Post your overclocks !

Is there a definite way of knowing which core is failing when a few P95 workers throw errors?
Edit: Think I can gather this from the cores that clock down in HWinfo, if my assumption is correct - pretty sure it is since I can see the effective clock falling on certain cores
 
Last edited:
roofrider said:
Is there a definite way of knowing which core is failing when a few P95 workers throw errors?
Edit: Think I can gather this from the cores that clock down in HWinfo, if my assumption is correct - pretty sure it is since I can see the effective clock falling on certain cores
Click to expand...
Yes. Cores that are being downclocked are the ones throwing errors as they fail to deliver payload.
This is mainly due to thermals which effects few cores more than others due to chip design and proximity to dispense heat.
 
Kratos47 said:
Yes. Cores that are being downclocked are the ones throwing errors as they fail to deliver payload.
This is mainly due to thermals which effects few cores more than others due to chip design and proximity to dispense heat.
Click to expand...
Thanks I was playing with curve optimizer. It was odd because CoreCycler did not give any error even after 12 hours but P95 was quick to fail, though both run P95 tests. I think it's because of the different modes. CoreCycler was in SSE while P95 was in AVX (2?) by default since I didn't disable AVX. I'm going to try CoreCycler with AVX today if my overnight SSE run didn't fail lol (shouldn't because it passed yesterday with even tighter offsets which I turned down a little because of P95).
 
Last edited:
Yes, AVX can make CPUs fart within an hour or two and is extremely brutal, especially AVX512 small FFTs (unrealistic PC scenario). I recommend using prime only on newly bought CPU running at stock to test out newly bought rig as a burn in test. OC profiles shouldn't be tested on Prime. RealBench, SuperPI is more suitable for real usage which stress the main CPU, Mem and their interconnect memory controller.

Higher temps of 95-100ºC+ for prolonged\overnight period isn't good for CPU cores and can permanently damage them, so it's good they're being throttling as a safety measure. Think Buildzoid even did a video on that. Also if using Ryzen 5000 with X570 MOBO, make sure you have the AMD AGESA 1.2.0.2 BIOS Firmware. I read a couple of reddit users reporting it reduced temps\better performance in stress tests\benchmarking.
 
roofrider said:
Thanks I was playing with curve optimizer. It was odd because CoreCycler did not give any error even after 12 hours but P95 was quick to fail, though both run P95 tests. I think it's because of the different modes. CoreCycler was in SSE while P95 was in AVX (2?) by default since I didn't disable AVX. I'm going to try CoreCycler with AVX today if my overnight SSE run didn't fail lol (shouldn't because it passed yesterday with even tighter offsets which I turned down a little because of P95).
Click to expand...
what are your system specs ?
 
And my overnight SSE CoreCycler run failed this time. I had set it to 1hr per core and Core 0 failed with 30 seconds left to reach that mark (The last time I did this was 30mins/core but 2 iterations so it probably didn't catch this error). Looks running long tests is the only way to truly weed out the errors.

I'll check out RealBench and SuperPI. I think AVX512 is disabled by default on P95 v30.3? Will need to double-check that. Mine is a 5600x on a B550 A Pro. Temps are under control, the only time it touched 90C was with PBO limits lifted and small FFTs on P95 I think. With offset curve it stays below that so I'm not even noting down the temps. Right now with CoreCycler it's 55C, but it's SSE and small ffts. I think it was 80C (Edit: 85C) with Prime, will check next time. But I have only run short Prime tests (20mins) till now.
I'm on AGESA 1.2.0.3b, I think the latest beta.
 
Last edited:
i've tried to over clock my ryzen 3 1200 today , it was decent until 3.8 ghz on 1.35 v
but ive seen a lot of forums that showed it goes till 4ghz on 1.35v , when i try it crashes

any fix?
 
BlueMoon said:
i've tried to over clock my ryzen 3 1200 today , it was decent until 3.8 ghz on 1.35 v
but ive seen a lot of forums that showed it goes till 4ghz on 1.35v , when i try it crashes

any fix?
Click to expand...
It is not that simple. Motherboard and RAM also contribute to a overclock failing.
 
BlueMoon said:
i've tried to over clock my ryzen 3 1200 today , it was decent until 3.8 ghz on 1.35 v
but ive seen a lot of forums that showed it goes till 4ghz on 1.35v , when i try it crashes

any fix?
Click to expand...
None. Use what is stable 24x7. No use hunting or better ram/mobo for more OC.
 
It depends on silicon lottery as well. 3 same model CPUs can overclock. like, 2 of them can overclock on 4GHz with different voltages. One on relatively small bump while the other requiring a lot. While the 3rd one refuses to go there despite pumping more voltage.
If you feel the CPU is lacking in performance then you can buy another AM4 compatible CPU which is more powerful instead of resorting to overclocking. AM4 socket is highly flexible.
 
Kratos47 said:
It depends on silicon lottery as well. 3 same model CPUs can overclock. like, 2 of them can overclock on 4GHz with different voltages. One on relatively small bump while the other requiring a lot. While the 3rd one refuses to go there despite pumping more voltage.
If you feel the CPU is lacking in performance then you can buy another AM4 compatible CPU which is more powerful instead of resorting to overclocking. AM4 socket is highly flexible.
Click to expand...
I won this lottery in 2007/8 when I got a E2140 that did 3.9 ghz! 3.6 Ghz 24x7 stable and another Q6600 at 3.9z ghz. Can't find screenshots of E2140 OC. It was an OC record (on air) on Anandtech IIRC.

1637601432614.png
 
BlueMoon said:
so it also depends on the freqnc that the ram is running on is t?
Click to expand...
when you overclock, FSB or front side bus frequency changes. in short the memory controller tells the RAM to run faster. Sometimes the memory can run, but sometimes it can't and that is when overclock fails. I may be wrong in explaining that.
 
I have Corsair Vengeance RAM 2 sticks of 8GB, both bought together and rated at 3000 Mhz but they never run at 3000
Anytime I enable XMP in BIOS, PC just refuses to start and then resets back to default 2400 Mhz
Anyone got any suggestion?
 
Back
Top