- Jul 15, 2019
- 15
- 0
- 10
Hi,
I am a total amateur when it comes to overclocking. Hell, a month ago I didn't have the faintest idea how to build a computer. I spend about 12 hours a day in front of a screen working, and decided to replace my 6 year old ThinkCentre. I built a computer, and one of the big reasons was constant issues with a particular application I use. Unfortunately, I am still experiencing a good amount of hanging / freezing. It is a trading platform, and often operates fine, but during surges of volume it will often hang. This is a single threaded application. I do plenty of other stuff on this computer, and it seems to be handling everything else well. I have an i9 9980XE. So the intention of my overclock would be to increase single threaded application performance. I took a stab at it with Intel XTU last night. I understand it will be ideal to change in the BIOS but figured I would start there.
Adjusting the package turbo frequencies, 4.5GHZ on all cores did not result in a crash but ran at 94C in the benchmark. So I did not bother with a stress test. Bumping back down to 4.4GHZ on all cores, the benchmark temp was 63C. It made it through a 30min stress test, but with some thermal throttling. So it seems I don't have any more thermal headroom. Given my use case, I imagine that normal use would not see a heavy load on all cores, but would often push one core to it's max. So I have some ideas, and wanted to get some input.
1) XTU allows me to adjust the frequency based on active core count. I could attempt to lower the frequency for more active cores and increase it for 1 active core. I imagine that would help with my temps? However, I am also concerned that I will not see the performance boost if other applications are causing more than 1 core to be heavily utilized.
2) Would I see improvement by adjusting the per core values rather than the package values? I was thinking I could attempt to undervolt 17 of my cores and lower the speed, while seeing how fast I can get a single core by incrementally increasing the frequency and then voltage on that particular core. I could then use thread affinity to assign the application to that core. Would this give me more thermal headroom to push a single core faster than the 4.4GHZ I achieved on all cores? (On a side note, I am exploring a third party solution for a Windows version of the isolcpus to keep other processes off that core, or determining whether it would be stable to set Priority to Realtime for that application if it was pinned to a single core)
3) Will increasing AVX Ratio offset help me with more thermal headroom?
Thanks for any replies. Specs are below:
Motherboard: Asus Prime X299 Deluxe II
CPU: Intel i9-9980XE
Memory: Corsair Vengeance 4x16GB DDR4 3600MHz
Storage 1: Intel Optane 905P NVMe PCIe 960GB
Storage 2: Samsung 860 Pro 2TB
Graphics: AMD Radeon Pro WX 9100
Cooler: Corsair Hydro Series H100i Pro 120mm Liquid Cooling System
PSU: Corsair RM850i
I am a total amateur when it comes to overclocking. Hell, a month ago I didn't have the faintest idea how to build a computer. I spend about 12 hours a day in front of a screen working, and decided to replace my 6 year old ThinkCentre. I built a computer, and one of the big reasons was constant issues with a particular application I use. Unfortunately, I am still experiencing a good amount of hanging / freezing. It is a trading platform, and often operates fine, but during surges of volume it will often hang. This is a single threaded application. I do plenty of other stuff on this computer, and it seems to be handling everything else well. I have an i9 9980XE. So the intention of my overclock would be to increase single threaded application performance. I took a stab at it with Intel XTU last night. I understand it will be ideal to change in the BIOS but figured I would start there.
Adjusting the package turbo frequencies, 4.5GHZ on all cores did not result in a crash but ran at 94C in the benchmark. So I did not bother with a stress test. Bumping back down to 4.4GHZ on all cores, the benchmark temp was 63C. It made it through a 30min stress test, but with some thermal throttling. So it seems I don't have any more thermal headroom. Given my use case, I imagine that normal use would not see a heavy load on all cores, but would often push one core to it's max. So I have some ideas, and wanted to get some input.
1) XTU allows me to adjust the frequency based on active core count. I could attempt to lower the frequency for more active cores and increase it for 1 active core. I imagine that would help with my temps? However, I am also concerned that I will not see the performance boost if other applications are causing more than 1 core to be heavily utilized.
2) Would I see improvement by adjusting the per core values rather than the package values? I was thinking I could attempt to undervolt 17 of my cores and lower the speed, while seeing how fast I can get a single core by incrementally increasing the frequency and then voltage on that particular core. I could then use thread affinity to assign the application to that core. Would this give me more thermal headroom to push a single core faster than the 4.4GHZ I achieved on all cores? (On a side note, I am exploring a third party solution for a Windows version of the isolcpus to keep other processes off that core, or determining whether it would be stable to set Priority to Realtime for that application if it was pinned to a single core)
3) Will increasing AVX Ratio offset help me with more thermal headroom?
Thanks for any replies. Specs are below:
Motherboard: Asus Prime X299 Deluxe II
CPU: Intel i9-9980XE
Memory: Corsair Vengeance 4x16GB DDR4 3600MHz
Storage 1: Intel Optane 905P NVMe PCIe 960GB
Storage 2: Samsung 860 Pro 2TB
Graphics: AMD Radeon Pro WX 9100
Cooler: Corsair Hydro Series H100i Pro 120mm Liquid Cooling System
PSU: Corsair RM850i
Facebook
Twitter
Instagram