jasoncodispotiI am running a 9800x... I have not pushed my OC very far as I have really just not found a need to do so yet. For all of my CPU testing I use OCCT Large Data Set, results below...Stock 4.1GHz (what I consider to be baseline/stock): Target CPU Frequency - 4.1GhHzCPU Multiplier Control - AutoCPU Multiplier Setting - Auto (41)Mesh Ratio - Auto (24)AVX2 Offset - Auto (0)AVX3 Offset - Auto (0)BCLK Frequency - 100CPU VIN - AutoVCORE Voltage - AutoVMesh Voltage - Auto (1.000v)VIN VDroop - AutoVSA - Auto (0.800v)VCCIO - Auto (1.000v)PCH 1.0v - Auto (1.000v) OCCT Temps Large Dataset 1HR:OCCT AVX Disabled - AVG 50c, MAX 56c, MAX Fluid 28c, Max Inlet Air 23cOCCT AVX - AVG 47c, MAX 56c, MAX Fluid 28c, Max Inlet Air 23cOCCT AVX2 - AVG 49c, MAX 58c, MAX Fluid 28c, Max Inlet Air 23cOCCT AVX512 - AVG 47c, MAX 56c, MAX Fluid 30c, Max Inlet Air 24c OC 4.4Ghz (max OC I can hit with baseline/stock voltages): Target CPU Frequency - 4.4GhHzCPU Multiplier Control - Manual RatioLimitCPU Multiplier Setting - 44Mesh Ratio - 29AVX2 Offset - 0 (4.4GHz)AVX3 Offset - 1 (4.3 GHz)BCLK Frequency - 100CPU VIN - 1.800vVCORE Voltage - 1.200vVMesh Voltage - 1.000vVIN VDroop - EnabledVSA - 0.800vVCCIO - 1.000vPCH 1.0v - 1.000v OCCT Temps Large Dataset 1HR:OCCT AVX Disabled - AVG 61c, MAX 68c, MAX Fluid 30c, Max Inlet Air 24cOCCT AVX - AVG 58c, MAX 69c, MAX Fluid 30c, Max Inlet Air 24cOCCT AVX2 - AVG 61c, MAX 68c, MAX Fluid 30c, Max Inlet Air 24cOCCT AVX512 - NA Note:AVX512 is currently bugged for the Skylake CPU's and it really cant be tested, more information regarding that can be found here in another post that I did. https://forums.evga.com/X299-Dark-BIOS-v118-Possible-BugIssue-m2975001.aspx 200F or 93C is too hot IMO, I would honestly question the stability of that OC especially long term... X299 does not like heat and I would not recommend going above 80C for any extended period of time. Also keep in mind that MAX TJunction per Intel Spec for the 9800x is 95C. Would be interesting to see if that OC with those mentioned temps would pass OCCT stability test for 1HR with a large dataset, but again I would not advise running the CPU that HOT. That being said AVX/AVX2/AVX512 tend to generate a lot of heat when using most stability programs out there, heat and a load that you will ALMOST never find in a real world implementation of AVX assuming that it was implemented correctly. OCCT with a large dataset IMO is one of the best tools for testing CPU stability and well worth the cost, you may find that your temps drop when using OCCT compared to other tools that you have been using. "...but scoring in cinabench or geek bench is not giving a noticable difference in numbers."- Would not be surprised if you are seeing this due to some sort of throttling that is taking place due to the mentioned temps. Also if memory serves me correct either R15 does not use AVX instructions or it was poorly implemented, this was one of the reasons why R20 was created. "So if i drop avx loads to a 10 offset or even 5, temps seem to drop."- This would put your AVX clock speed at 4.00GHz down from 5.00GHz, far below what I am able to get at basically stock voltages. Something seems off here... again I would try testing these overclocks with OCCT and keep an eye on temps as well. "So what would be favored here. more clock speed like 5ghz, with higher offsets, or step below like 49 with very low offset."- As you increase core clock you will need to increase the voltage and as you increase voltage you will increase the amount of heat. AVX already runs HOT as it is so adding more voltage will compound this issues... There is a delicate balance between voltage and temperature and at a certain point you will find that you have to drop your AVX clocks below what you can could run at a lower voltage. For example at 1.2v you might be able to run AVX with no offset @ 4.4GHz, but at 1.3v you will have to run AVX at 4.0GHz; the added voltage causes too much heat. As to what is favored, well that depends on the workload that is being applied to the CPU, if its mostly a non AVX workload than who cares what AVX frequencies are as long as they are stable. I will say that it can be challenging to determine which applications and games use AVX instructions. There is a also a lot of misinformation on the internet about which applications use it. "I do notice in single core test in cinabench r15, i get 217, at 5 and 213 at 49."- A score of 217 vs 213 is within margin of error of testing.
bob16314Intel AVX is designed for use by applications that are strongly floating point compute intensive and can be vectorized. Example applications include audio processing and audio codecs, image and video editing applications, financial services analysis and modeling software, and manufacturing and engineering software.I use an AVX Offset that returns my overclock to the default Intel Turbo Boost speed when/if an AVX workload is running..Default for my mobo is an AVX Offset of 3 (minus 300MHz).
dmann304Ok this is very strange, i defaulted memory to 3600 cas 18, the uncore back to audo from +400, and VCCIO to auto, put vin back to 2.0v and cpu to 1.310v and set cpu to 50, and ran Cinabench, and AIDA, and temps went down 10c from thememory overclock of 4ghz at 16cas 1T command, so maybe that was the issue. Will i lose any performance in system? dropping memory back to its default?
Sultan.of.swingI must of had a damn good 9800x then.50x all core1.8 VIN1.310 Adaptive Vcore32x Mesh1.15 Mesh Voltage Ran it for 2 years like that and never crashed in anything I used it for and now my Roommate uses it.
dmann304Thats what i would like to have as settings. What temps? and cooling? I am thinking of grabbing a 10980XE soon to avoid all the comotion
Sultan.of.swingYou will lose some performance. I don't know which GPU you have but I have 1st place in Timespy for a 9800x and a single 2080ti https://www.3dmark.com/spy/13988569I like the CPU test in Timespy because it shows gains with clock frequency/mesh frequency and memory speed/timings. Maybe could give you a CPU score to aim for.