-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
AMD 7950X ?
- Thread starter BruceUSA
- Start date
50/50 I would say ... 5.3GHz typically works at 1.2-1.3V, depending on luck, so 1.15V is like a cherry-picked sample (barely any reviews shows over 5.2GHz and less than 1.2V). Then it will be ~150W under load, so 137.8W is possible but depends on the workload. Max temp of 55.3°C is impossible on any popular cooling under full load. If someone has a temp like that, then probably also 137.5W wasn't under full load.
- Joined
- Aug 11, 2015
I went with the simplest method for overclocking my 7950X for my all-core BOINC computing. I tried all the other methods and nothing was stable. Very different from my 5950X.
So went with the AI Optimization method and just changed my cooler evaluation score to 200. So running 5.4Ghz on my good CCD0 die and only 5.175Ghz on my mediocre CCD1 die.
Using the development version of corefreq-cli, puts Vcore at 1.2875V and temps around 94° C. on CCD0 and 88° C. on CCD1.
So went with the AI Optimization method and just changed my cooler evaluation score to 200. So running 5.4Ghz on my good CCD0 die and only 5.175Ghz on my mediocre CCD1 die.
Using the development version of corefreq-cli, puts Vcore at 1.2875V and temps around 94° C. on CCD0 and 88° C. on CCD1.
- Joined
- Dec 19, 2012
Here's a couple of examples, both use a "running" voltage below 1.15 V and a speed of 5.3 GHz
Cinebench R23 easily runs like this and temps aren't bad 5.3/ 1.105V/ ~70°C/ 180 W CPU package
Then I tried Prime 95 small FFT, and had to up my negative offset a few times, but finally, it would run
5.3 GHz/ 1.14V/ 91°C/ 245 W CPU package.
So as Woomack said those numbers you are giving us are all quite relevant to the test being run. The voltage lines up but temperature and wattage are off IMO unless the test was very light work-wise.
E_D says he can easily run 5.4 GHz with a 7950X but that is still relative to the testing method. I can run all our tests at 5.5 GHz but there's no way I'd be able to run P95 small FFTs at that speed. After the SS I posted above at 5.3 GHz/1.14 V the temps were already creeping up to 95° C so were already maxed out.
Cinebench R23 easily runs like this and temps aren't bad 5.3/ 1.105V/ ~70°C/ 180 W CPU package
Then I tried Prime 95 small FFT, and had to up my negative offset a few times, but finally, it would run
5.3 GHz/ 1.14V/ 91°C/ 245 W CPU package.
So as Woomack said those numbers you are giving us are all quite relevant to the test being run. The voltage lines up but temperature and wattage are off IMO unless the test was very light work-wise.
E_D says he can easily run 5.4 GHz with a 7950X but that is still relative to the testing method. I can run all our tests at 5.5 GHz but there's no way I'd be able to run P95 small FFTs at that speed. After the SS I posted above at 5.3 GHz/1.14 V the temps were already creeping up to 95° C so were already maxed out.
My 7950X is like 5.4GHz 1.25V CCD0, and 5.3GHz 1.25V CCD1 or 5.4GHz 1.28V CCD1. Long high load passes only at 5.3GHz when all cores are set to the same frequency (at 5.4GHz it's or too hot or I see bluescreens). Full load at 1.25V = ~99-97°C on 360 AIO (ASUS or Deepcool AIOs).
I noticed that CPU voltages are slightly different on various motherboards. On Gigabyte my CPU needed slightly lower voltages, like 1.22V, while on ASUS about 1.25V to make the same clock.
Now check other reviews around the web and you see that in most of them, 5.2GHz is assumed as good OC at all cores set manually. 5.15-5.20GHz is about average that my CPU does at auto settings under full load what is much above the base frequency. As far as my CPU isn't any special, then looking at average results in reviews around the web, it still looks above average.
Btw. idle temps are close to 50°C with 360 AIOs. Mixed-load temps (something like PCMark 10 benchmark) are 95°C as at least one of the cores is bumping to the power/thermal limit. Max load is also 95°C, but at slightly lower frequencies if set at auto. All that looks weird but this is how these CPUs work at auto clock/voltage adjustments and with given power/thermal limits. I still feel it's a huge step forward looking at the last gen CPUs.
I noticed that CPU voltages are slightly different on various motherboards. On Gigabyte my CPU needed slightly lower voltages, like 1.22V, while on ASUS about 1.25V to make the same clock.
Now check other reviews around the web and you see that in most of them, 5.2GHz is assumed as good OC at all cores set manually. 5.15-5.20GHz is about average that my CPU does at auto settings under full load what is much above the base frequency. As far as my CPU isn't any special, then looking at average results in reviews around the web, it still looks above average.
Btw. idle temps are close to 50°C with 360 AIOs. Mixed-load temps (something like PCMark 10 benchmark) are 95°C as at least one of the cores is bumping to the power/thermal limit. Max load is also 95°C, but at slightly lower frequencies if set at auto. All that looks weird but this is how these CPUs work at auto clock/voltage adjustments and with given power/thermal limits. I still feel it's a huge step forward looking at the last gen CPUs.
Ive never broken them down by CCD, but my 7950X takes around 1.25-1.3V (depending on board - I saw the same thing Woomack has as well in my mobo testing) as well to reach 5.4 GHz all c/t.
Indeed. I can run 5.4 GHz against AIDA64 stress test (CPU/FPU/Cache/Memory, NO AVX 512). And will looop Cinebench for hours (figured that out by accident, lol).
I don't remember where I saw it but some guys on the web tested multiple Ryzen 7000 CPUs and typically CCD0 boosts 100-200MHz higher on the best cores than CCD1. It's like most cores are the same in CCD0 and CCD1, but usually some single cores are better in CCD0 and that's why we see CCD0 at 5.3-5.4GHz and CCD1 at 5.1-5.3GHz when we set manually all cores. The best way would be to set each core separately, but let's say I never wanted to waste time to test all 16 cores separately in my 7950X. Ryzen Master performs this test on all cores but it takes a lot of time (like 3h per pass), and the results are not perfect. Also, older BIOS had no option to set it separately for each core, so the only way was Ryzen Master and I dislike software overclocking. I think they added this option to ASUS and Gigabyte BIOS in the last releases.
One more thing. I don't know if you noticed but there is no new BIOS for X670E/B650E motherboards since about November. I think they can't improve anything as even initial BIOS for most motherboards was stable and with a memory clock limit they have nothing to improve. The next BIOS will be probably for CPUs with 3D cache support. On the other hand, most Z790 motherboards have no new BIOS too. I wonder if all are working on the next gen motherboards already or there is another reason.
One more thing. I don't know if you noticed but there is no new BIOS for X670E/B650E motherboards since about November. I think they can't improve anything as even initial BIOS for most motherboards was stable and with a memory clock limit they have nothing to improve. The next BIOS will be probably for CPUs with 3D cache support. On the other hand, most Z790 motherboards have no new BIOS too. I wonder if all are working on the next gen motherboards already or there is another reason.
FlanK3r
New Member
- Joined
- Jul 11, 2019
- Location
- Czech Republic
