceptive advertising.
Always like seeing the difference between PBO and Stock, would like to see 3d slicer performance. Obviously everyone uses different slicers but seeing the difference between old cpu and new cpu on Orca, Cura or Prusa slicer would be a huge help when deciding to upgrade.
Im pretty sure just "PBO Enabled" in the BIOS means 95°C Tjmax, since its the default Tjmax they probably felt like mentioning it would be more confusing or something.
If you compare multicore performance the 9700x with pbo just barely beats the 13600k and 12700KF which cost 265€ and 190€. So, it's significantly too expensive. The 9700x should cost 300€ and the 9900x 400€. Currently the 7700X at 290€ is 10€ cheaper than the 9600x.
Didn't they move the core thermal sensors to a cooler spot? That was the gist of their "magic trick" to make 9000 series run cooler than 7000 series, from what I heard. Hence why default Tjmax is also lower.
I believe its 85 c because amd said they adjusted the locations of where they now get their temps from to more sensitive areas so overall the temps are lower for the same to more performance
@@Jayztwocents It's not 85 degrees C. AMD states the 9000 series has a 95 C TjMax on their own website / cpu spec page. So obviously it would (and should) show 95.
@@Jayztwocents maybe the bios isn't the best? I mean they also made new boards for these processors too. Maybe the new boards will do better with reporting?
Cool vid as always. Be nice to see either the 7900/50 (or x variant) included in these charts, as they're closer to intel in terms of thread count - for the multi core stuff it'd make a lot of sense as a decent price competitor for people on those types of workload
My 3700X is my first solid CPU since my Athlon 4200+ X2. 1700: had 3 of them, twice replaced by AMD. It now crashes under full load out of warranty. 5900X: recently developed a "Cache hierarchy" BSOD problem.
I still remember the hype when it came out. It was such a beast compared to any mainsteam desktop cpu at the time. Infact it was beating intel’s 2000 dollar 9980xe😅. It’s sad that amd didn’t increase the core counts to match intel, especially in the midrange.
from my limited experience playing with PBO and CO with my 5800x, ive realized that CO helps single and all core workloads while PBO mostly only affects multi core perfeormance
Great vid. As a car guy, I am pretty sure that you are more about performance than efficiency. I am the same way. Will be nice to get a more efficiency chip but is always a tertiary criteria for me when choosing a platform. I hope manufacturers just keeps things simple.
@28:32 chart looks like a long AC signal holding pulse signal inside it I imagine the performance of any system would be affected by this, but especially faster clock speeds would see less power at these intervals! Almost makes me wonder if different power supplies might provide more stable power? Or is this how the cpu is using the power?
@@gamingwithegoon They idle at about 23,5W though. 500%+ higher power consumption vs about any intel just to run a calculator or to nothing. (And these are "efficient" by AMD standards.) They also use way more at simple tasks. Like watching youtube, etc..
@@TSracer1idle power usage only really matters for a server use case where its always powered on. I think power consumption under load is more important for consumers. Consumption under load effects heat output and noise where you probably wont notice a 20w idle vs a 2w idle but will notice a 50w load vs a 300w load.
You have many people are stuffing 13900k and 7950x3d in tiny cases like the NCASE T1 without issues. This CPU is a waste of time due to crappy performance gain from last gen.
Thank you for the fantastic video Jay! I’ve been getting REAL tired of people saying “oh but it just needs to be juiced and it’s an amazing uplift” just because they completely ignored/misinterpreted most of a Der8auer video.
@@brasiliafc2138Turbo boost isn’t the same as PBO. AMD’s CPB is the same which they’ve also had for years. Intel doesn’t have anything similar to it since they left it up to the motherboard manufacturers with something called MCE.
Why are you listening and not using ryzen master? I haven't herd the reason? Is it okay still to use for tuning then as long as we set the settings in the bios?
@@ABaumstumpf Because everything wirtten is always 100% correct? It's like adressing you with Steve. And based on your logic it's written, so it can be wrong, ain't I correct Steve?
@@johnscaramis2515 "Because everything wirtten is always 100% correct?" No, but there are 2 very different things - the BIOS-preset is applying a 30mV offset, where as the curve-editor is using 3mV steps but that one never showed 30 but 33. So he is mixing up 2 things there and i was asking WHICH he meant. On the other hand there are people like you......
I feel like the most important thing to come out of this video is that it's pretty clear the bios and software and so on are still pretty cooked for the 9000 series. I'd wait 2 months or so before getting one to let them sort this out. And make them cheaper!
To be honest, I have experience with AMD in retail back in the 90s and personal experience with the 1950X Threadripper, 3960X Threadripper and 5000G APUs on AM4. AMD always has been janky and weird and it didn't change to this day. You can wait years, I doubt it will get significantly better. The only thing you can do is cope with how the platforms of AMD are. Generally, AMD's monolithic Ryzen designs seem to be much more stable. These include the 5000G and 8000G series of desktop processors. I think they're still having massive issues with the chiplet design.
Awesome to see these follow ups. And you do it right Jay. I am in the camp that launch reviews, are out of box "open it and plug it in" reviews. The overclocking and configurability are features worth mentioning, but they are not the product being sold. Launch videos dont need to be a 1 hour deep dive covering out of box AND overclocking/PBO/tweaked settings. It becomes a mess.
pbo works fantastic for boosting single core performance but you need the temps to have it boost, did you put scalar x10 and plus 200mhz? and did you undervolt all cores and testing them one by one? don’t think so
I saw a news article about a msi bios update for zen 5 that adds up to 10% more performance i feel as though the bios and software is just funky still.
honestly if they make the x3d chips spicy performance wise and power efficient I am sold. Power efficiency is good for me specially for the summer months where my room cooks. Coming from a 5800x these are very interesting specially once you look at the 12 and 16 core x3d parts. Power and efficient at the same time would be a dream come true
TLDR: I’m unsubscribing. I’ve really enjoyed Jayz channel over the years. However, not providing a comprehensive/unbiased (or, honestly at the bare minimum, at least commenting on *potential* shortfalls in his analysis) review turns me away. Lampooning his audience in the first few seconds of his follow-up video does not engender trust or goodwill. He had some issues with one of his processor and wrote it off since Gamers Nexus had a similar problem. GN actually worked with an AMD engineer though; that doesn’t seem to be the situation here though. I don’t know if we will see responses to the comments on mV adjustments here, but I’ll try to find multiple other reviewers that have more experience with AMD’s hardware.
21:50 Amd will make a 2 chiplet 8x Zen 5 core and 16x Zen 5c core. They have not announced it. But if you have 2 chiplets. Why not do 1 Zen5 chiplet and 1 Zen5c chiplet. 48 threads.
If you are getting random restarts on AMD 7000 systems, go into BIOS and disable "global c-states". Doing this fixed the random restarts I had with windows 11 and have had no issues since - AMD 7950x/Asus ProArt X670e-creator/Corsair 2x48gb DDR5 5200mhz
it's easy to think AMD laid en Egg....... I don't think they did though..... Lets wait and see what the 9000 series 16 core parts look like....... Intel ............................
I stopped overclocking in the mid 2000's. I spent more time chasing 1 - 2 frames than I did playing games. Not worth it IMO.
But that's the real fun 🦾
I actually found that with all the advancements of technology, I was able to get 10-15 more average. Which May not seem like much, but with the higher resolutions and stuff 10fps can be huge. ( In my experience anyway )
@@garedroessel4315 10fps gains today is noticeable for sure. My comment was in reference to the P4 and ATI 9800 AIW I was running at the time. I should have elaborated a bit more in my original comment.
My monitor is a 27" 1080p @ 75mhz. I vsynch at 75 and just play. Occasionally I'll unlock it to see what my frame rate can be. I always put it back though. I don't play demanding games however.
You stopped a little too early. The Intel E2140/60/80 were true kings of overclocking (2006/9) After that it went downhill and that is when I stopped.
@@rockfella27It was fun in the beginning, but we (friend and I) spent WAY too much time chasing something that really didn't matter.
-40 on the curve is way too aggressive in my experience, Jay. -30 might work (typically only the best 7000 series hit that all core) and -20 should work fine
-40 might only work for a test (might even not work) and even -30 might only work for a few dozen minutes before crashing in a game or just going on the web.
I could only get -10 on my 7950X3d, but someday I'd like to do a per core undervolt. My cinebench scores went up after the undervolt, so imo it's definitely worth the time if you're into optimizing things
Yeah you cant just crank PBO and expect it to work so high especially all core. Needs to tone it down a bit, find the sweet spot then start going core by core. It has nothing to do with Fanboyism like Jay said. The crashes are due to pushing PBO without tuning it manually.
@@FireCestina1200 not only the crashes but having too low voltages weakens the single core performance or overal performance aswell
The random restart on my 7800x3d came from to aggressive PBO setting, i went from -30 to -25 to get it stable.
Realistically still two gens away from getting the feeling of FOMO. 5800x still going strong
I've got a 5950X and don't really feel any need to upgrade any time soon.
I think I'll change my 5800x3d when the DDR6 generation comes out... And even then a couple years later, when the platform prices come down.
Yep, my 5900X is a powerhouse (in my gaming/video editing PC) and I have no need to upgrade it.
I went from a 1600 to a 5700x. Might maybe get a friend's 7600x and push it hard. But otherwise, I am in no rush. Probably just wait for the Zen 6 X3d stuff
I just barely changed from my 2600 to a 5600x only because I felt like repurposing my old parts into another PC lol. Had absolutely no need for "more"
That is why AMD didnt do the raw power thing. It didnt make sense to crank power so much for such little gains. At 90% of full performance, you get a huge efficiency buff.
Meanwhile Intel: "May Zeus and Thors power push this CPU a few MHz higher!"
@@Fabian3331234333 CPUs getting thunderstruck by Ogun
It is the same story for ryzen 7000
They just rebranded the non X versions as X, they match exactly to the non X variants last year. Zen4 was also as efficient when at those wattages
I think what AMD did made sense. If people didnt care about poor efficiency then turn on PBO. But by default at 65W it is very efficient. The disapointing thing is Zen5 is very small upgrade over 7xxx series.
A -30 curve optimizer offset is not -30 millivolts. It is actually somewhere between -90 and -150 millivolts depending on the part of the frequency curve you’re looking at.
The board literally had the unit on screen right there. -40mV! Ryzen Master claimed -33 counts, he did NOT set this up manually. It was the result of Ryzen Master's algorithm which seems to be utter crap. Might just be an early adopter issue.
I mean, why could he be bothered to research from someone like Scatterbencher to better understand how to overclock 9 Series. Just toggle button with garbage RAM.
@@kyoudaiken "The board literally had the unit on screen right there. -40mV" yeah, becasue everything inside the UEFI is a 100% correct? No, the UEFI programmers simply put mV behind the number, because they did not understand the issue either.
There's a good explanation from AMD what the values mean. You do not shift the voltage-frequency curve in terms of voltage, but you shift the curve in terms of frequency at the same voltage. And if you don't need the higher frequency, the CPU will downclock and with that reduces the voltage.
28:20 IMO AMD made the right decision, default profile consuming half the power, for almost no loss in performance is absolutely worth it
It's kind of a rip off. IDK what you're talking about.
@@battlephenom8508 If you mean 9600/9700 CPU is bad value at its current price, I agree. If you think pumping power for minimal gain...nah
This was how chip designers used to do, they default ship you the most balanced settings and anything above you tune by your own, even the corrupted Intel used to be like this. Jay is no longer the Jay we loved, he is just biased to Intel and pick every issues that are non issue against AMD. He said he wanted the first video as close to stock as possible, but he didn’t even mention PBO one time in the video as this feature did not exist. What a big disappointment.
Would you say the same thing if intel did this exact same thing? Don’t make any excuses for a company who is all about the money.
@@Blaquegold I would, especially in Intel's case. Their idle-lower usage power consumption is very good, but on high loads its awful
I'm a truck driver and I care about efficiency. Less wattage = less heat in my truck.
😂
7700 still do the same while costing less but guess what people still delusion about 9700x think its great value, pure cope
@@Raghav-gd3fn its not. 9700x is better than previous gen. It just works differently. I have it now for 24hrs. Difference is power consumption. If you need CPU which runs all cores at maximum frequency, then dont even look at Ryzen 9000. Low TDP is enough to boost few cores to maximum performance. How often do you really need all cores to run at maximum power? If you play older games then they can use only few cores, and for this type of workload 65W TDP is more than enough. And you can see those 5,4-5,5GHz frequencies on few cores. This is super easy to cool. Compare to my previous i7-7700k I decrease my fan curve by 20%. PC is very silent.
I dont need more performance out of it.
For all my tasks I do -> my PC draw less Watts and run super fast.
AMD SUCKS
You mean less miles per hour.
I do see most people talking about offset as mV. This is not correct for AMD. It is steps. Each step=3-5mV. So a -30 offset is 30 steps of 3-5mV each = -90mV to -150mV.
yeah... this video was painfull to watch. I love how people pretend to know something without study it. 2024 I guss... they want everything in not time. thanks un-social networks... thanks tik tok videos.
AMD should make that more clear then.
@@conorstewart2214 absolutely.
I'm glad you explained this here. It's not at all obvious looking at the software or in the bios. It very much gives the impression that it is actual mV, and not steps.
That restart into 95c feels like some old recovery fallback/defaults not getting updated to 85c, though it is weird that most of the other settings stayed.
Kind of dangerous since that might be well over 100 given the lower temp reporting
The default temperature limit on the 9700X is still 95°C, and there was no change to the spec from the 7000 series. The AMD product page also shows this. Not sure where he even got that idea. (X3D 7000 series parts were 89°C).
@@AK-Brian Far as I remember, Steve from GN also said in his video that AMD told him 95°C is indeed the max safe temperature.
@@AK-Brian I would assume these are old Asrock preset settings that stayed in the bios from the times when Motherboards and x3d chips were burning 1 year ago, and they were probably good for x3d chips and probably still are, for manual tuning you press PBO enabled and you do PBO manually, which is not rocket science, but Jay doesn't care to do any research or preparation for his videos
That could have been one of the reasons for the recall.
Perhaps they changed the max temperature from 95 to 85 in light of Intel's degradation issues.
why is 7800x3d missing from the charts I get the new chips are not x3d but you have intel on the charts it would be nice to include the best selling cpu of this generation for reference
Because he just established these charts just now and hasn't tested a whole lot of CPUs yet, because he has health issues and is in the middle of moving his whole studio...
i am run an intel but was wondering the same thing because when the 50s come out i want to build a AMD system never had one and my intel is not giving me any problems its a 14700k/ 4080 first system in 20 years
@@kyoudaiken he should have tested first
I think Jay is not very much into 7800x3d, he never emphasized it in his videos like GN and HUB. Would like to see it though.
running 5200mhz unknown timings is kinda criminal lol
5200MT/s @ 1.1V with 40 - 44 - 44 timings, who runs EXPO @1.1 v!
@@jvgoransson try to running 5600 1.1V is probably the cause of all his crashes lol.
he is always shooting his AMD systems in the foot with the RAM then complains it doesnt run well or whatever, when people been running 6000+mhz on AM5 for almost 1 year at this point, its all around the internet, not to mention all of his Hardware channel colleagues always have their AM5s running at least 6000mhz with semi-decent timings, not to mention Buildzoid gave settings for decent 6000mhz with a bit tighter timings that were compatible with almost everything shortley after launch, I would assume it works on 99.9% of the systems right now, Jay never cared to put them in even once
@@tsvetanbliznakov7694Yup, 100% accurate, that's why you watch this guy for entertainment only and not for hardware tweaking advice lmao💀
Well, this does seem to be a 9000 series related issue.
7000 series runs 6000MHz fine, but a lot of tech channels seem to be having issues getting the 9000 series to run stable with current BIOS revisions available on some boards. 🤔
You got curve optimizer completely wrong. The number you enter in the curve optimizer is called count and 1 count correspond to 3 to 5 mv.
So -33 CO = -99mv to -165mv which would be great if that worked but most peeps won't even get that on there best cores lol
Start at -15 CO(-45mv to -75mv)
The BIOS has shown it in mV in its presets, NOT counts, and Ryzen Master "tested" -33 counts to be "stable".
Depends on motherboard, yeah?
@@kyoudaiken Look up "Precision Boost Overdrive 2" to understand it better. Ryzen master is awful to test if stable.. Better to use Prime95 or OCCT.
@@Haydatsanime It should all be the same.. only diff is Ryzen 5000 chips are capped at -30/+30 and Ryzen 7000/9000 are capped at -50/+50
Either you or I got curve optimizer completely wrong. I was under the impression that your best cores ran closer to zero, as they are the cores doing the very heavy and very light workloads. They require more volts to improve stability and to hit higher boost. For example, my 58x3d has -15 and -17 on the 2 best cores, and the other cores all in the mid to high -20's. But this...
"-99mv to -165mv which would be great if that worked but most peeps won't even get that on there best cores lol"
...reads like you believe the opposite to me?
MSI just did a bios update claiming a 10% gain with 9000 through memory tuning.
never gonna see this in this channel, he is reluctant to deal with any kind of Ram set up, if it doesn't work he just puts something lower until it works blindly, its super dissapointing, like every other channel and person who has a bit of hardware knowledge is running Ryzens past 6000mhz easily, this is not 2022 anymore
I heavily doubt the gain is that large. Most people aren’t even using high quality DDR5 especially on AMD platforms.
@@tsvetanbliznakov76947600x here running 6400mhz no issues
The difference in price between DDR5 5200 and DDR5 6000 is negligible at best. High quality price vs. whatever anyone else buys is almost non-existent in that speed range. The fact that Jay selected 5200 instead of the memory's rated speed of 6000 seems odd.
@@ZackSNetworkthis is an odd argument. Given Ryzen 7, 8, and 9000 series CPUs involve a sweet spot clock speed of 6000 MHz, why not push for 6000 MHz low latency? Is there something wrong with spending around $120 USD on a 2x16GB kit? We aren't buying 8000 MHz RAM.
Random black screen restarts happen because of light load instability imo. There is always a side effect of curve optimization for higher loads which is that it moves the entire curve including for low loads. It’s why their curve shaper feature might actually be pretty useful. It happens more often when you are doing ECLK overclocking.
Yup after playing around with CO on my 5950x for the last year, low load is almost always what shows instability for at least this gen of chip. I could game at -30 CO and +75mhz offset ALL DAY, but it'll crash within an hour of browsing the internet. At -30 CO and +50 offset, its rock solid.
I agree with this too, when I'm on the limit with my 7950X3d using curve optimizer (-20 iirc) it would only crash at low load. Someday I want to do a per core undervolt
Yep. I remember when setting up my 5900x that I often thought that I've got it right now, only for it to randomly crash while on desktop doing nothing much like a day or even a week later. Took me a while to get this one set right, but was very glad I did it per core because I've had some cores only take as little as -13, while others were fine with up to -27. Was a bit of work, though.
Looks like jay has not so much experience with CO
@@majestynlm1206 more like 0 and its very easy to see, idk how u justify not preparing for a video on a topic you know nothing about
Definitely sticking with my 5800X3D for another few years
i'm willing to bet that the 9800x3d will be massive improvement of performance. these zen5 are bandwith starved. i'm also willing to bet that the reason they went with the 65W is to shift the attention to the power limits and not the bandwith limits. Waiting for Wendell or phoronix to test how much faster the Epyc version of zen5 with 8 channel bandwith is going to be.
To be fair, anyone who wants to move that fast from ANY X3D chip soo fast, just have to much money and nothing else
cuz on performance point of view, it js just pointless to move from it if you have it
UNLESS you really want to have ddr5 machine
me too, plug and play. amd should stick with what actually works.
I am updating to the 9800x3D from the 5800x3D
@@Karti200 the 7800x3d is substantially faster than the series 5x3d you have anything above a 4070 ti
PBO per core optimizer on a 32-core TR took 5 hours. Just fyi for anyone wondering. I didnt keep the settings afterwards, i thought they were trash.
And 5 hours still tells you basically nothing about single-core stability. 😁 Just that it won't immediately crash.
Asrocks limit is 85c on 7000 series as well, just change the platform limit to 95c and it'll override asrocks limit
@8:00
Check your ram multiplier when over clocking it. Good Tip.
In the past, ive gotten new speeds to populate in the option table by changing the bin interval.
Thanks for the walk through. Don't understand though why you expect that undervolting the CPU, esp. with an aggressive offset of -35mV shouldn't lead to crashes. Finding a value where it runs stable is the point of tuning that value in, no?
The offset was not -35mV, but 35 units, so roughly 105-175mV.
One change made was location of internal temperature probe points...the result is a 7-9 degree drop in "reported" temp (under like for like temps) so dropping max temp is needed.
what do you mean by "reported" temps?? whats the source??
Classic AMD move. Make the CPU cooler by having the tprobe report a lower temperature, then lower tmax by that much.
@@RavTokomiSteve explained this in the gamers nexus review. They moved some probes around and it just happens to cause it to report lower. It's not a big thing but just know that's why the tj max has been lowered.
@@RavTokomi That's NOT why they did it, they did it to get a better safer boost curve algorithm and increase the overall efficiency
Enough with the Conspiracy Theory nonsense, we get enough of that BS from politics
@@RavTokomi Would that explain the TDP drops? Unless AMD has also changed how they calculated them, it would still require the same level of cooling if the temperature is actually the same, and only differs because they're measuring it differently
You could have done a power parity comparison with the 7700X (PPT: 142, TDC: 110, EDC: 170 on the 9700X). Another thing is you went too aggressive on curve optimizer, typically to get -30 to -45 CO you need to bump the LLC up one, unless you have a golden sample. You would have been better off test -20 which is more realistic. Also test with DDR-6400 since this will probably be the new standard on these chips, it seems like every one of them does it just fine with a 2133.33 FCLK.
OCCT was a lifesaver for me with my 7800x3D cpu I was able to realize that in the first 7 cores it supported -22 and only the eighth -20
What I like most is that you can configure it to stop the WHEA errors and tell you which thread failed and with CPUZ you can get the thread order list to which core it belongs to and know which one does not support the offset and lower it a bit. real lifesaver
i had a falcon NW system back in the early 2000's, it was a BEAST at the time with its overclocked q6600 and 4 gb ram and 8800gts (512 gb) and it lasted for almost a decade before i eventually went to a LGA 1366 setup, funny thing is i still have that 1366 system but but i really wish i had kept my old falcon instead these days
Went to Falcon Northwest, and my wallet ran away.
Even if it was to just keep that old case, which would’ve been awesome
These specs were my dream system back in the day. Had a E4400 and a 7600gs back then.
@@HesterDW Went to Falcon Northwest, and my wallet ran away
@@HesterDW well if you had the headroom you could of done a little overclocking on those, but i feel yeah the dual cores were nice compared to P4's but the q6600's felt like melted butter compared to the duos frozen margarine XD
Jay, I mine with my 7000 series and pbo plus CO is popular to keep clocks as high as possible while remaining efficient. All core PBO usually leaves performance on the table since theres usually a few weak cores. Best to do per core core offset which will require a lot more tweaking. For 7000 series, CO -30 was rare and usually -20 or worse was the norm depending on silicon lottery.
Would've been cool if you'd spent a bit longer getting curve optimizer to work - generally I think anything below -30mV is a stretch anyway.
yeah expecting -35mV to just work is pretty optimistic. At least for 7000 series chips that would be a notable silicon lottery win if that just works.
Yeah, it was too ambitious to go for -30 but to be fair finding the optimal undervolt pbo can take MONTHS.
I have had a 5800x undervolt around -25 -28(per core) for 8 months and I still get some crashes sometimes but it's quite rare, so I have to tweak my curve optimizer by time to time
@@dragoon0anime its easy to find if pbo is stable, use occt , cpu+ram and core cycler, if no errors there, then do all core test.
@@dagnisnierlins188 I did it, got some values (no more crashes using occt or core cycler)
A few months later my PC just crashed and reboot, so it wasn't stable.
It's just my personal experience, core cycler was really good to find the base per-core values to use.
Now I just use my computer if it crashes, check the logs and change the values in the BIOS.
Nowadays I rarely get any crashes but at least for me it took a lot of time and actually using my PC (gaming, watching videos, editing videos, etc...)
to get per core values where it doesn't/rarely crash.
@@dagnisnierlins188pro-overclocker has entered the chat 😂
The manner by which you talk about and comment upon data makes this video vastly richer and more informative than others.' Regardless of whether your testing methodology is or is not ideal by everybody else's standards. Chapeau! 👍
I wonder if these chips might be getting optimized for X3D? Maybe they want an efficient, stable base chip so they can crank X3D more?
If anything this video shows why I still prefer tinkering in the bios myself and finding what works best. Ryzen master is rubbish at finding stability
Isn't this just an issue because it's early for these CPUs now? Doesn't the optimizer work better on 7000 CPUs?
Now that I think of it, AMD has always been janky with some features...
@@kyoudaiken My experience trying out Ryzen master is that it's overly optimistic. If all cores are 100% stable with a -30 offset Ryzen master will try to go beyond. Whilst it may pass it's own tests you have a very high chance that it'll just turn off in other stress tests/games.
My 5900x was a brilliant undervolter. 200+ MHz and -30 offset all core in PBO, with a further -0.075v offset in the base mbo overclocking menu. Rick solid stable, scores over base pbo and runs very cool (I'm a silence freak so want fans going as slow as possible)
Everybody gangasta in the gaming charts until 7800X3D opens the door.
My i9 destroys 7800X3D...so no.
Your i9 is destroying itself, so no lmao @@battlephenom8508
@@battlephenom8508 Lol
@@battlephenom8508 You mean your i9 destroys ITSELF :D
@@battlephenom8508 begin selfdestruction TDP protocol.
I'm still rocking an AMD 8000 series. FX-8350 that is, lol. I also have a Xeon E3-1271 V3 (equivalent to a i7 4790). I don't do a ton of gaming, so it's not a big issue for me.
My hope is that I'll be able to pick up a decent AM4 system soon when everyone around me upgrades to AM5.
why are people mad that the cpu isn't pulling 1.21gw out of the box and isn't turning their pc into a goddamn fusion reactor?
or do people really want their future phones to have a lead acid car battery built in for it to last only 5h all while reaching temperatures that can melt steel in seconds?
it's both cooler/more efficient and better performing, why all the "ryzen 9000 sux" videos?
also weren't we laughing at intel/amd when their cpus were impossible to cool because they were drawing so much power? did we just do a 180 in that regard in just a few days for some reason?
Because people want more performance? Especially for the price of cpus
most people don't care about efficiency they care about numbers go brr...
and then they go and play 10 year old games lol
@@chrisk3127 OUCH! Your comment hurts. I bought a new PC in June and I am now playing FFXIV :P
@@chrisk3127 I just built a new 5700X3D system to primarily play modded skyrim on lmao
It’s like 5% faster for 20% more money than the 7000 series equivalent (current pricing). That’s why people are annoyed - it’s pretty simple tbh
The temperatures between 7000 series and 9000 series are not directly comparable because they moved the temperature sensor locations. All the boost algorithms have been adjusted for a new temperature curve. People obsess with temperature but what really matters is the Wattage which is the true test of how much energy is being used and the efficiency of the device.
100Watts @ 70C = 100 Watts @ 90C
Yes temperature (and voltage) is still import because if either is too high it can cause electromigration (Intel's current problem) and chip degradation but that has been accounted for in the new temperature curves
@11:15 - "This is kinda like intel territory on all-core by the way" - completely ignores that all of 7000-series boosted all core almost exactly as high as single core, and higher than the result here, 7700X hit 5.4 GHz all core all day. Also, (Auto) Curve Optimizer is undervolting the CPU / increasing frequency during the setup / testing - of course it is going to crash, why would you say it shouldn't? It's supposed to find the limit, even though the super quick testing can't catch every possible error and you should be doing this yourself. If you want to push single core with PBO on or Curve Optimizer on, you must set a boost clock override / "Auto OC" up to +200 MHz. A lot of CPUs are going to have some headroom there.
"7700X hit 5.4 GHz all core all day." A select handful did... or *yours* did. That doesn't mean it was 100% that did, or even the majority.
I've seen a total of 1, RUclipsr actually use PBO and overclock correctly.
@@tim3172 It depends on the workload, but something like Cinebench R23 would be above 5.3 GHz with PBO (~20K score) even on a bad CPU, assuming you have the thermal headroom (ie. 360mm AIO). Even if we assume 5.2 GHz (pretty much all 7700X at stock) you're missing the point of the comment - "This is kinda like intel territory on all-core by the way" is what he said when PBO was enabled on the 9700X and hitting 5.2 GHz in Cinebench.
Jay, what I’m more interested in is the comparison of 65w parts… why aren’t you comparing the 65w 9700x to the 65w 7700(non-x)? Let’s compare it to its own weight class please.
I think the PBO with just -20mV could have been interesting
I always enjoy giving Jay a watch even if the current topic isn't completely relevant to me. Keep rockin Jay!
Dude, Skaterbencher did just fine.
Skatter knows what he’s doing...Sry but sometimes jay just seems lost to be honest no hate 🤷🏽♂️
@@johnlong6197 it aint hate when its true
Thats because SkatterBencher is the OC God.
Just make the settings in the BIOS and you will not have any problems with resetting the parameters
That still doesn't negate the fact that it's not working correctly. Ryzen Master is supposed to write the values to the BIOS. Unless he turned that off.
Ryzen Master is not now and has really never been stable. Make your adjustments in the BIOS after figuring out your two best cores using Ryzen Master to test. If you want AMD to work, make the adjustments in the BIOS. And don't start with -40 or even -30 on Curve Optimizer. That works with some chips .. it does not with most.
Its a 65 watts part coz there was no peformance ulpift with more power and it wasnt an uplift from previous gen so they just said atleast make it 65 watts so that people dont get that dissapointed
Also people saying that this cpu is efficient by comparing 105watt 7700x to 65watt 9700x but unstead they should compare to 65watt 7700
I'm with you, but on that last part, I blame AMD - why name the successor of the 7700 as though it were the successor of 7700X, unless they wanted reviewers to compare the 9700X with the latter? They did the same thing when the 2700 successor became the 3700X rather than the 3700, but the generational jump was large enough then that it was still an upgrade from the 2700X.
Never bought an ASRock motherboard but I really like the Bios layout, very clean and precise.
I've used 4 of them over the last year, they've been rock solid for me so far 👍
In my opinion the efficiency is nice for SFF builds, but having the option to power it up will be nice for other users too. Options are always a good thing
u mean the efficiency that the 7700 non x already basically has (single digit % difference)
@@Frozoken Very true. I always forget about that non X version.
@@blainehock754 Also consider that for multi thread workloads this new CPUs do not seem to offer a real improvement over intel's 13th and 14th gens (due to the core count disadvantage). For instance if I lock my 12600k at 90W (and stock voltage) I get 16'211pt in R23 which is already quite close to the 16'900pt of the 9600x, thus 13th and 14th gen are gonna be more efficient at that TDP. Besides all the recent issues with intel cpus, I believe their bad reputation for efficiency is only Intel's fault, the architecture is actually decent, but they decided to push it to the brink of catching fire just for winning benchmarks, what a shame (and massife f***up).
Now we only need some decent SFF GPUs.
And intel 14thgen can be really efficient - not even hard to achieve, but the default settings were just atrocious and it finally boiled over.
But random crashes is a big red flag
I feel like amd expects users to have PBO enabled. But its still good to test both since one is technically overclocking. Also if you want to mess with curve optimizer again. Try up to -10max then add negative curve shaper on medium/high frequency. This will get the performance boost without messing with stability. Or at least its working so far for me.
C'omon do a research before, and learn how to do the things on bios side. I have a ASRock MB and they have a lot of options for manually adjusts. And yes, here the memory profiles don't work a content. I have to set the timmings and frequency manually. PBO -33? It's too much... Starts with -10,-20 and test stability.
Derbauer had some nice extra preformance with PBO. Had a few extra points in R23 multi with 50% power usage compared to your intel scores.
Only Jay would talk about turning on PBO with other Ryzen CPUs, then make the "AKTUALLY" mocking meme when someone points it out for the 9000 series
Yeah, I notice that too. What's up with that?
@@CallMeRabbitzUSVI he has 0 idea what he is doing, thats why he was reluctant to do it, very clueless attempt in the bios shows it very clearly
@@CallMeRabbitzUSVI "intel defenders"
@@tsvetanbliznakov7694 i've been saying this since the launch, none of them know what they are doing with PBO, let alone even notice their chips are only boosting to 4.4 ghz on Asus boards
*There is a 3-8% uplift in fps gains when using 8,000MHz ram, plus an additional 4% when tightening the timing, plus PBO, etc. The 9700X is actually a whole league ahead of the 7700X. It keeps up perfectly with the 7800X3D. And with using less electricity, and keeping cooler.*
*But only a couple of people went against AMD's stupid rule book.*
*7000 series support SPD 5,200MHz. 9000 series support SPD 5,600MHz. It is designed to run 6,800MHz to 8,000MHz ram and get better results.*
Hey there... I am still only on AM4 with a Ryzen 9 3900X with 2x32GB ram (tried 4x32GB but behaved weird)... Waiting on the test results on the new AM4 cpu's... I do not experiment with overclocking, because i want a rock solid system that just works... Keep up your great work, because it is very interesting... The best part is the tongue twisters that happens sometimes, because it happens for all of us...
It reverted to 95 because it detected the crash and reapplied defaults. Look at 9:50 in the video where it is also set to 95 after the crashes when you tried to set the voltage offset. You then lowered it from 95 to 90.
I don't know why it was 85 at the start of the video. I can only assume it got set there while you were trying it before recording.
No, the platform isn't perfect, but this was user error my friend. ;) You even mention in the video how it will go back to defaults on crash.
Defaults are 85 for this cpu....
@@Jayztwocents It's literally in the video, before you load profile 1 it displays 95. Applying a preset or auto OC then expecting perfect stability isn't always going to happen.
Lastly, cinebench is a terrible test and has nearly zero bearing on real world productivity.
@@Jayztwocents Perhaps nobody told Ryzen Master yet. 😅 Also weird it would choose an Expo profile as a "safe" post crash default. I'm curious how much of this is BIOS vs Master. My MSI X570 will go straight to JEDEC defaults when it detects crashes.
@@Jayztwocents "Max. Operating Temperature (Tjmax)
95°C" from the page for the 9700x on AMDs website...
@@BigMan7o0 As someone who just built my AM4 system last year long after AMD said the TJmax was 95C, the default set in the bios is 85. I noticed that when I was just running tests out of the box and I had to change it myself to get it above 85 to avoid throttling. This is the exact same things that AM5 will do, even if you update the bios.
Appreciate the video! I much prefer separate videos for OOB and "overclocked" performance. This way you can do a bit of a deeper dive into both subjects. Again, great video!
"Stop", "Collaborate and listen..." WTG Phil 🤣🤣🤘
Best part of the whole video!
That was 🤌
@@masterofnonetech absolute gold. Total JTC style!!
11:48 for anyone wanting to scroll to it
I can't believe I just got Iced by Phil.
I find it interesting that all the benchmarks showing big leads for Intel include a 14700k and 14900k while not including the 7900x and 7950x, which are both much closer in MT workloads and both cheaper. At the launch prices for the 9600x and 9700x, they certainly aren't good value though.
I have been running AMD since I bought my Ryzen 5 3600, it have never restarted for me. Didn't know that was a "thing" for AMD, currently I'm running a Ryzen 9 3950x because I found it for a good price and I like to play with VM's. But still havn't had any random restarts.
I do however had some random strange crashing. Removed one stick of RAM to run only 16Gb and it worked fine for month, regardless of which of the sticks of RAM I was running it worked fine. But as soon as I used both at the same time (no matter the slot) it randomly crashes again.
same ive had phenom, fx, ryzen 2000,ryzen 3000 and since 2022 i7 12700kf. Never had an AMD restart
but all AMD i've had default voltage by MB was too high and could be ajusted for more efficiency and less heating
I can go back as far AM386SX and that thing never crash on me … unless DOS/Windows wanted to. Doesn’t mean the newer ones won’t crash.
I’m currently using 3700x, still solid all year around.
5600X never had a single issue
sounds like a memory channel problem on the CPU itself.
I could be totally wrong but I think that was quite frequent on 3000 series
Ive never seen that stuff happen on my 3600 or 5800x, and im ptetty sure no ones going intel till they fix their shit. @George_Taylor_
should have prefered cores turned on. I just manually tuned curve optomiser on my 5900x and ended up with -30 on all. Should have tried -20
the problem is people are expecting an overall improvement but mainly the biggest improvements are within the AVX instruction sets, especially AVX-512. I'd be interested to see a comparison of CPUs using the PS3 emulator RPCS3
Zen 5 is actually Zen 4.2
PS3 emulation performance is just as disappointing as gaming performance on 9000 series in general sadly...
I'm not completely in the loop but from my use of RPCS3, I don't have a single-issue running any game at a locked 60 that I've tried. The only issue I seen people ever had performance-wise was God of War at 4K due to a memory bandwidth limitation on VRAM of some kind.
So from my understanding AVX-512 will help convert the PowerPC code to x86 more efficiently?
@@NewishJordan what cpu you using? I got a 11900k and it actually outperforms some 12th and 13th gen i9 in Rpsc3 cause of avx512
@@gex581990 Yeah after doing some research AVX512 helps a lot depending on where the bottleneck is for each game.(SPU or RSX)
Once you start pushing more than 60fps the bottleneck switches too. Seems like the 12900k with AVX is currently the best but the chart did not have Ryzen 9000 series yet.
I've seen several different reviews now and its all over the place. But not everyone is using the same memory or mb, so that makes it interesting. I'd like to see a little bit of a do over in say 3 weeks to a month because to be sure amd is gonna release an updated bios. Perhaps that will make memory more stable and we can get a better feel for these altogether as a group. How all 4 cpus stack together is what I curious about.
To AMD;
Thank you for listening to the enthusiast community. Fun overclocking has been missing from the high end desktops due to factory setting pushing chips for all they're worth. Thanks for putting it back in our han... fingers.
I was really happy to see you include Handbrake in your testing suite. That is what I really look at when considering a CPU upgrade. (I do a lot of video encoding)
Per AMD website: TjMax = 95c
RAM 2x1R = ddr5-5600
2x2R = ddr5-5600
4x1R = ddr5-3600
4x2R = ddr5-3600
I agree, for the out of the box experience they need to start testing these with the ram specs listed. I am pretty sure they can use the DDR5 6000 to void your warranty.
To be clear same with Intel
@@TheJmac82 Agree there. Not worth risking the RAM speed with Intel now they've increased Warranty period from 2 to 5 years. THAT alone brings them back into the equation again. The other 2 AMD CPUs with higher core count should be more appealling.
That's what AMD guarantee that will works out of the box. They say that 6000 MT/s (CL30) is the sweet spot. Intel's 13th and 14th gens are on paper also supported up to 5600MT/s, but you can enable XMP to get higher speed (your mileage may vary).
@@N3v3r_S3ttl3 That just further pushes my point though, so based on that if for what ever reason I happen to get one that cant run over 5600, i cannot RMA that but i will never get the speeds that are being reviewed. I really think it is deceptive. And where do they say that? I see tons of documents to the press saying that but no where have I seen AMD state to the public to up their ram speeds to 6000... why? because they dont want to be sued when there are issues. So with that its at minimum deceptive advertising.
Always like seeing the difference between PBO and Stock, would like to see 3d slicer performance. Obviously everyone uses different slicers but seeing the difference between old cpu and new cpu on Orca, Cura or Prusa slicer would be a huge help when deciding to upgrade.
level1tech quoted an AMD Rep saying they are stable and don't degrade at 95° 24/7
Welp, case closed then, who wants hot potato's?
Reps will say anything to sell you shit .
It's all relivent to me. Thanks for the data Jay. Love the Intel, AMD compitition. 😊
@0:05 that was good ... really good
Inknow cause he sound like that and look like that most of the time hahaha 😅
Im pretty sure just "PBO Enabled" in the BIOS means 95°C Tjmax, since its the default Tjmax they probably felt like mentioning it would be more confusing or something.
14700k is more expensive in germany 410 euros to 387 euros. wild that a 14th gen is still that expensive for everything happening
If you compare multicore performance the 9700x with pbo just barely beats the 13600k and 12700KF which cost 265€ and 190€. So, it's significantly too expensive. The 9700x should cost 300€ and the 9900x 400€. Currently the 7700X at 290€ is 10€ cheaper than the 9600x.
14700k is far better at multi core perf than all of amd CPUs ever made.
I would say the 7950x would still be faster it would also use less power.
9700x doesn't beat 13600k in multicore lol.
@@tmsphere 7800x3d shits on your degrading cpu "lol" and what now? i didnt even talked about 13600k :D
Didn't they move the core thermal sensors to a cooler spot? That was the gist of their "magic trick" to make 9000 series run cooler than 7000 series, from what I heard. Hence why default Tjmax is also lower.
the 9700x with pbo has the same performance like my 5900x with pbo , but 70w less consumption and 4 core less.
Of course less power if less cores.
Yes 5900x and 9700 x are equivalent in terms of performance .
I believe its 85 c because amd said they adjusted the locations of where they now get their temps from to more sensitive areas so overall the temps are lower for the same to more performance
As we showed though it kept changing, and sometimes shows max at 95 even though I set 90
@@Jayztwocents It's not 85 degrees C. AMD states the 9000 series has a 95 C TjMax on their own website / cpu spec page. So obviously it would (and should) show 95.
@@Jayztwocents maybe the bios isn't the best? I mean they also made new boards for these processors too. Maybe the new boards will do better with reporting?
I have never been able to get ryzen master to work for anything lol i do not know why i must go bios and set manually any thing
RM is a confusing mess.
Great video idea!
Very good and relvant tests.
I loved the idea to sort the chart big the price of the cpu
maybe the x870 boards will fix a few things. and yes i put my glasses on the say that :D
Cool vid as always. Be nice to see either the 7900/50 (or x variant) included in these charts, as they're closer to intel in terms of thread count - for the multi core stuff it'd make a lot of sense as a decent price competitor for people on those types of workload
Also, the reason the temp limit is lower is because they relocated the temp sensor so people stopped being so scared of the high temps
😂 righttt
Showing Core voltage while overlooking I found very helpful.
this is why Im still running a 3950x.... its never crashed or given me any problems... gonna just keep waiting till they get it right
im still with 3950x too and still amazing
My 3700X is my first solid CPU since my Athlon 4200+ X2.
1700: had 3 of them, twice replaced by AMD. It now crashes under full load out of warranty.
5900X: recently developed a "Cache hierarchy" BSOD problem.
*BIOS huh..?*
I still remember the hype when it came out. It was such a beast compared to any mainsteam desktop cpu at the time. Infact it was beating intel’s 2000 dollar 9980xe😅. It’s sad that amd didn’t increase the core counts to match intel, especially in the midrange.
The CPU is right, the marketing and the software on the other hand....
from my limited experience playing with PBO and CO with my 5800x, ive realized that CO helps single and all core workloads while PBO mostly only affects multi core perfeormance
88 watts! You're lucky you didn't wind up in the 1950s!
Seriously, thanks for the Science!
Great vid. As a car guy, I am pretty sure that you are more about performance than efficiency. I am the same way. Will be nice to get a more efficiency chip but is always a tertiary criteria for me when choosing a platform. I hope manufacturers just keeps things simple.
Love the RUclips troll voice in the beginning. 😂
Yeah of course I don't think he likes the teacher RUclipsr. He just mocks the dude each time lol
@28:32 chart looks like a long AC signal holding pulse signal inside it
I imagine the performance of any system would be affected by this, but especially faster clock speeds would see less power at these intervals!
Almost makes me wonder if different power supplies might provide more stable power? Or is this how the cpu is using the power?
This CPU is great for ITX builds
and a self build home server or nas
Just built one using 7600, it's very good
@@gamingwithegoon They idle at about 23,5W though. 500%+ higher power consumption vs about any intel just to run a calculator or to nothing. (And these are "efficient" by AMD standards.) They also use way more at simple tasks. Like watching youtube, etc..
@@TSracer1idle power usage only really matters for a server use case where its always powered on. I think power consumption under load is more important for consumers. Consumption under load effects heat output and noise where you probably wont notice a 20w idle vs a 2w idle but will notice a 50w load vs a 300w load.
You have many people are stuffing 13900k and 7950x3d in tiny cases like the NCASE T1 without issues. This CPU is a waste of time due to crappy performance gain from last gen.
Thank you for the fantastic video Jay!
I’ve been getting REAL tired of people saying “oh but it just needs to be juiced and it’s an amazing uplift” just because they completely ignored/misinterpreted most of a Der8auer video.
Happy to see PBO numbers. I feel like most people that are gonna watch a video on a CPU would probably turn PBO on or do some kind of tweaking anyways
Yes it's a brainer . Intel's processor have turbo boost which is directly enabled on the bios .
@@brasiliafc2138Turbo boost isn’t the same as PBO. AMD’s CPB is the same which they’ve also had for years. Intel doesn’t have anything similar to it since they left it up to the motherboard manufacturers with something called MCE.
@@brasiliafc2138 AMD CPUs also have a turbo boost which is usually on. So what's your point exactly?
PBO is for additional tweaking.
Why are you listening and not using ryzen master? I haven't herd the reason? Is it okay still to use for tuning then as long as we set the settings in the bios?
-30 CO equates to 90mv of 150mv less Vcore. -30 CO offset does not mean -30mv.
Do you mean the BIOS dropdown that literally is showing "-30mV" ?
@@ABaumstumpf Because everything wirtten is always 100% correct? It's like adressing you with Steve. And based on your logic it's written, so it can be wrong, ain't I correct Steve?
@@johnscaramis2515 "Because everything wirtten is always 100% correct?"
No, but there are 2 very different things - the BIOS-preset is applying a 30mV offset, where as the curve-editor is using 3mV steps but that one never showed 30 but 33.
So he is mixing up 2 things there and i was asking WHICH he meant.
On the other hand there are people like you......
@@ABaumstumpf I read a few comments from you, I can only say "on the other hand there are people like you"
Jay, do you get those random restarts if you don't use PBO? What settings make it totally stable?
I feel like the most important thing to come out of this video is that it's pretty clear the bios and software and so on are still pretty cooked for the 9000 series. I'd wait 2 months or so before getting one to let them sort this out. And make them cheaper!
To be honest, I have experience with AMD in retail back in the 90s and personal experience with the 1950X Threadripper, 3960X Threadripper and 5000G APUs on AM4. AMD always has been janky and weird and it didn't change to this day. You can wait years, I doubt it will get significantly better. The only thing you can do is cope with how the platforms of AMD are. Generally, AMD's monolithic Ryzen designs seem to be much more stable. These include the 5000G and 8000G series of desktop processors. I think they're still having massive issues with the chiplet design.
Awesome to see these follow ups. And you do it right Jay. I am in the camp that launch reviews, are out of box "open it and plug it in" reviews. The overclocking and configurability are features worth mentioning, but they are not the product being sold. Launch videos dont need to be a 1 hour deep dive covering out of box AND overclocking/PBO/tweaked settings. It becomes a mess.
As a 7700X and 7800X3D owner, I’m feeling pretty confident so far
As a 13900KF owner I'm feeling confident as well.
@@battlephenom8508 what a brave comment to make. Obviously you know your way around the BIO’s pretty well to set limits for a stable experience
pbo works fantastic for boosting single core performance but you need the temps to have it boost, did you put scalar x10 and plus 200mhz? and did you undervolt all cores and testing them one by one? don’t think so
Yea do this, that, that, and this and that, to get an increase lol, how about no.
none of those suggestions were the point of the video.
I assume the 7700x numbers are at stock right? How does 9700x PBO stack up against 7700x PBO?
7700 x get 21000 on average with pbo on .
I thought that said 9700x pro for a second
don't give AMD any ideas
Great video Jay :) and i'm one of those people who thought you were a "little" Intel bias
I saw a news article about a msi bios update for zen 5 that adds up to 10% more performance i feel as though the bios and software is just funky still.
honestly if they make the x3d chips spicy performance wise and power efficient I am sold. Power efficiency is good for me specially for the summer months where my room cooks. Coming from a 5800x these are very interesting specially once you look at the 12 and 16 core x3d parts. Power and efficient at the same time would be a dream come true
TLDR: I’m unsubscribing.
I’ve really enjoyed Jayz channel over the years. However, not providing a comprehensive/unbiased (or, honestly at the bare minimum, at least commenting on *potential* shortfalls in his analysis) review turns me away. Lampooning his audience in the first few seconds of his follow-up video does not engender trust or goodwill. He had some issues with one of his processor and wrote it off since Gamers Nexus had a similar problem. GN actually worked with an AMD engineer though; that doesn’t seem to be the situation here though. I don’t know if we will see responses to the comments on mV adjustments here, but I’ll try to find multiple other reviewers that have more experience with AMD’s hardware.
21:50 Amd will make a 2 chiplet 8x Zen 5 core and 16x Zen 5c core. They have not announced it. But if you have 2 chiplets. Why not do 1 Zen5 chiplet and 1 Zen5c chiplet. 48 threads.
AMD Power
If you are getting random restarts on AMD 7000 systems, go into BIOS and disable "global c-states". Doing this fixed the random restarts I had with windows 11 and have had no issues since - AMD 7950x/Asus ProArt X670e-creator/Corsair 2x48gb DDR5 5200mhz
Bro.. this cpu is for low consumption.. google and other companies will love them
This is the video i was waiting for!
it's easy to think AMD laid en Egg.......
I don't think they did though.....
Lets wait and see what the 9000 series 16 core parts look like.......
Intel ............................