Yeah it is funny I am still rocking an old ass 4770K Overclocked to 4.5 GHZ stable 24/7. I am even able to go up to 4.7 GHZ but 4.5 is fully stable with good temperatures and only 1.26 volts. I guess Intel might be superior guys?
I would really love to see that long term degradation testing, also (if you see this comment) then would you you recommend just undervolting a ryzen CPU right out of the box instead of turning on pbo?
Undervolting is probably the best option for tweaking. My advice is not to mess with anything outside of PBO. PBO is still running within spec, it just removes the power limit.
I'm giving my 3900x the full send. PBO on, voltage set to auto. Even if this cpu faces issues later on, the 4000 series CPUs will be out later this year and I'll pick one of those up.
I do have the Crosshair Hero 8 and 1.4v is TOO HIGH, even for a high-end system... I wonder if the latest BIOS will set these voltages to a more reasonable level... 1.47v is too aggressive
The issue is you can validate an o/c but only for the short term, the threshold of stability is changing because of the high volts used to achieve it. In the past you tested a new CPU, backed off by 1 step on both V and frequency and a stable system tuned to it's CPU resulted, which (usually) ran problem free for years until it caught fire due to dust build up. Your time is better spent playing Doom Eternal rather than doing experiments on your CPU which are likely to screw up the careful factory binning and dynamic CPU management
@@RobBCactive just saying, ive ran my 1700 at 4ghz on a b350-f since the week after ryzen launched and i still have the same overclock set today, at the same voltage (1.4-1.43v) and ive had no degradation
Jay, I just want to say thanks as your channel has helped me get through this pandemic. The chill music, your casual tone, I learn new interesting things every day. Thanks man.
When I built a ryzen pc for a friend of mine, at first I tried manually overclocking but then also decided to turn on the motherboard's auto boost because it was just a better option.
I just let PBO enabled (5950x) with a good cooler, and monitoring sotware shows me 4 cores over 5K Mgh and most others ranging 4.5 to 4.9 .. I don't think I need to push all that further and damage anything too : I decided to quit learning about OC that specific cpu/ram/mb alliance.. fair enough power pseudo stock.
J I have 3 stations with Zen cpus in and like you say they are full sending the voltage on stock settings. I would love to know if my CPU are dying. Maybe run a Intel CPU next to it for a comparison? Thank You for the awesome content as always.
Because they can leave a rig sat somewhere for a year or so.......... Deg is vastly overtalked about and completely blown out of proportion by reddit experts. You won't deg a chip unless you put stupid voltage through it.
I'm personally more interested in stability over absolute performance these days, but it would be interesting to see how fast they degrade. I also would be interested to know if after the degradation was measured, a fresh coat of thermal paste, a fresh cooler and perhaps a fresh motherboard and memory would restore performance. Probably asking a lot...
I've got a scratch on my forehead that I don't remember getting, and while I have got 40 or so vehicles lying around outside, I don't remember working on any of them lately.
Yes, please make that video/experiment! Would be real interesting to actually get some "evidence" either way - and please do folding@home, because all that power will go to a good cause.
This is an excellent idea! :) Folding or Rosetta, both have CPU WUs. I think Folding@home has problems with larger thread count processors, though, so maybe Rosetta would be the way to go.
Buildzoid has already reported this behavior though it would be good to have a 2nd example or "confirmation on the behavior". Would also be good to ramp the voltage up slowly so we can tell at what point the voltage degradation is occurring. That would be super useful to the community.
22:06 - Long term degradation test would be great to see. Been siting on a 4790K with the hopes to jump to the next series of Ryzen this fall. But, as someone who tends to keep PC hardware for a long time, hence 4790K, this latest issues gave me some pause on Ryzen.
I think when running stock or just PBO and Auto OC, degradation will not really be a problem that will affect lifespan, just keep your bios up to date so that mobo manages power correctly.
Been running my 6700k at 4.7ghz 1.35v since day one....4 years later it still scoring the same in tests. Silicon degradation is so minor it only becomes an issue when you are doing extreme OC (for me that would be 4.9ghz at 1.4v and after that nothing will run.) Every cpu is different.
I think that the long term degradation project sounds good. When testing performance, you shoot for best case scenario e.g. CPU testing with a 2080ti on the test bench so this test should be shooting for worst case scenario. If you do it with Folding@Home then you can test with CPU pegged at 100% AND help towards a bigger cause... Win Win situation imo.
Would be really interesting to see the results of a long term test of degradation. 👍 I've just finished my 3950X build , and on the ROG Impact mobo the BIOS looked like it was feeding the cores 1.47V so I dropped it down to 1.35V(it's on water).........but in reality it wasn't. That setting in the BIOS actually just gives that voltage to the scheduler, the cores are only drawing off that one source so it's spread out quite thin really. I would like to see if the scheduler itself is able to take that voltage so could you really go deep on this one and do the test in increments. 🙏
I think that the long term degradation project sounds good. When testing performance, you shoot for best case scenario e.g. CPU testing with a 2080ti on the test bench so this test should be shooting for worst case scenario. If you do it with Folding@Home then you can test with CPU pegged at 100% AND help towards a bigger cause... Win Win situation imo.
Would love to see it. Maybe compare the XT to an early 3000 chip, and for that matter maybe an intel chip like an 8700k that you don't use anymore. Which one lasts longer? Also, folding seems like a good use.
Dont waste your time on a revamped 3700x. It was made, the 3800xt, to raise the average sale price of amd across the board. If you already have a 3000 series, hard pass. If not, go with a 3600xt, or 3900xt. The 3800x should have never really existed.
I would like to see the long-term test and then at the end of it I’d like to see you through that processor in another motherboard to verify that it’s actually processor
22:30 I think for this kind of test you will need 2 set of test-benches 1 with a Ryzen 3000 XT running with stock settings and the other running with your OC settings.
3point1 I don’t think it would matter about cpu quality because you’re running one at stock and one overclocked, then comparing the changes in speed over time, therefore if the overclocked cpu decreases in speed more (has a greater delta) than the stock cpu (or vice versa) you have a conclusion
You do not need one as long as you document well enough the performance of the chip you start with. Like voltages Vs stable clock speed, performance at various clock speed etc. Doing these initial tests with a precise set of 2x8 decently timed ram. Then you take off those ram chips to preserve them and run the degradation load with a sacrificial ram kit. This way at the end you'll have the CPU as only variable.
@@dremoralord6730 This has been well established for a while now. Auto settings on zen2 will run ~1.45v during single core and very light workloads. It uses a very complex automatic algorithm that dials back voltage to around 1.2v-1.3v during all core workloads. You can even see it yourself using monitoring software. It will never run anywhere near 1.45v during an all core workload with auto settings, even with PBO settings maxed out. I've had zen2 since launch day and never once messed with the voltage and I haven't had any degradation. It's best to leave the cpu multiplier and voltages on auto. Manually tweaking RAM frequency and timings using DRAM calculator is a good way to squeeze out some extra performance. More gains to be had here than you would by manually OCing the CPU.
@@NigguhMake yep, since the BIOS is a light workload, it's entirely possible it runs at about 1.45V if you have PBO enabled. And yep, I'm pretty sure Jay did not notice that he left it on a static 1.46 when using ryzen master. With 7 cores at 3.7/3.9 and one at 4.7, it will never apply 1.46V on auto settings. I don't blame him, ryzen master will not warn you. It will only warn you if you do not include the manual voltage setting when setting core clocks manually. There is a way to find safe voltages and that is by enabling PBO, run different worst case allcore workloads, and extrapolate a voltage/frequency curve. Also, it's entirely up to you but don't use DRAM calculator, take your time to manually tune everything over a month or so. I guess it could be somewhat useful for the most popular ICs (i.e. B-die), but even then it can't account for all the other parts involved (mobo, CPU clocks, IMC, Data Fabric) Jay, about the other cores compensating clock speeds as you do per-core overclocking in Ryzen Master, it's because of the clock dividers. How they work I have no idea. If you set clocks per-CCX, you won't have that problem of course.
Am I glad Jay always has a video with his solid advice on different topics. Good thing I search for a video from you before I went and overclock my Ryzen 9 5600x
He wanted the Fabric to be stable at 1:1 aka 1800Mhz. Prior to the XT series, unless you won the silicon lottery you could almost never get above 1900mhz and that would have been 1934."Most people" would be fine at DOCP and it sets memory timings as well so Jay would likely benefit in this case
ram speed does matter a little on AMD but ram speed matching the infinity fabric matters most. fastest STABLE IF speed is usually right around 1800mhz, so running that setting and ram at 3600 with some tighter timings will get you MUCH faster performance across the board
@@ghomerhust Ram speed absolutely matters a lot - up to a point. There are diminishing returns past a point... but more important than ram speed is ram timings, they are HUGE for ryzen and zen 2 specifically. 3600mhz CL14 (and just as important, optimized subtimings... 14-14-14-14-30-56 etc etc) is the sweet spot for zen 2 tuning. Why? because it gets you 95% of the maximum potential gains, without needing the very best most expensive exotic ram kits. Even CL15 or 16 is pretty great (at 3600-3800mhz). If you can do 3800mhz (1:1 with 1900 fclock) and keep the timings at 14, that's just about as good as it gets. DOCP settings are TRASH, and thus Jay's settings are trash. it wont matter for cinebench or most other all-core workloads, but for gaming it's significant. it doesn't matter if he runs 1:1 at 3600mhz, his timings are garbage... he lowered his ram speed without even tightening the timings... his DOCP timings where what... cl 18-19? makes me puke. he's leaving 5-10% gaming performance on the table. he doesn't know how to optimize/tweak zen 2 period, he's a novice... one level over a beginner. the average AMD subredditer knows more than Jay on this topic. he's flat out wrong about average all-core overclocks for zen 2, which have always been minimum 4.3, and mostly 4.4ghz. all these reviewers that stuck with their SUPER early silicon apparently never realized how quickly and drastically it improved, within months. my 3800x from feb 2020 hits 4.6 all core ~1.4v, and 4.5ghz all core ~ 1.35v. it can boot at 1.3 but it's not stable. but it could probably do 4.4 at 1.25v literally NO ONE is stuck at 4.1-4.2 like he says, except for trash early silicon or NON X series parts with garbage CCD's by design... they're the lowest bins. even the average 3700x, a mediocre bin, hits 4.4ghz all core... at least in 2020.
After leaving my 3800x stock for a while, recently I tried going manual. Running 4.4 ghz at 1.275v with cooler temps (~66c on a 10 min Cinebench R20 run vs. mid 70s [Corsair h115i]). All core score went up a few hundred points while single core is functionally the same (+/- 2-3 points).
@Warm Soft Kitty a person who runs a printing press, I started my apprenticeship in 1986 when I was 15, I used to print Letterheads (Before Email), Business Cards, Wedding Invites, Hospital Forms, Advertising Brochures, Annual reports, Raffle Tickets, all sorts of stuff
Here is jay on a 3800xt talking about how you can sometimes get 4.3ghz all core overclock. And here am I running 4.75ghz on my 5800x. Generational improvement!
If you have grey hairs like him or almost no hair anymore, tapping del will be hard to let go. Been tapping del to reach bios for, dare I say it, almost thirty years now ...
Best video in a long time, Jay. From any Tech Tuber. Thank you for going against public opinion which is based on artificial metrics like "price to performance" and telling it like it is with actual issues with a product. I would definitely watch 100% of your videos testing degradation on different scenarios over time for both AMD AND Intel platforms. Keep up the great work.
+1 on Ryzen 3xxx long-term analysis video. Whether VID from the cores or MoBos responding to it too aggressively, it DOES seem (anecdotally, at least) to be an issue.
I have no interest in trying to squeeze more performance out of my cpu, but I would certainly like to try undervolting to lower the risk of degradation.
Yeah probably, but Intel actually needs to overclock to touch Ryzen performance. Not to mention Ryzens are so so much more efficient. 45w compared to 95 and even 105w? Not even a question. Not to mention 7nm is king, and 14nm is old hat.
@@Bobdylan12121 I mean you are right but we cannot ignore that Intel can challenge AMD and even overcome it in some cases if over clocked to its limit at least Intel is trying to give competition
and another thing is Ryzen Cpu dont scale good in games with higher frequencies because their overall latency they scale in tile base rendering that's pretty much it.... that's the main problem with Ryzen 3000 series...
Still, my 3600 with a 2080 super is almost on par in performance with a 8700k. The only time you see a big difference is if you have say a 9900k with a 2080ti, and playing at 1080p with medium or low settings to really get the frames high. However if you are playing with graphic settings maxed or on ultra, the cheapo 3600 almost matches the flagship expensive Intel cpus, as well as the 3900x.
@@christophervanzetta but it's hardly enforced because it's basically impossible for AMD to tell if someone actually used it or not. Unless you're dumb enough to tell them.
Hey Jayz! I am building a new gaming pc with a 2080 ti GPU and I have a few questions and I would appreciate it if you or someone could help me choose! I am choosing from three graphics cards: MSI Gaming X TRIO, MSI Gaming Z TRIO, Asus ROG Strix. So if I am correct the MSI Z Trio is better than the X Trio. So I would either choose the Asus ROG or the Z Trio. I would love it if you could tell me which one is better and why... because I am very confused some people say the z trio has better fps and some say the Asus rog so yeah I don't know... If you could help out I would be very thankful! Thanks for your response in advance.
@JayzTwoCents Total power seems to be what kills Zen 2, not just voltage. By default, voltage drops for each additional core hitting maximum boost. More cores = more amps = more load on the power distribution part of the silicon, at least from what I have read about 7nm.
These are very interesting results. It begs a question on my 3900X overclock. I OC’d my 3900X to 4.2 GHz all core while setting my voltage to 1.25 V. I actually got my system stable to 1.225 V but nudged it up a little to have some headroom. With these settings I saw a good 300 to 400 point boost on Cinebench R20 while load temps went down by almost 10°C. (Peaks around 70°C single test and around 75°C on a 10 minute sustained test. Running a Corsair H100i AIO) Am I right in saying that since I am under volting my CPU I won’t see the kind of degradation this video talks about? Or at least it won’t degrade as fast as this video suggests? I mostly use the extra power for video rendering. Also, as usual great video. Been watching your content for years now and it is always been informative as well as entertaining.
Yea. I'm wondering the same. I for example have a 3600x(stock cooler), CCX0 on 4.224Ghz, CCX1 on 4.175Ghz (effectively all core at 4.2Ghz) that was 3 hours stable in Aida64 at 1.275V. If I let ryzen do its thing i would see voltages of 1.3 and even abit more. So I'm basicly undervolting it aswell? However, it does stay at +/-1.275V at all times according to CPUz and HWinfo. Idle temperature are in 30-40°C's, prolonged load will go to 78 ish degrees. As it stays at +/-1.275V, is there a proper C-state option in the bios? Hope to get an answer from Jay :DD
@@klaasm950 This video is hilarious. AMD Boost that ruins everything with those voltages is why you don't use it. It pushes temps too high and you get damage indeed. Locking all cores and undervolting is much much safer and prolonged in usage terms. So you're doing very good. His entire explanation of 4.7 with fluctuations is exactly the problem the boost programs cause, voltage induced thermal issues. If you can find the stable lowest + ish voltage (for headroom) values you are in the sweet spot. Funny enough it seems Jay also doesn't know that these cpu's have very specific DDR4 speeds to sync up. There is actual loss when going faster than the synced speed, so seeing a blind 3600 means loss as well. So, as for degradation ... the biggest degradation comes from turning a system off (thermal expansion/shrinkage), the second one comes from voltage misusage (too high)... undervolting + not turning the system off extends the lifetime by a lot.
I've been confused as to how to get the most out of my system since I built it... Build: AMD 3700X, Asus TUF X570, Corsair MP600, 5600 XT (went cheap because upgrading to Ampere when released), EVGA 850W, Arctic Freezer 2 360, Lian Li 011-Dyanic, G. Skill 3600Mhz ddr4 32gb. I have 3 fans on bottom for intake, rad is side mounted with 6 fans for full push/pull, 2 fans on top for exhaust. My question is, Should I just enable PBO and set it to get the highest gains (Think you can tell it to go like 200% or mhz or idk) and would that be the best bet for me? I haven't done a build since AMD Phenom II 550 BE (Basically a 950). Any advice is much appreciated!
Dont touch the CPU, tune the RAM. If you can get 3733, or even 3800 with Infinityfabrickclock in sync and low latencies stable on the Memory, you will gain way more than anything else. The boost of those ryzen2s is way better, than anything you can achieve manually.
@@Gottwtf So ryzen chips are simply much worse without these tweaks. It seems ryzen is not as powerful as some people might say. So why even with these tweaks, it cannot beat a $400 Intel CPU like 10700K ? It should mean Ryzen without tweaks is way worse than a 10700K.
I can only give you my personal experience with my 3800X: Out of the Box with 3200RAM C20 Score: 4900 PBO and 3200 RAM: 4950 (insanely hot, 90C on Custom Water) syncing IF manually to 1600: 4850 (no idea why my score went down, but i could consistently reproduce this with different settings) All core frequency locked @ 4,3GHz with a VID of 1,26V an 3200RAM gave me the best score 5150 and only 65°C...
That's what you call cheap? That GPU isn't bad. It's a good mid range card. A lot of people would be happy to have. As it seems like you don't have financial problems and you are just waiting for the new gen I hope that you will sell that RX 5600XT cheaply on the used market for people who might need it😊
Yes, please do the long term test. I'm thinking of upgrading to a ryzen system soon. I currently do a lot of heavy calculations with financial data analysis, machine learning, etc. Often have folding ticking over in the background when just doing casual work. A test like this would be fantastic. Just doing folding at Max settings would be really interesting to see number of work units completed during this time, as well as any degredation.
10:20 So...i should be veeeery happy, that my R7 1700 is running at 4.15GHz right? Weeeellllll... (1.3625V, never above 68°C, 75€ 280mm AIO, Asrock B450 MPro4 & X470 Aorus Ultra Gaming)
dude the best I could get on a 1600x was 3.85 all core, it defaults to single core 4.15-ish. is that 4.15 all-core? if so, damn, talk about good silicon lottery
What voltage? Mine runs perfectly fine at 1.35v 3.8GHz and I feel like stressing it at 1.4v 3.9GHz. Might do 4.0 but that's probably 1.45v and is not a good idea for a 24/7 schedule.
@@yammarques depends heavily on the board. first gen ryzen had crap boards with crap VRMs that are no good for OC. second gen boards were a little better but ASRock and cheap ASUS stayed being crap for OC.
This is why I changed my attitude towards OS based overclocking tools. With my 3900x, I set the CPU to a lower voltage and 3.8GHz in the BIOS for lower power consumption and longevity. When I need more from the CPU - this is when I use Ryzen Master's profiles. Each profile plays with the voltages a little bit and brings up the ratio. I dunno - maybe I'm just retro ... but it's like a software turbo button. Works for me and I'm not overly concerned about the longevity with most tasks.
Steve was tinkering with the voltages on the 3600xt and dropping volts to 1.287 was the sweet spot. It punched out a lot more performance with such a low voltage, even with letting the other auto settings go to town.
I ran 3.8 on my Ryzen 5 1600 with a stock cooler for a while, then had to drop to 3.6 after some trouble-shooting Fast forward ~2 years after getting the CPU, I had trouble rendering videos on Premiere with stock voltage at 3.4GHz, because the CPU would overheat and blue-screen halfway. I am now running the stock 3.2 speed with an offset of -100MHz and it's incredibly stable, honestly better than when it came out of the box.
Went through all these exact issues when overclocking my 3600x/3800x/3900x, the individual ccx adjustment just isnt really individual ccx adjustment imo... Gave up on it and went into all core for most of them (mainly the watercoolled) I own around 12 Ryzen CPU's of all gens, all of them have been a nightmare to overclock. However, with an insane amount of patience and time you can eventually get some pretty decent improvements, however not worth it for most people.
I’ve managed to pull 4.4ghz all core at 1.28v bagging around 5250 on Cinebench. I’ll be honest, I’ve been doing oodles of research but I haven’t got the experience and knowledge to ensure I’ve done 100% the right thing for this. Intel was just far less complicated. 3800x ASUS viii crosshair hero Gskill 3600 running at 3800 with IF at 1900 and tighter timings than XMP (docp, whatever)
I guess what I’m saying is Everything I’ve read and watched contradicts what my results are, and I’ve verified my results so many times various ways. I just can’t help I feel I’ve done something wrong because it appears to be so good, but I’m not finding it.
As an hardware engineer for a renderfarm and consultant for a datacenter i can clarify a bit about the high voltage the CPU asks for on auto settings. (We're using Threadrippers 3960/3970 and i got a 3800XT for my own rig) . These voltages (up to 1.5v) are directly linked to the Amperage of the load your asking. The 1.3v limit rule for all-core OC is there for the highest loads, i.e. AVX2 small FFT, long sustained heavy render loads and such with amperage above 70A. Gaming and single core loads hoover around the 40-50A and down.. Thats when you'll see auto voltages above 1.4v and should be OK for the CPU. !!The important bit!!: The reason it is ok is the algorythm (as show here in the video) switches from core to core to spread the load and temperature and brings the other cores to sleep. When you overclock an allcore OC these algorythms stop working and thats when your OC gets dangerous even with lower amperage loads. Thats why your 3800XT @ 1.4v and 4600Mhz is degrading as we speak. No sleeping cores, no low voltages, no core switching.. 1.4v all the way.. try to load a sutained renderproject and your cpu will degrade every day.
Wow, over 4 GHz,,,I am impressed really. My old 10+ year old PC is using a AMD Phenom II X4 965 Processor × 4 at 3.4GHz on 3 cores with one core at 800 mhz. I love this PC of mine, how ever I can't wait till I can afford to get a new computer. Any type of upgrade will feel like going from 0-60 in 1.5 secs. I don't play games on this PC any more I am mainly console, but the work stuff I do is kinda starting to get a bit longer.
Jay, I hit the silicone lottery. I have a R5 3600 that can run 4.4 with temps around 50c gaming. Yet only get my wife's to 4.1. I only changed the clock speed and left the rest at auto. Plus XMP on my 3200 ram. MSI B450M Gaming Plus AMD R5 3600 Corsair H115 pro AIO front mounted Corsair Vengeance 32G @ 3200 MSI GTX1660 Gaming (vertical mount) WD 970 500G m.2 Samsung 860 1T SSD WD 1T black HDD Corsair 760T case 2 - 140mm front (plus 2 - 140mm on AIO for push pull) 3 - 120mm top exhaust 1 - 140mm rear exhaust Case fans are Thermaltake Pure Plus RGB
i did a 4.2 all core OC on my 3950x using a 3.5 year old Thermaltake Water3 240mm AIO with 2 Noctua NFP12s, it ran R20 repeatedly and stayed at about 75C, and that was on 1.01V stable. also, the auto overclock in RyzenMaster SHUT OFF HALF OF MY CORES and wouldnt let me get them back until i completely uninstalled the program. even going back to auto settings in Master, it still had me locked at 8 cores 16 threads. so i ditched it and went back to my BIOS controlled setup. without the manual 4.2 OC, i do see TWO cores regularly bumping against the 4.7 max for this chip, and most others bounce between 4.3 and 4.5. and this is on a cheap asus prime-p x570 board. very pleased with that. as far as long term testing, you could always do "folding at home" on team 231300 (barnacles' team). i do folding with my 3950x when im not using it, goes for a good cause, you know, like saving lives via disease research and stuff
I'm convinced my 3600 has degraded just from running at stock because of the voltage, my system is pretty much always on and I've had it from new since Feb. it would sometimes hit 4.2 on a couple of cores so I just figured I'd leave it be as I have a decent cooler for it (h100i v2 in push pull with thermal grizzly kryonaut) and noticed a few weeks ago that it will no longer hit 4ghz. So I went into bios and dropped the voltage to 1.25 and pinned all cores to 4.2, runs way cooler now and getting better cinebench scores than when I first got it, fingers crossed all is good now.
My 3800x is at 4.5 all core with the 200mhz pbo enabled, and smt off, when i run smt on i lose 100mhz so 44x +200pbo . Runs great in both configs at safe voltage no stability issues. I actually have a couple of the HWbot top spots for different ram and cpu stuff on the 3800x so im very happy with it. Every video ive seen about 3800xt and whatnot kindve just shows a very minimal improvment at best. Just to Clarify what Jay is getting here with the 3800xt over my 3800x is a 100mhz improvement on pbo boost being 4.7ghz on the pbo cores
I suggest 2x AMD units for degradation testing, one as suggested (highest sable clock) but second one at lowest stable voltage (to gain highest clock) even down to 1.38 or 1.39, both running the same tasks. Reasoning: if degradation IS a real problem, then the comparison will also show that, and may show where it is degrading, either from high clock (both will degrade) or high voltge (one wil degrade) - thnx for the work done on this video, cheers
YES! Please make the long-term degradation video. Got my wife one of the Origin PC laptops you recently highlighted and I'll be acquiring her Ryzen 7 3700X, but there is possibly a significant upgrade coming in my future as well that would have me abandon that as my main workstation. I'd be very interested to see what happens to Zen 2 long-term under heavy load when overclocked because this machine would likely become a render server.
Instead of doing per core, you can do per ccx. You will be limited to the slowest core on each ccx, but the ccx's are binned differently. On my 3600, I have one ccx that does 4.2, but the other does 4.3 @1.3125v. the beauty is I can maintain these frequencies with lower temps than pbo +EDC bug can. On a 3800x, I'd expect 4.3 on one ccx and 4.4 or 4.5 on the other ccx. The results tend to be even better on the 4. Ccx chips and they see up to 4.6 on their ccx's
This is where you can overcome overvolting too! A manual OC locks the volts (for the most part) instead of letting the volts go crazy. Anyway, that's my argument next to my other comment here.
The best practice is to run a test like Prime95 for 10 minutes to see where the processor considers the all-core voltage a same limit, and to then not exceed that voltage in manual overclocks. It's typically around 1.3 volts, but can be higher with fewer cores. 1.45 volts on one core is fine since the current draw is much lower, but definitely aim for the Prime95 output voltage, with voltage droop enabled. For 1.3 volts, that will mean a setting of around 1.35 volts with medium droop levels.
My ryzen 9 3900x is showing 20 degrees high. It is actually 35-40 degrees idle, but it shows 55. Everyone solved this problem almost, the indicator problem. My problem is not getting better.
The long term degradation testing sounds like a really good idea!
Please do this!!!!
How would you even test that? Live it runnuning cinebench 24/7 for months? Yeah maybe like that.
^^^ jay !
Yeah it is funny I am still rocking an old ass 4770K Overclocked to 4.5 GHZ stable 24/7. I am even able to go up to 4.7 GHZ but 4.5 is fully stable with good temperatures and only 1.26 volts. I guess Intel might be superior guys?
I have a busted 9900k to confirm 🙄
I would really love to see that long term degradation testing, also (if you see this comment) then would you you recommend just undervolting a ryzen CPU right out of the box instead of turning on pbo?
Undervolting is probably the best option for tweaking. My advice is not to mess with anything outside of PBO. PBO is still running within spec, it just removes the power limit.
Undervolting is all good if power prices are high where you live, otherwise leave it stock and tweak RAM instead.
I'm giving my 3900x the full send. PBO on, voltage set to auto. Even if this cpu faces issues later on, the 4000 series CPUs will be out later this year and I'll pick one of those up.
I do have the Crosshair Hero 8 and 1.4v is TOO HIGH, even for a high-end system... I wonder if the latest BIOS will set these voltages to a more reasonable level... 1.47v is too aggressive
@@thestig007 Ah to be rich
"it's just not worth your time"
you underestimate how little my time is worth
same bundescock
The issue is you can validate an o/c but only for the short term, the threshold of stability is changing because of the high volts used to achieve it.
In the past you tested a new CPU, backed off by 1 step on both V and frequency and a stable system tuned to it's CPU resulted, which (usually) ran problem free for years until it caught fire due to dust build up.
Your time is better spent playing Doom Eternal rather than doing experiments on your CPU which are likely to screw up the careful factory binning and dynamic CPU management
@@RobBCactive just saying, ive ran my 1700 at 4ghz on a b350-f since the week after ryzen launched and i still have the same overclock set today, at the same voltage (1.4-1.43v) and ive had no degradation
Jay, I just want to say thanks as your channel has helped me get through this pandemic. The chill music, your casual tone, I learn new interesting things every day. Thanks man.
PLANDEMIC...
Scamdemic...
When I built a ryzen pc for a friend of mine, at first I tried manually overclocking but then also decided to turn on the motherboard's auto boost because it was just a better option.
I just let PBO enabled (5950x) with a good cooler, and monitoring sotware shows me 4 cores over 5K Mgh and most others ranging 4.5 to 4.9 .. I don't think I need to push all that further and damage anything too : I decided to quit learning about OC that specific cpu/ram/mb alliance.. fair enough power pseudo stock.
I would really like to see the long term test being the owner of an overclocked 3800X
J I have 3 stations with Zen cpus in and like you say they are full sending the voltage on stock settings. I would love to know if my CPU are dying. Maybe run a Intel CPU next to it for a comparison? Thank You for the awesome content as always.
ditto
Dito
What’s your overclock at? I’m at 4.3ghz all cores at 1.25volts. Runs really cool
same
I definitely want to see if the chip degrades. Folding at home seems like a good choice because your sacrifice will be for some good
Please do make the degradation study. It's interesting to know.
I second this.
@@OlaffLudwig I third this.
Yup, Jay please do it!
Plz
Because they can leave a rig sat somewhere for a year or so.......... Deg is vastly overtalked about and completely blown out of proportion by reddit experts. You won't deg a chip unless you put stupid voltage through it.
Jay, I think you're yanking my chain here with that Windows 95 wallpaper :-)
I'm personally more interested in stability over absolute performance these days, but it would be interesting to see how fast they degrade. I also would be interested to know if after the degradation was measured, a fresh coat of thermal paste, a fresh cooler and perhaps a fresh motherboard and memory would restore performance. Probably asking a lot...
Jay lies! He knows exactly how he got that scatch on his forehead, more 67 Camaro shenanigans! 😂😂
Its a deepfake, that's not Jay.
Revenge of the hammer!
He’s a shill for big bandaid
I've got a scratch on my forehead that I don't remember getting, and while I have got 40 or so vehicles lying around outside, I don't remember working on any of them lately.
@@HauntedCorpseGaming so you are saying linus took his channel over?
Yes, please make that video/experiment! Would be real interesting to actually get some "evidence" either way - and please do folding@home, because all that power will go to a good cause.
Agreed or Rosetta@home as it is more CPU intensive. Folding@home is more suited for GPU calculation.
This is an excellent idea! :) Folding or Rosetta, both have CPU WUs. I think Folding@home has problems with larger thread count processors, though, so maybe Rosetta would be the way to go.
Buildzoid has already reported this behavior though it would be good to have a 2nd example or "confirmation on the behavior". Would also be good to ramp the voltage up slowly so we can tell at what point the voltage degradation is occurring. That would be super useful to the community.
22:06 - Long term degradation test would be great to see. Been siting on a 4790K with the hopes to jump to the next series of Ryzen this fall. But, as someone who tends to keep PC hardware for a long time, hence 4790K, this latest issues gave me some pause on Ryzen.
I think when running stock or just PBO and Auto OC, degradation will not really be a problem that will affect lifespan, just keep your bios up to date so that mobo manages power correctly.
@@tortuga_za I mean why would anyone who wants to use hardware for a "long" time do that anyway when it provides auto OC itself?
I have the same plan.. im on a i7 3770
Been running my 6700k at 4.7ghz 1.35v since day one....4 years later it still scoring the same in tests. Silicon degradation is so minor it only becomes an issue when you are doing extreme OC (for me that would be 4.9ghz at 1.4v and after that nothing will run.) Every cpu is different.
I think that the long term degradation project sounds good. When testing performance, you shoot for best case scenario e.g. CPU testing with a 2080ti on the test bench so this test should be shooting for worst case scenario. If you do it with Folding@Home then you can test with CPU pegged at 100% AND help towards a bigger cause... Win Win situation imo.
Would be really interesting to see the results of a long term test of degradation. 👍
I've just finished my 3950X build , and on the ROG Impact mobo the BIOS looked like it was feeding the cores 1.47V so I dropped it down to 1.35V(it's on water).........but in reality it wasn't. That setting in the BIOS actually just gives that voltage to the scheduler, the cores are only drawing off that one source so it's spread out quite thin really.
I would like to see if the scheduler itself is able to take that voltage so could you really go deep on this one and do the test in increments. 🙏
I think that the long term degradation project sounds good. When testing performance, you shoot for best case scenario e.g. CPU testing with a 2080ti on the test bench so this test should be shooting for worst case scenario. If you do it with Folding@Home then you can test with CPU pegged at 100% AND help towards a bigger cause... Win Win situation imo.
Would love to see it. Maybe compare the XT to an early 3000 chip, and for that matter maybe an intel chip like an 8700k that you don't use anymore. Which one lasts longer?
Also, folding seems like a good use.
Like that windows 95 background, imagine trying to do this on windows 95...
lmao!!
lol I saw it last video as well. I still have a Win95 system
Not gonna lie I was like “damn didn’t even think could run it at first”
Lol I thought he was from the thumbnail 😁
@Atlantean If you know how to code maybe not!
i definitely would like to see a long term result of this particular CPU!!!!!
I VOTE YES!
Dont waste your time on a revamped 3700x. It was made, the 3800xt, to raise the average sale price of amd across the board. If you already have a 3000 series, hard pass. If not, go with a 3600xt, or 3900xt. The 3800x should have never really existed.
It would be interesting to see tutorial on how you reached that 4.6GHz, would you make a video about it?
I would like to see the long-term test and then at the end of it I’d like to see you through that processor in another motherboard to verify that it’s actually processor
22:30 I think for this kind of test you will need 2 set of test-benches 1 with a Ryzen 3000 XT running with stock settings and the other running with your OC settings.
The problem from that though is you get different qualities of the same chip. Who's to say the best or worst chip will be the 1 he's overclocking?
3point1 that’s true but it’s still better than not having a control
3point1 I don’t think it would matter about cpu quality because you’re running one at stock and one overclocked, then comparing the changes in speed over time, therefore if the overclocked cpu decreases in speed more (has a greater delta) than the stock cpu (or vice versa) you have a conclusion
You do not need one as long as you document well enough the performance of the chip you start with. Like voltages Vs stable clock speed, performance at various clock speed etc. Doing these initial tests with a precise set of 2x8 decently timed ram. Then you take off those ram chips to preserve them and run the degradation load with a sacrificial ram kit. This way at the end you'll have the CPU as only variable.
9:02 its actually not on Auto, you *manually* set it to 1.4625v in the Manual OC tab
It's both. When he's in the BIOS you can see the 1.46v on the right hand side
@@dremoralord6730 This has been well established for a while now. Auto settings on zen2 will run ~1.45v during single core and very light workloads. It uses a very complex automatic algorithm that dials back voltage to around 1.2v-1.3v during all core workloads. You can even see it yourself using monitoring software. It will never run anywhere near 1.45v during an all core workload with auto settings, even with PBO settings maxed out. I've had zen2 since launch day and never once messed with the voltage and I haven't had any degradation. It's best to leave the cpu multiplier and voltages on auto. Manually tweaking RAM frequency and timings using DRAM calculator is a good way to squeeze out some extra performance. More gains to be had here than you would by manually OCing the CPU.
@@NigguhMake I forgot about that. Aida64 will show you that live. Auto or load line calibration will do that.
@@NigguhMake yep, since the BIOS is a light workload, it's entirely possible it runs at about 1.45V if you have PBO enabled. And yep, I'm pretty sure Jay did not notice that he left it on a static 1.46 when using ryzen master. With 7 cores at 3.7/3.9 and one at 4.7, it will never apply 1.46V on auto settings. I don't blame him, ryzen master will not warn you. It will only warn you if you do not include the manual voltage setting when setting core clocks manually. There is a way to find safe voltages and that is by enabling PBO, run different worst case allcore workloads, and extrapolate a voltage/frequency curve.
Also, it's entirely up to you but don't use DRAM calculator, take your time to manually tune everything over a month or so. I guess it could be somewhat useful for the most popular ICs (i.e. B-die), but even then it can't account for all the other parts involved (mobo, CPU clocks, IMC, Data Fabric)
Jay, about the other cores compensating clock speeds as you do per-core overclocking in Ryzen Master, it's because of the clock dividers. How they work I have no idea. If you set clocks per-CCX, you won't have that problem of course.
@@NigguhMake Que the "I was wrong" video at some point. As much as i like Jay sometimes he need to engage brain before opening his mouth..lol
Am I glad Jay always has a video with his solid advice on different topics. Good thing I search for a video from you before I went and overclock my Ryzen 9 5600x
Jay try turning off all core boost then manually over clocking.
I think you'll be pleasantly supprised
Love to see a longitudinal study of the overclocking on the Ryzen 3000's :) Be amazing to see what actually happens.
2:27 "Set it to DOCP and... call it a day"
Jay proceeds to override DOCP and downclock the memory to 3600 MHz
He wanted the Fabric to be stable at 1:1 aka 1800Mhz. Prior to the XT series, unless you won the silicon lottery you could almost never get above 1900mhz and that would have been 1934."Most people" would be fine at DOCP and it sets memory timings as well so Jay would likely benefit in this case
ram speed does matter a little on AMD but ram speed matching the infinity fabric matters most. fastest STABLE IF speed is usually right around 1800mhz, so running that setting and ram at 3600 with some tighter timings will get you MUCH faster performance across the board
@@blatcher2 I know exactly what he was doing.
blatcher2 XTs should run at 1900 IF stable Steve from GN was demonstrating this on his 3600XT OC.
@@ghomerhust Ram speed absolutely matters a lot - up to a point. There are diminishing returns past a point... but more important than ram speed is ram timings, they are HUGE for ryzen and zen 2 specifically. 3600mhz CL14 (and just as important, optimized subtimings... 14-14-14-14-30-56 etc etc) is the sweet spot for zen 2 tuning. Why? because it gets you 95% of the maximum potential gains, without needing the very best most expensive exotic ram kits. Even CL15 or 16 is pretty great (at 3600-3800mhz). If you can do 3800mhz (1:1 with 1900 fclock) and keep the timings at 14, that's just about as good as it gets.
DOCP settings are TRASH, and thus Jay's settings are trash. it wont matter for cinebench or most other all-core workloads, but for gaming it's significant. it doesn't matter if he runs 1:1 at 3600mhz, his timings are garbage... he lowered his ram speed without even tightening the timings... his DOCP timings where what... cl 18-19? makes me puke. he's leaving 5-10% gaming performance on the table.
he doesn't know how to optimize/tweak zen 2 period, he's a novice... one level over a beginner. the average AMD subredditer knows more than Jay on this topic.
he's flat out wrong about average all-core overclocks for zen 2, which have always been minimum 4.3, and mostly 4.4ghz. all these reviewers that stuck with their SUPER early silicon apparently never realized how quickly and drastically it improved, within months. my 3800x from feb 2020 hits 4.6 all core ~1.4v, and 4.5ghz all core ~ 1.35v. it can boot at 1.3 but it's not stable. but it could probably do 4.4 at 1.25v
literally NO ONE is stuck at 4.1-4.2 like he says, except for trash early silicon or NON X series parts with garbage CCD's by design... they're the lowest bins. even the average 3700x, a mediocre bin, hits 4.4ghz all core... at least in 2020.
jay was abducted by aliens in his sleep they gave him the scar after giving him an implant
...and then Scott Baio gave him pink-eye...
Hate it when they go through the forehead to plant an anal probe.
Hopefully the Alien didn't have Covid.
That explains why he can't speak properly then. 😲
Maybe hit will make him "digress" less.
I would absolutely love to see the long term effects of running it at those clocks!
Great idea running the chip on Folding@Home!
Also interesting video, thanks for making it
I set my 3700x to 1.25v 4.25Ghz all core, 65c max with Corsair H115i Platinum at cinebench R20.
Stock it's was 1.3v 4.0Ghz..
Managed 4.2ghz all core with 1.15v just hated the amount of heat from pbo this was on a 3600
@@cjttuapawa 5.3 ghz on 8 cores here on a NH15 aircooler.
I gave up using pbo cuz i kept getting crashes, i think i have a poorly binned 3900x, i have a noctua nh15 cooler and a high spec machine too.
After leaving my 3800x stock for a while, recently I tried going manual. Running 4.4 ghz at 1.275v with cooler temps (~66c on a 10 min Cinebench R20 run vs. mid 70s [Corsair h115i]). All core score went up a few hundred points while single core is functionally the same (+/- 2-3 points).
@@sambarnett8041 nice undervolt im running same speeds but at 1.375v
I usually don’t know I have cuts on my hands till I use hand sanitizer or cleaning products. Then I find out real quick.
"Owwww, WTF.... How did I get that" yea.... I've said that many many times...
Always the outside of the pinky finger.
I was a Printer for many years & would constantly get paper cuts without noticing until I washed my hands
@Warm Soft Kitty a person who runs a printing press, I started my apprenticeship in 1986 when I was 15, I used to print Letterheads (Before Email), Business Cards, Wedding Invites, Hospital Forms, Advertising Brochures, Annual reports, Raffle Tickets, all sorts of stuff
Windows 95. The Legend..
Here is jay on a 3800xt talking about how you can sometimes get 4.3ghz all core overclock. And here am I running 4.75ghz on my 5800x. Generational improvement!
Jay's usually my go-to channel for busy work background videos.
Please do the long term test, I'd love to see that!
thanks for your videos built my first computer last year thanks to you!!!
I really, really hope he does a Ryzen RAM OC video.
Would love to see an updated video for the 5000 series and curve optimizer undervolting
0:52 You no longer have to keep tapping on the DEL key to access the BIOS, holding it down will work. Found out recently :)
If you have grey hairs like him or almost no hair anymore, tapping del will be hard to let go. Been tapping del to reach bios for, dare I say it, almost thirty years now ...
yeah, I would like to see that long term stuff, man.
(says like Palpatine) ...DO IT!
Thanos : JUST DO IT
Best video in a long time, Jay. From any Tech Tuber.
Thank you for going against public opinion which is based on artificial metrics like "price to performance" and telling it like it is with actual issues with a product.
I would definitely watch 100% of your videos testing degradation on different scenarios over time for both AMD AND Intel platforms. Keep up the great work.
Yes i would love to see you make that video!
+1 on Ryzen 3xxx long-term analysis video. Whether VID from the cores or MoBos responding to it too aggressively, it DOES seem (anecdotally, at least) to be an issue.
Long term degradation testing all the way please. Thank you for the videos! They have helped me build my first PC!
What keyboard was that bc i like it and depending on price I might get it
I think that is the one he modded for the star wars build
He's the long lost Harry potter of pc's with that scratch on his head xD. He got struck by an unforgivable cabinet door.
Fascinating overclock results.
My 1600 was 4.05 with a bit of voltage and it went 4.1 with a lot of voltage but yeah, ended up at 3.9 because it was so much cooler.
3.9 is the sweet spot for the 1600.
I have no interest in trying to squeeze more performance out of my cpu, but I would certainly like to try undervolting to lower the risk of degradation.
Is Jay telling me that over clocking is much better on a Intel chip when compared to AMD and it yields better performance boost on Intel then AMD
Yeah probably, but Intel actually needs to overclock to touch Ryzen performance. Not to mention Ryzens are so so much more efficient. 45w compared to 95 and even 105w? Not even a question. Not to mention 7nm is king, and 14nm is old hat.
@@Bobdylan12121 But is 14nm++++++++ old hat?
@@Bobdylan12121 I mean you are right but we cannot ignore that Intel can challenge AMD and even overcome it in some cases if over clocked to its limit at least Intel is trying to give competition
Thanks for the video and preventing me from wasting my time or blowing up my cpu ,love your videos!!
Thanks very much for this overview of this process.
and another thing is Ryzen Cpu dont scale good in games with higher frequencies because their overall latency they scale in tile base rendering that's pretty much it.... that's the main problem with Ryzen 3000 series...
Still, my 3600 with a 2080 super is almost on par in performance with a 8700k. The only time you see a big difference is if you have say a 9900k with a 2080ti, and playing at 1080p with medium or low settings to really get the frames high. However if you are playing with graphic settings maxed or on ultra, the cheapo 3600 almost matches the flagship expensive Intel cpus, as well as the 3900x.
The 8 cores have dual 4 core ccx thats enough cores in 1 ccx to not cause issues benchmarks prove
MAKE IT!!! I'd love to see how degradation really is on AMD :)
“Use of the feature [Precision Boot Overdrive] invalidates the AMD product warranty”
Is that actually what they say?
@@gangigooga7710 Yes. Read their warranty information
@@christophervanzetta but it's hardly enforced because it's basically impossible for AMD to tell if someone actually used it or not.
Unless you're dumb enough to tell them.
@@fireonyxiaz That's a pretty nasty stab in the back, I thought that AMD was the 'customer friendly underdog' you just gonna do them dirty like that?
Hey Jayz! I am building a new gaming pc with a 2080 ti GPU and I have a few questions and I would appreciate it if you or someone could help me choose! I am choosing from three graphics cards: MSI Gaming X TRIO, MSI Gaming Z TRIO, Asus ROG Strix. So if I am correct the MSI Z Trio is better than the X Trio. So I would either choose the Asus ROG or the Z Trio. I would love it if you could tell me which one is better and why... because I am very confused some people say the z trio has better fps and some say the Asus rog so yeah I don't know... If you could help out I would be very thankful! Thanks for your response in advance.
@JayzTwoCents Total power seems to be what kills Zen 2, not just voltage. By default, voltage drops for each additional core hitting maximum boost. More cores = more amps = more load on the power distribution part of the silicon, at least from what I have read about 7nm.
I believe the reason the cores are dropping is because there are power limits baked into it.
RUclips: There's 30 comments
Also RUclips: Shows One
I'm sure RUclips will correct it's ways now that you commented this
@@deminybs hehe i wish but that's how RUclips works :/
Multimillion dollar company guys
Yt broken everywhere :(
For me 1hour later youtube says 16 comments and there's hundreds.
24min08sec Video posted 2 mins ago, already 2 dislikes.. pple disliking this fast, are you ok ?
These are very interesting results. It begs a question on my 3900X overclock.
I OC’d my 3900X to 4.2 GHz all core while setting my voltage to 1.25 V. I actually got my system stable to 1.225 V but nudged it up a little to have some headroom.
With these settings I saw a good 300 to 400 point boost on Cinebench R20 while load temps went down by almost 10°C. (Peaks around 70°C single test and around 75°C on a 10 minute sustained test. Running a Corsair H100i AIO)
Am I right in saying that since I am under volting my CPU I won’t see the kind of degradation this video talks about? Or at least it won’t degrade as fast as this video suggests?
I mostly use the extra power for video rendering.
Also, as usual great video. Been watching your content for years now and it is always been informative as well as entertaining.
Yea. I'm wondering the same. I for example have a 3600x(stock cooler), CCX0 on 4.224Ghz, CCX1 on 4.175Ghz (effectively all core at 4.2Ghz) that was 3 hours stable in Aida64 at 1.275V. If I let ryzen do its thing i would see voltages of 1.3 and even abit more. So I'm basicly undervolting it aswell? However, it does stay at +/-1.275V at all times according to CPUz and HWinfo. Idle temperature are in 30-40°C's, prolonged load will go to 78 ish degrees. As it stays at +/-1.275V, is there a proper C-state option in the bios? Hope to get an answer from Jay :DD
@@klaasm950 This video is hilarious.
AMD Boost that ruins everything with those voltages is why you don't use it. It pushes temps too high and you get damage indeed. Locking all cores and undervolting is much much safer and prolonged in usage terms. So you're doing very good. His entire explanation of 4.7 with fluctuations is exactly the problem the boost programs cause, voltage induced thermal issues. If you can find the stable lowest + ish voltage (for headroom) values you are in the sweet spot.
Funny enough it seems Jay also doesn't know that these cpu's have very specific DDR4 speeds to sync up. There is actual loss when going faster than the synced speed, so seeing a blind 3600 means loss as well.
So, as for degradation ... the biggest degradation comes from turning a system off (thermal expansion/shrinkage), the second one comes from voltage misusage (too high)... undervolting + not turning the system off extends the lifetime by a lot.
Please do the long-term testing! Would love to see those results!
The scratch probably happened in your sleep
No he had visitor From Intel HQ because he's not doing enough to promote intel :)
"OC NOT worth it"
Jayztwocents
I've been confused as to how to get the most out of my system since I built it...
Build: AMD 3700X, Asus TUF X570, Corsair MP600, 5600 XT (went cheap because upgrading to Ampere when released), EVGA 850W, Arctic Freezer 2 360, Lian Li 011-Dyanic, G. Skill 3600Mhz ddr4 32gb.
I have 3 fans on bottom for intake, rad is side mounted with 6 fans for full push/pull, 2 fans on top for exhaust. My question is, Should I just enable PBO and set it to get the highest gains (Think you can tell it to go like 200% or mhz or idk) and would that be the best bet for me? I haven't done a build since AMD Phenom II 550 BE (Basically a 950). Any advice is much appreciated!
Dont touch the CPU, tune the RAM. If you can get 3733, or even 3800 with Infinityfabrickclock in sync and low latencies stable on the Memory, you will gain way more than anything else. The boost of those ryzen2s is way better, than anything you can achieve manually.
@@Gottwtf So ryzen chips are simply much worse without these tweaks. It seems ryzen is not as powerful as some people might say.
So why even with these tweaks, it cannot beat a $400 Intel CPU like 10700K ?
It should mean Ryzen without tweaks is way worse than a 10700K.
I can only give you my personal experience with my 3800X:
Out of the Box with 3200RAM C20 Score: 4900
PBO and 3200 RAM: 4950 (insanely hot, 90C on Custom Water)
syncing IF manually to 1600: 4850 (no idea why my score went down, but i could consistently reproduce this with different settings)
All core frequency locked @ 4,3GHz with a VID of 1,26V an 3200RAM gave me the best score 5150 and only 65°C...
That's what you call cheap?
That GPU isn't bad. It's a good mid range card.
A lot of people would be happy to have.
As it seems like you don't have financial problems and you are just waiting for the new gen I hope that you will sell that RX 5600XT cheaply on the used market for people who might need it😊
@GraveNoX Well, where i live, the 10700k costs 50% more than the 3800X with basically the same performance...
Yes, please do the long term test. I'm thinking of upgrading to a ryzen system soon. I currently do a lot of heavy calculations with financial data analysis, machine learning, etc. Often have folding ticking over in the background when just doing casual work. A test like this would be fantastic. Just doing folding at Max settings would be really interesting to see number of work units completed during this time, as well as any degredation.
I got my 3700x to 4375 on ccx0 and 4350 on ccx1 at 1.4v. Been running this setup since December of last year. No issues.
10:20
So...i should be veeeery happy, that my R7 1700 is running at 4.15GHz right?
Weeeellllll...
(1.3625V, never above 68°C, 75€ 280mm AIO, Asrock B450 MPro4 & X470 Aorus Ultra Gaming)
Yeah my Gf has a 1200 that I was able to get running at 4.225, but I’ve left it at 4.1 because I don’t want to risk it being unstable
dude the best I could get on a 1600x was 3.85 all core, it defaults to single core 4.15-ish. is that 4.15 all-core? if so, damn, talk about good silicon lottery
What voltage? Mine runs perfectly fine at 1.35v 3.8GHz and I feel like stressing it at 1.4v 3.9GHz. Might do 4.0 but that's probably 1.45v and is not a good idea for a 24/7 schedule.
@@yammarques depends heavily on the board. first gen ryzen had crap boards with crap VRMs that are no good for OC. second gen boards were a little better but ASRock and cheap ASUS stayed being crap for OC.
@@rawdez_ my cheap asrock lets me OC my 2600 to all core 4.2 so idk about the whole asrock being crap for OC
"Per clock overcore"
the last time i was this early she left me
Degradation testing would be a great video to see!
Long-term testing is a great idea and could help in future
"I have no idea how i got this scratch on my forehead", tell the wife to cut her toe nails lol
toenails? you mean fingernails right?
PapayaMan I think he meant what he said 😂😂
@@papayaman123 Does 69 sound familiar?
@@bukhosim2829 you're disgusting
@@iBoolGuy haha oh gosh
Me: *Thinks Im First*
But,246views 30comments
This is why I changed my attitude towards OS based overclocking tools. With my 3900x, I set the CPU to a lower voltage and 3.8GHz in the BIOS for lower power consumption and longevity. When I need more from the CPU - this is when I use Ryzen Master's profiles. Each profile plays with the voltages a little bit and brings up the ratio.
I dunno - maybe I'm just retro ... but it's like a software turbo button. Works for me and I'm not overly concerned about the longevity with most tasks.
Steve was tinkering with the voltages on the 3600xt and dropping volts to 1.287 was the sweet spot. It punched out a lot more performance with such a low voltage, even with letting the other auto settings go to town.
yes, please do a long term testing, sound like a great idea :)
I ran 3.8 on my Ryzen 5 1600 with a stock cooler for a while, then had to drop to 3.6 after some trouble-shooting
Fast forward ~2 years after getting the CPU, I had trouble rendering videos on Premiere with stock voltage at 3.4GHz, because the CPU would overheat and blue-screen halfway.
I am now running the stock 3.2 speed with an offset of -100MHz and it's incredibly stable, honestly better than when it came out of the box.
Went through all these exact issues when overclocking my 3600x/3800x/3900x, the individual ccx adjustment just isnt really individual ccx adjustment imo... Gave up on it and went into all core for most of them (mainly the watercoolled) I own around 12 Ryzen CPU's of all gens, all of them have been a nightmare to overclock. However, with an insane amount of patience and time you can eventually get some pretty decent improvements, however not worth it for most people.
I’ve managed to pull 4.4ghz all core at 1.28v bagging around 5250 on Cinebench. I’ll be honest, I’ve been doing oodles of research but I haven’t got the experience and knowledge to ensure I’ve done 100% the right thing for this. Intel was just far less complicated.
3800x
ASUS viii crosshair hero
Gskill 3600 running at 3800 with IF at 1900 and tighter timings than XMP (docp, whatever)
I guess what I’m saying is Everything I’ve read and watched contradicts what my results are, and I’ve verified my results so many times various ways. I just can’t help I feel I’ve done something wrong because it appears to be so good, but I’m not finding it.
As an hardware engineer for a renderfarm and consultant for a datacenter i can clarify a bit about the high voltage the CPU asks for on auto settings. (We're using Threadrippers 3960/3970 and i got a 3800XT for my own rig) . These voltages (up to 1.5v) are directly linked to the Amperage of the load your asking. The 1.3v limit rule for all-core OC is there for the highest loads, i.e. AVX2 small FFT, long sustained heavy render loads and such with amperage above 70A.
Gaming and single core loads hoover around the 40-50A and down.. Thats when you'll see auto voltages above 1.4v and should be OK for the CPU. !!The important bit!!: The reason it is ok is the algorythm (as show here in the video) switches from core to core to spread the load and temperature and brings the other cores to sleep.
When you overclock an allcore OC these algorythms stop working and thats when your OC gets dangerous even with lower amperage loads. Thats why your 3800XT @ 1.4v and 4600Mhz is degrading as we speak. No sleeping cores, no low voltages, no core switching.. 1.4v all the way.. try to load a sutained renderproject and your cpu will degrade every day.
Wow, over 4 GHz,,,I am impressed really. My old 10+ year old PC is using a AMD Phenom II X4 965 Processor × 4 at 3.4GHz on 3 cores with one core at 800 mhz. I love this PC of mine, how ever I can't wait till I can afford to get a new computer. Any type of upgrade will feel like going from 0-60 in 1.5 secs. I don't play games on this PC any more I am mainly console, but the work stuff I do is kinda starting to get a bit longer.
Jay, I hit the silicone lottery. I have a R5 3600 that can run 4.4 with temps around 50c gaming. Yet only get my wife's to 4.1. I only changed the clock speed and left the rest at auto. Plus XMP on my 3200 ram.
MSI B450M Gaming Plus
AMD R5 3600
Corsair H115 pro AIO front mounted
Corsair Vengeance 32G @ 3200
MSI GTX1660 Gaming (vertical mount)
WD 970 500G m.2
Samsung 860 1T SSD
WD 1T black HDD
Corsair 760T case
2 - 140mm front (plus 2 - 140mm on AIO for push pull)
3 - 120mm top exhaust
1 - 140mm rear exhaust
Case fans are Thermaltake Pure Plus RGB
i did a 4.2 all core OC on my 3950x using a 3.5 year old Thermaltake Water3 240mm AIO with 2 Noctua NFP12s, it ran R20 repeatedly and stayed at about 75C, and that was on 1.01V stable. also, the auto overclock in RyzenMaster SHUT OFF HALF OF MY CORES and wouldnt let me get them back until i completely uninstalled the program. even going back to auto settings in Master, it still had me locked at 8 cores 16 threads. so i ditched it and went back to my BIOS controlled setup. without the manual 4.2 OC, i do see TWO cores regularly bumping against the 4.7 max for this chip, and most others bounce between 4.3 and 4.5. and this is on a cheap asus prime-p x570 board. very pleased with that.
as far as long term testing, you could always do "folding at home" on team 231300 (barnacles' team). i do folding with my 3950x when im not using it, goes for a good cause, you know, like saving lives via disease research and stuff
I love when you deep dive software
Yes long term testing sounds great.
I'm convinced my 3600 has degraded just from running at stock because of the voltage, my system is pretty much always on and I've had it from new since Feb. it would sometimes hit 4.2 on a couple of cores so I just figured I'd leave it be as I have a decent cooler for it (h100i v2 in push pull with thermal grizzly kryonaut) and noticed a few weeks ago that it will no longer hit 4ghz. So I went into bios and dropped the voltage to 1.25 and pinned all cores to 4.2, runs way cooler now and getting better cinebench scores than when I first got it, fingers crossed all is good now.
I like how the focus is on the back of his neck's shirt, as opposed to his face.
My 3800x is at 4.5 all core with the 200mhz pbo enabled, and smt off, when i run smt on i lose 100mhz so 44x +200pbo . Runs great in both configs at safe voltage no stability issues. I actually have a couple of the HWbot top spots for different ram and cpu stuff on the 3800x so im very happy with it. Every video ive seen about 3800xt and whatnot kindve just shows a very minimal improvment at best. Just to Clarify what Jay is getting here with the 3800xt over my 3800x is a 100mhz improvement on pbo boost being 4.7ghz on the pbo cores
I managed 4.475ghz at 1.275v on my 3900x, and managed 8010 multi/510 single on Cinebench so pretty happy!
I suggest 2x AMD units for degradation testing, one as suggested (highest sable clock) but second one at lowest stable voltage (to gain highest clock) even down to 1.38 or 1.39, both running the same tasks. Reasoning: if degradation IS a real problem, then the comparison will also show that, and may show where it is degrading, either from high clock (both will degrade) or high voltge (one wil degrade) - thnx for the work done on this video, cheers
even Intel XTU would allow you to click on all core OC in windows, and add voltage.
YES! Please make the long-term degradation video. Got my wife one of the Origin PC laptops you recently highlighted and I'll be acquiring her Ryzen 7 3700X, but there is possibly a significant upgrade coming in my future as well that would have me abandon that as my main workstation. I'd be very interested to see what happens to Zen 2 long-term under heavy load when overclocked because this machine would likely become a render server.
Instead of doing per core, you can do per ccx. You will be limited to the slowest core on each ccx, but the ccx's are binned differently.
On my 3600, I have one ccx that does 4.2, but the other does 4.3 @1.3125v. the beauty is I can maintain these frequencies with lower temps than pbo +EDC bug can.
On a 3800x, I'd expect 4.3 on one ccx and 4.4 or 4.5 on the other ccx. The results tend to be even better on the 4. Ccx chips and they see up to 4.6 on their ccx's
This is where you can overcome overvolting too! A manual OC locks the volts (for the most part) instead of letting the volts go crazy. Anyway, that's my argument next to my other comment here.
The best practice is to run a test like Prime95 for 10 minutes to see where the processor considers the all-core voltage a same limit, and to then not exceed that voltage in manual overclocks. It's typically around 1.3 volts, but can be higher with fewer cores. 1.45 volts on one core is fine since the current draw is much lower, but definitely aim for the Prime95 output voltage, with voltage droop enabled. For 1.3 volts, that will mean a setting of around 1.35 volts with medium droop levels.
i love how he starts his videos
It has been my experience just set memory and CSM and you're good to go.
My ryzen 9 3900x is showing 20 degrees high. It is actually 35-40 degrees idle, but it shows 55. Everyone solved this problem almost, the indicator problem.
My problem is not getting better.