Message boards :
Number crunching :
What are your OC settings and config for an i9-9900KS?
Message board moderation
Author | Message |
---|---|
6BQ5 Send message Joined: 7 Dec 18 Posts: 29 Credit: 12,725,636 RAC: 357 |
I've been tweaking my i9-9900KS and would like to hear from others who have done the same. When it comes to overclocking and performance boosting I am far from being an expert. What are your performance characteristics and settings? I'm using CPUID HWMonitor to read the status of my processor. SETI@Home pushes my 8 cores to 4568 MHz - 4773 MHz depending on temperature spikes. The ~300 MHz drop from 5000 MHz seems normal if it is caused by the processor detecting AVX instructions being executed. Instantaneous temperatures bounces between 84 deg C and 92 deg C. My CPU Core Voltage is 1.280 Volts. Power consumption is reported at 180W. In my BIOS ... ... BCLK is set to 101.5625. ... CPU Multiplier is set to 50. ... Cache/Ring Multiplier is set to 47. ... CPU Voltage is set to 1.315V. On the one hand, I'm OK with the sustained performance but I can't even begin to imagine how some people go higher than this. Did I lose the silicon lottery? What do they cool their system with? I must have some humble cooling. 2x120mm fans on the bottom blowing upward. 2x140mm fans on the front blowing inward. 1x140mm fan on the rear blowing inward. 3x120mm fans on top w/radiator blowing upward like jet turbine. The rear fan blowing inward seems counter intuitive. I didn't have anything there initially. When I held my hand up to the rear vents I didn't feel any air flow. I put a piece of tissue paper up to the vents and it was sucked in. The 3x120mm fans up top must really create a vacuum! I can probably float a small ball on the air flow from 3x120mm fans. So, to help air that is already going in to go in easier I have the rear fan blowing inward. -=- Boris |
Tom M Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 |
While I can't help you on that specific CPU I can say that in general if you run your Boinc at 90% of the available CPU's freeing up at least 1 thread your total production should go up and your average CPU processing time should go down. This has been the experience of any Setizen who is running CPU tasks. For a dedicated GPU-only processing system it may not matter. Tom A proud member of the OFA (Old Farts Association). |
6BQ5 Send message Joined: 7 Dec 18 Posts: 29 Credit: 12,725,636 RAC: 357 |
Hi Tom, Thanks for the reply! Right now my computing preferences are set to use 100% of the CPUs for 100% of the time. Since I have 16 threads, would each thread get 6.25% of the CPU? If yes then should I set my computing preferences to use 93.75% of the CPUs available 100% of the time? Why would that lower the CPU processing time? It sounds counter intuitive to running everything at 100%. -=- Boris |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
Unless you've installed an anonymous platform and a select few apps, you are not running any AVX instructions that would cause the standard AVX instruction downclock penalty on Intel processors. Typically the highest SIMD instructions the stock apps run is SSSE3. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
Cruncher-American Send message Joined: 25 Mar 02 Posts: 1513 Credit: 370,893,186 RAC: 340 |
Running fewer threads will cut system thrashing IF you have overcommitted your CPU. And using 100% of the CPU all the time WILL overcommit, as you leave no threads available for system (as opposed to app - BOINC/SETI) use, so the OS, the SETI apps and BOINC will be fighting for the CPU. For example, I run 2 dual Xeon systems with 40 threads each. When I was using 95% (38) of them for SETI, I noticed a lot of red in Task Manager (Win 7) which means the system was thrashing, so I cut that to 90% (36). On both systems, the OS usage went down, and I actually got more work done. |
6BQ5 Send message Joined: 7 Dec 18 Posts: 29 Credit: 12,725,636 RAC: 357 |
Very cool! I opened up my Task Manager and from there opened up my Resource Manager. The CPU tab in my Resource Manager showed CPU utilization at 114%. My Computing Preferences had BOINC using 100% of the CPUs 100% of the time. Using only 80% of the CPUs 100 % of the time lowered my CPU utilization to 99%. Four tasks went from "Running" to "Waiting to Run". My CPU temperature now peaks less. It seems like the temperature runs between 84 deg C and 88 deg C with less jumps. The clock rate no longer bounces around either. It's pegged at 4773 MHz. Power consumption stabilized a bit too. It used to bounce between 170W and a whopping 190W. Now I am seeing a steady 176W to 179W. I still wish I I could see a steady 5 GHz like when I run an artificial CPU stress tester. -=- Boris |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
In all this you never told us what board you are using. I run an i7 9700 @5 Ghz on an ASUS ROG Strix board. In the BIOS there is a switch to run all cores at 5 GHz. I don't know what board you have so can't tell if you have that option or not for yours. I would say thou that your Temps, while acceptable, could be better. What kind of cooler do you have on it? I run AlphaCool 280mm. Never seen the temp go over 60C. |
Keith Myers Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 |
He needs to look for MCE or Multi-Core-Enhancement or what ASUS calls it in their BIOS or equivalent in his board. Set all the cores to a locked multiplier at whatever clock the cooling solution can support for sustained 24/7 loads. Seti@Home classic workunits:20,676 CPU time:74,226 hours A proud member of the OFA (Old Farts Association) |
6BQ5 Send message Joined: 7 Dec 18 Posts: 29 Credit: 12,725,636 RAC: 357 |
My motherboard is an ASRock Taichi Z370. The Multi-Core Enhancement was disabled when I fixed my CPU Multiplier to 50 and fixed my CPU Voltage. My cooler is a DeepCool Castle 360 RGB V2. I tuned my voltage by running Geekbench at different Voltage settings starting at 1.3V and bumping it up by 10 mV. My Geekbench numbers would rise and then dip. After seeing the dip I pulled back 10 mV. Oh, and I made a typo in my initial post. My CPU Core Voltage is actually 1.320V ... not, 1.315V. -=- Boris |
Lazydude Send message Joined: 17 Jan 01 Posts: 45 Credit: 96,158,001 RAC: 136 |
I suggest that you find max power you could use. My z370 mb could only supply the CPU with about 160w before mu I9-9900k (95W TDP) was starved of power. My new Z390 i have set power limit 1 to 220watts and yes its running hot with HT on and only 8 threads for SETI-cpus http://mikesworld.eu/download.html I am using Seti V8 CPU x64 3330 AVX it running about 90-105C. at 5100MHz On my card I can set Tjmax to 115C AVX offset are set to 0 V-core according to CPU-Z 1,43v I9-9900ks (127W TDP) I suggest that you test with Prime 95 with is the hardest test i know of. And have an eye on HWiFOS CPU power https://www.mersenne.org/download/ I'm testing this in about 20min before i do adjustments. Then it only for you to do the settings YOU are comfortable with. More or less what OP asked for //lazy EDIT: Have a lock athttps://setiathome.berkeley.edu/forum_thread.php?id=84694 For drivers of your GPUS |
Ian&Steve C. Send message Joined: 28 Sep 99 Posts: 4267 Credit: 1,282,604,591 RAC: 6,640 |
set your AVX offset to 0. Seti@Home classic workunits: 29,492 CPU time: 134,419 hours |
6BQ5 Send message Joined: 7 Dec 18 Posts: 29 Credit: 12,725,636 RAC: 357 |
I suggest that you find max power you could use. Wow, that's a meaty message with lots to unpack! I need your help. :) I ran another benchmark program last night before your message here. It's called OCCT v5.0 and it seriously pushed my processor, especially when running AVX2 instructions. Power consumption jumped to 190W and temperatures skyrocketed to 95 deg C to 98 deg C. My clock rate fell to a perfect 2^12. Yes, 4096 MHz. :-D I think it was just coincidence. The clock fluctuated +/- 2 to 3 MHz so it wasn't locked in at 4096 MHz. So, I think my ASRock Taichi Z370 motherboard can supply the necessary power. Does Intel limit T(jmax) of the i9-9900KS to 100 deg C? I have my AVX offset set to Auto now. When I come home tonight I will set it to 0 but I'm afraid the processor will get so hot that the clock rate will throttle down the same. I tried running Prime95 before and I got an error message. The message said there was an error starting the program. It came up right away and didn't run. Did I do something wrong? The link you provided, www.mersenne.org, is different than where I was downloading it from. I'll try downloading that version later tonight. Regarding the NVIDIA drivers ... yep, I saw that when I noticed some of my tasks were freezing at 0.603%. Personally, I don't think NVIDIA is going to do anything to help. It will be up to the SETI developers to come up with a workaround. What is the mikesworld.eu link? It looks like a hidden treasure chest of goodies! This is all new and fascinating to me. :-D -=- Boris |
Lazydude Send message Joined: 17 Jan 01 Posts: 45 Credit: 96,158,001 RAC: 136 |
Hi Boris! I we try to answer your Q OCCT : I know that program - I have not use it since 4,4 so i dont know how it vorking today Does Intel limit T(jmax) of the i9-9900KS to 100 deg C? It's the stock value -my Mobo I can set it to 115C So, I think my ASRock Taichi Z370 - Its accept I9-9900K but im not sure if your motherboard can deliver the power I9-9900K (s) wants do go at maximum I have set my MB to max 220W My Z370 board did not AVX offset=0 more and more programs uses AVX and above -prime95 uses FMA3 I tried running Prime95 before and I got an error message. Thats Good! It overwelming the CPU Start low 4 thread in the blend test its powerhouse beast What is the mikesworld.eu link? The stickie https://setiathome.berkeley.edu/forum_thread.php?id=71867#1375942 more info EDIT In my Radio Active Days my sig was SM7JXZ |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.