How Much FPS Can You ACTUALLY Gain From A PC Overclock?
HTML-код
- Опубликовано: 16 сен 2023
- Hello all! Today we are taking a look at what a full PC overclock actually gives in terms of FPS gains on a ryzen + nvidia system.
My Optimization Services:
www.beefytechoptimizing.org/
Full System Specs + Overclock Details
Windows 10 Revi OS
1440p 270HZ Acer Monitor (With Backlight Strobing)
RTX 4090 with 450Watt Power Limit
7950X3D (Prefer Cache enabled, 89C TjMax, Per Core Undervolt)
360MM Cooler Master AIO (6 fans push pull, Noctua NT-H2 thermal paste)
DDR5 6200MHZ CL30-38-38-30-68 (With tuned subtimings)
FCLK 2067MHZ : UCLK 3100 MHZ : MEM FREQ 3100 MHZ
X670E Asrock Steel Legend Motherboard
1350 Watt Thermaltake PSU
Cooler Master H500 Mesh Case (white)
3 Terabytes of NVME Storage
My WZ2 Config File
drive.google.com/drive/folder...
Follow my socials here:
Discord - / discord
Background Music (Royalty Free Music)
William Rosati - Head Candy 
Игры
Here is the thing. I’ve noticed some very inflated benchmark numbers over the years that can be manipulated to look like the hardware is performing well above expected, but then when your average Joe gets the exact same hardware, they can’t hit those numbers and get very frustrated. It can be a bit misleading. People don’t get the fact that there are little trivia here and there to use for certain benchmarks. Good video👍
Truthfully the people showing the benchmarks either don't show the whole picture, or the average Joe never sets up right the hardware.
Whatll often happen is that your casual gamer will miss a crucial step that may seem obvious, but it's not to someone who hasn't had to do it before.
On ryzen 7000x3d for example, dual ccd cpus often don't work at all without 5 different things all being set correctly.
Bios updated, chipset updated, game mode on, game bar on, and balanced power profile.
If 1 of those is wrong, the cpu will begin using wrong cores for wrong tasks, which is why process lasso is so clutch, even if you fuck up a crucial setting, lasso will force correct cores, making performance good, even if not perfect.
I had roughly 40fps gained in minimums on vondel just from tuning memory on my 7800x3d. 160fps dips to 200. My old 7700x had similar went from about 140fps dips to 180. Same thing on my 13700k. 160 dips to 200. Warzone eats up memory performance. Especially good latency.
Ull see varying results depending on map and system. These results were done a a 6200mhz m die kit, which is not as good as some of the a die kits out there. On the 6400Mhz cl28/2133mhz fclk kit I'm running on 7800x3d system I saw 10 fps higher than with this kit, so warzone scales well with better tunes.
You know you can move that sound card down to the low x16 slot. It'll still only run at x1. It doesn't need to be an x1 slot for a sound card. Any pcie slot will work fine. That way it won't block airflow to your gpu.
It doesn't fit cause of my motherboard IO sadly. The pin connectors block the sound card from slotting in at the bottom.
If the clock on your device is higher than the other devices, aren't you supposed to calculate frames from another device that operates with the same clocks as the other devices?
I tried the auto curve optimizer and then the pc became very unstable with crashes but all core oc was apperently much more stable. prob becuse the cpu undervolted to much
Would it make sense if I overclocked my 13600k and 7900xt
Beef good shit as usual. I'm just curious what are your 0.1 lows at stock on WZ?? 13900k DDR4 my 1% are the same as your non tuned 7950x and you average is 30 FPS higher in that area. Just curious how the 0.1 compare. Mine are 190.
Hey fam, thank you for the support!
Before I removed my A die ram kit from the 7950x3d system, I ran a final test in al mazrah using CAP X FRAME (an app that graphs out your performance) and I ran through al mazrah city for 2 minutes. Due to having that faster ram kit in at the time, I averaged slightly higher, at 313avg for the run, with 226 1% Lows, and 169FPS for the 0.1% lows. This means that yes, intel still has better 0.1% lows despite the lower averages. This will also means that even with lower averages the perceived smoothness will be better on intel in thoery.
I fully believe that if the 14900K will be capable of 6.2Ghz + ddr5 8400mhz-8600mhz with Overclocks, itll easily catch up in averages too! I have a buddy with a 13900KS + DDR5 8000Mhz and he gets only 5% less performance in averages than me, and 10% less than some of the best 7000x3d rigs out there (as there is better 7000x3d bins than mine).
How is 200 Hz extra for the CPU = 10%? It's it more like 4%?
Hello again. I also have a 7950X3D on an X670E Steel Legend motherboard with a 4090 Gamerock OC, and while watching your tutorials, I noticed that you don't set the RAM voltage above 1.4V on this configuration. Is this because you have Mdie memory modules, or do you think it's better not to exceed this threshold? In my configuration, I have TeamGroup T-Force Delta RGB RAM, DDR5, 32 GB, 7200MHz, CL34 (FF3D532G7200HC34ADC01) - as far as I know, these are Hynix Adie modules.
I've tested both settings from your videos - the first one with 6000MHz CL28 and all subtimings works perfectly, and the one with 6200MHz CL30 from that video also works. I'm wondering which one is better and whether increasing the MHz by 200 will make any noticeable difference, as the benchmark results in games like COD show the same performance (FHD - CPU Avg = 512 FPS, GPU Avg = 450 FPS / 1440p - CPU Avg = 494 FPS, GPU Avg = 369 FPS).
Is it worth increasing the voltage to 1.45V and trying 6200/6400 CL28/30, or is it better not to exceed this threshold, as the performance increase may be minimal?
In this video I used the M die kit at 6200 cl30 with 1.4v. The reason is because at 6200 cl30 m die is fine at 1.4 volts.
On my A die kit, I ran 6200 CL28 with far more tight timings at 1.55volts.
I have a fan on the ram to keep it cool mind you, but up to 1.8 volts won't damage ddr5, but I'd recommend not going above 1.6v for daily use on DDR5, and also using a dedicated ram fan.
Your A die kit can most likely easily do 6200mhz cl28 + tuned subtimings if you do 1.55vdd and 1.45vddq, so give that a shot if you haven't tried already.
@@Beefy-Tech Thank you very much for the very informative answer. I will definitely try to set it up as you described, and I'm also considering purchasing additional fans for my RAM. Does it help a lot? What fans do you recommend for ram?
@@BenchmarkingTom any fan you can point directly at the ram will help, but if you want to purchase a dedicated small ram fan, the only one available for sale new today is from corsair I believe.
It brings ram temps down by about 10 degrees, so definitely worth it if you are pushing higher voltages during ram OC.
@@BenchmarkingTomyou also want your fclk higher. try 2133-2167
how about on low bandwith gpu like 3050 or 4060 ? i really want to see that
Hi.
It's crazy how big the framerate deviation/dispersion is with the 7950x3d in Forza Horizon 5 on the benchmark graph. I've ran the same benchmark with the same settings, the main differences are that I'm on Windows 11 + HAGS enabled + nV driver 537.34 (setup: 13700k @stock, RTX 4090 @stock, 6400MT/s CL32 @XMP), and according to the benchmark graph both my CPU and my GPU are pumping out more constant/uniform values on the graph, tho less fps. Could it be that Intel is capable of producing a bit less fps but it's more stable/consistent, thus making it a smoother experience?
My CPU render results are: avg: 380, min: 311, max: 425
CPU simulation: avg: 418, min: 382, max: 450
GPU render results are the same as yours.
Average latency: 7.8
The "achieved" average that is in the pink/magenta box for me is 330 which is 24 fps worse than your stock setup (pre-OC), but the deviation/dispersion is much-much smaller. If you would like to, I can provide a screenshot about the FH5 results on Twitter. Even with the OC results, the devation/dispersion is so big.
What could be the cause of the huge framerate deviation/dispersion with AMD cpus?
The chiplet desing is the problem and the low fclk achive speed 2133 at max..
There were no actual "stutters" or massive FPS dips that would result in the game feeling anything but smooth. If you look at a frametime line during the test ull notice it remains flat the whole way through, meaning regardless of the graphs detailing the CPU performance on its own, the end gameplay was indeed perfectly smooth.
@@tacoterito22 also, this 7950x3d I have doesn't do more than 2067fclk stable, it's a bad bin.
@@Beefy-Techthe highest I’ve seen is 2200 & the guy was using 6400 cl26 🥲
Hey mate, more gains are made with gpu overclock on 4k. Or 1440p high (im assuming your all on low). You should find it higher than 5%. Also x3d chips are kinda useless to compare overclocks with, not huge scaling. More scaling with ram/cpu tune on non x3d chips amd and intel. Great video though.
It's true, I forgot to test 200% render res☠️, even intended too. But yeah, the ram doesn't scale basically at all. If I did just ram on its own, it's like +5FPS on M die generic tune, and +10 FPS 6400/2133mhz FCLK with CL28. Without the undervolt, ram on its own doesn't scale much if at all in cod.
I'm going to sell the 7950x3d though, as I have a terrible bin and it's just annoying to deal with over a 7800x3d🤣
Thanks for the comment;)
I was going to comment saying you guys should collaborate 🤩
Great video as always👌🏼👌🏼👌🏼what did u mean with the correct cores? Do u uncheck core 0? 6:40
He means the cores using the 3d vcache, and possibly his best cores as gaming only uses a few cores max. As the cores are separate groups 3d/reg Project Lasso is a program that you can setup so it forces only the optimum cores are used while gaming.
Ive had several instances of the core parking and core assigning just not working out of nowhere after it worked for weeks, so I now just play it safe and force affinity on cores instead of risking it assigning wrong cores lol.
yes, that means first 16 threads are forced to be used within gaming tasks in process lasso.
@@Beefy-Tech thanks for the fast response man 🌹 i played around with proces Lasso before. Somehow for cod.exe it doesnt let me set the priority to realtime or high. Idk if its the Game or my System. Did u try that for cod?
@@nucl3arboNg Ty 😊
+40W that's around 10% more power usage for barely 5% gains. Not worth it
hey man could you do this with your 13700k and ddr5 kit and 4090 defaut vs tuned ?
My 13700K will be part of my streaming PC with a generic 6000mhz cl36 kit I have so I wouldn't be able to show good results with it sadly, but I can still show the exact overclock settings for the CPU specifically.
@@Beefy-Tech okay yeah that would still be great content thanks
Hi, may I know the model of the ram cooler? Great vid!
Hey! The ram used in this video is a 6400Mhz CL34 32GB M die ram kit from kingston. The ram cooler is a very very old G skill dual fam cooler, which isn't available anymore today. If you want a ram cooler, the only readily available one will be from Corsair.
Hello, what is your Refresh Interval (tREFI), because I dont see it in Ryzen Master?
65535 (the maximum)
@@Beefy-Tech ok, thanks
@@Beefy-TechI can’t set anything about 47000 you’re lucky lol
@@silfrido1768 it's ram kit specific. I had some kits even hold 200k + on intel (intel allows much higher Trefi) but some other kits struggle above 40k due to heat.
@@Beefy-Tech yea my kit can’t do more then 4700. this is a 7200cl 34 kit that I clocked down & tuned to 6400 cl28-36-36-30-66 i could possibly lower voltage but it’s stable @ 1.49v.
Original value for tref was 46800
It’s also because I’m on amd with a 7800x3d.
can you do a video with intel 13th gen overclocking for call of duty
I'm using my 13700k for my upcoming streaming pc part swap, so once it's in there I can show CPU overclocks. The ram it's Paired with is cl36 6000mhz 32GB (which us insanely slow for intel) so it won't be scoring well as a result, but the cpu oc is fairly simple.
the real question is:
is it worth shortening the life of the components to gain nothing compared to a simply swap to more powerful component?
The real question is who told you that you are shortening the life of your components by undervolting the cpu and tuning the ram? An undervolt applies LESS voltage and heat for the same achieved frequency, and the RAM doesn't degrade from anything below 1.8 volts, which at the usual 1.4-1.5v that an OC uses, it'll never degrade.
Even with intel, where you do use more voltage, the potential degradation comes for people pushing to the utmost extreme limits...as a simple 200mhz + OC will probably make 0 difference to longevity.
The only time it's NOT worth doing is if you believe your own inexperience will damage the components through improper tuning, otherwise a free 30-40FPS never hurt anyone, even if it's not a huge difference.
@@Beefy-Tech no m8
Overclocking cant stay in pair with undervoltage
The PC just stall
When i pass from 1357 to 1411 MHz but dont move up mV....the pc stall (just because 50 mV)
And of course overclocking make components life shorter
You working above 100%
I have 1444 MHz RAM never touched
And CPU clocked at 3.6 GHz (overclock at 4.0)
I dont like to play with overclock
Ill just buy better components when ill upgrade the PC next year
Overclock is for who need powaaa! Xp
Someone who use engines or blender or davinci with 4k/8k
And of course you dont mount 1gpu, you use a PC far better then a gaming PC
@@RealNotallGaming that's alright fam, you know what they say, if someone's made up their mind, it's pointless to try convince them otherwise, so agree to disagree👍
@@Beefy-Tech 🙄
Bruh how are you getting 3ghz @ 865mv !? Lol
On my strix I can only put +120 but it’s not stable so I have to use +100 on the core.
For memory anything passed 500 is unstable
Update - I had to update my driver to the latest
Hey, what driver you have in the 4090?
This was on 537.13 I believe
@@Beefy-Tech thanks boss 🙌🙌🙌
The last minute of this video, I know which guy you talking about 😅
Lol. I'm not even referring in specific to someone tbh, just the entire scene.
Doing this financial impacts my own consulting business, but I'm fine with that because I want to change the way my business works anyway, essentially making consults obsolete, and providing ALL of the information for free, at the downside that it requires the person takes the time to learn how to apply it all.
I intend to keep consults up, but make them available as a last resort, or rather for those who just really can't be asked to learn how to apply them and have the spare money to throw at consulting.
Sadly, the people who NEED a consult, are generally also those who shouldn't be spending a big chunk of money just to get their pc working. So I'm trying to help those people out now.
I am guessing you turned off Core Isolation?
Core isolation has not been touched here no.
@@Beefy-Tech I just benched my RTX 4090 Suprim X + 7950X3D + G Skill DDR5 6000 CL30 (no PBO or curve optimizer) with 3DMark Time Spy. I only had a score of 25382, Graphics Score of 28762 and CPU Score of 15236. So pretty bad (at the lowest 6% of people with the same setup). I turned of Core Isolation and I got a score of 30825, Graphics Score 36506 and CPU Score 16381. A HUGE difference. I'm scoring better than the average for my set up now. I'm guessing most people turn Core Isolation off when submitting their 3D Mark score.
@@nossy232323 I've never ran a 3D mark test so I honestly don't know, but I'll try it out right now and get back to you with results from games👍
@@nossy232323update, my custom Revi OS windows install has no support for core isolation features. My assumption is that it's off by default as a result.
@@Beefy-Tech Yeah, that makes sense. Thanks for checking!
Yoooo🎉
🥨🙌