- Joined
- Jul 23, 2011
- Messages
- 1,586 (0.34/day)
- Location
- Kaunas, Lithuania
System Name | my box |
---|---|
Processor | AMD Ryzen 9 5950X |
Motherboard | ASRock Taichi x470 Ultimate |
Cooling | NZXT Kraken x72 |
Memory | 2×16GiB @ 3200MHz, some Corsair RGB led meme crap |
Video Card(s) | AMD [ASUS ROG STRIX] Radeon RX Vega64 [OC Edition] |
Storage | Samsung 970 Pro && 2× Seagate IronWolf Pro 4TB in Raid 1 |
Display(s) | Asus VG278H + Asus VH226H |
Case | Fractal Design Define R6 Black TG |
Audio Device(s) | Using optical S/PDIF output lol |
Power Supply | Corsair AX1200i |
Mouse | Razer Naga Epic |
Keyboard | Keychron Q1 |
Software | Funtoo Linux |
Benchmark Scores | 217634.24 BogoMIPS |
I was overclocking my Bulldozer and I hit a wall very quickly. No matter if I increased my Vcore over the default a little or a lot, it would fail the very same way.
Then I noticed a strange thing:
Whenever my CPU was under load the Vcore would drop, the more load on more cores it had, the more it would drop. With 2 Cores on full load drops by ~ 0.05 V; With 4 Cores - by 0.09 to 0.12 V (would start making occasional calculation errors); With all 8 cores on full load at once, I would see the Vcore to drop to <1.2 V almost in an instant. That is way below the voltage needed to run at such clock speeds so my computer would freeze in a second or two.
I thought it my be my PSU at first. I could see the PSU was running on less than third it's maximum capacity, thus barely loaded. And that PSU is quite new. So unlikely a PSU problem.
It seemed is as if something would try to enforce the TDP in any means available. And since I had all the power saving features turned off, it suspect it goes after my Vcore to to so.
Now it get to the weird part:
I was doing it on Windows. Almost solely because I wanted to use CPU-Z.
Now, I went back to Linux, where I usually "live". I can see the clockspeed of my CPU there, but can't monitor the Vcore. I kept the same clocks and voltages. I then went on to stress my CPU there. I was surprised - it ran rock-solid with all of my cores on full load. As if the CPU would get the voltage I intended.But running rock solid on Linux with what would would crash Windows in mere seconds? That is really odd.
I suspect it could be that Windows or it's chipset drivers are enforcing the TDP.
I also suspect one other thing: I am using the latest BIOS. And a couple of BIOS versions back, the changelog says "Add VRM MOS protection". Maybe the BIOS is F***ing with me by enforcing that TDP? I suppose Windows honor what BIOS decides, while the Linux kernel is known to sometimes not give a f*** over what BIOS says when it wants / knows better and overriding it's functions (thus explaining not enforcing this presumed limit).
Does that mean "Windows be bad for 'dem bulldozers & Piledrivers"? Since A BIOS update to support PD CPUs probably wont remove this.
Expert advice on what to do in such situation would be great.
Then I noticed a strange thing:
Whenever my CPU was under load the Vcore would drop, the more load on more cores it had, the more it would drop. With 2 Cores on full load drops by ~ 0.05 V; With 4 Cores - by 0.09 to 0.12 V (would start making occasional calculation errors); With all 8 cores on full load at once, I would see the Vcore to drop to <1.2 V almost in an instant. That is way below the voltage needed to run at such clock speeds so my computer would freeze in a second or two.
I thought it my be my PSU at first. I could see the PSU was running on less than third it's maximum capacity, thus barely loaded. And that PSU is quite new. So unlikely a PSU problem.
It seemed is as if something would try to enforce the TDP in any means available. And since I had all the power saving features turned off, it suspect it goes after my Vcore to to so.
Now it get to the weird part:
I was doing it on Windows. Almost solely because I wanted to use CPU-Z.
Now, I went back to Linux, where I usually "live". I can see the clockspeed of my CPU there, but can't monitor the Vcore. I kept the same clocks and voltages. I then went on to stress my CPU there. I was surprised - it ran rock-solid with all of my cores on full load. As if the CPU would get the voltage I intended.But running rock solid on Linux with what would would crash Windows in mere seconds? That is really odd.
I suspect it could be that Windows or it's chipset drivers are enforcing the TDP.
I also suspect one other thing: I am using the latest BIOS. And a couple of BIOS versions back, the changelog says "Add VRM MOS protection". Maybe the BIOS is F***ing with me by enforcing that TDP? I suppose Windows honor what BIOS decides, while the Linux kernel is known to sometimes not give a f*** over what BIOS says when it wants / knows better and overriding it's functions (thus explaining not enforcing this presumed limit).
Does that mean "Windows be bad for 'dem bulldozers & Piledrivers"? Since A BIOS update to support PD CPUs probably wont remove this.
Expert advice on what to do in such situation would be great.