Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Hey guys so I am trying to overclock my system. Being that I’ve never done this I was hoping someone can guide me along the way and take a look at what I’ve come across. I’ve looked up YouTube videos and searched thru google but I can’t seem to find someone with my configuration.
I am running a i9-9900x skylake with the EVGA Dark X299 MB with 32gb ram 3200mhz 14C with a custom watercooled system using 2 360mm radiators and a Barrowch Cpu block and a Laing DDC pump which runs at full speed at all times. Flow meter says it pushes about 9-10L per minute
In the mean while I ran HWinfo with Prime95 to see what room I had for overclocking temps and seemed to me while testing in SMALL FFTS I was getting some really hot temps on stock. From what I read online these skylake-x cpus run really hot. I was getting 96-97c on 1 or 2 cores. Not really sure what to look for when looking at HWinfo other then I’m assuming anything that goes red is not good. Then when I reduced -50 on VCore and ran HWinfo again and my temps were below 88c at all times. If anyone can help that personally has a Dark X299 MB that would be awesome. I’m anxious to overclock my system as I feel I’m leaving money on the table performance wise if I left the system stock.
Hwinfo pic posted with RED readings is stock.
HWinfo pic WITHOUT Red reading is VCORE set at -50
Settings are back to stock
Thanks
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2019/06/08 09:19:06
(permalink)
I have the Intel i7-9800x (8 Core 16 Thread) paired up with the X299 Dark and a custom loop. Few tips... - If you are using HWInfo I would ditch that software ASAP and switch over to HWInfo64. Similar name, but very different software HWInfo64 will allow you to see all of the sensors (of which there are many) on the X299 Dark. The naming of the sensors in HWInfo64 can be a little misleading/confusing and some of the sensors dont report "valid" data... To help you out I have attached my settings that I use for HWInfo64 that should clear out the non-valid/non useful sensors and re-name some of them to make them more identifiable. After you have installed HWInfo64 you should be able to just double click on the attached file and it will import my settings for you.
- Utilizing AVX Offset's for both AVX-512 and AVX/AVX2 will be critical to being able to overclock on the X299 platform, if you are not familiar with what I am talking about let me know.
- The newest version of Prime95 defaults to using AVX-512 which generates an extreme amount of heat and will almost certainly require you to use an AVX Offset in order to keep temperatures under control. I am going to guess that the reason why you are seeing such high temperatures in your initial testing is due to the lack of an AVX-512 offset.
- I found that the X299 Dark was NOT setting the correct TJunction value (max CPU temp) in the BIOS and I had to manually set this. For your CPU it is 92C and I would highly suggest that you change this setting from AUTO to 92C in the BIOS, think its under CPU Configuration if you cant find it let me know. An incorrect TJunction value in the BIOS could allow the CPU to overheat and it will result in incorrect temperature values being read by the BIOS/Software.
- While Intel puts the MAX temp for your CPU at 92C I would suggest not exceeding 80C. 96C is way too HOT...
- Before doing any overclocking I would suggest establishing a baseline with default/auto settings than start overclocking from that point. This will give you an idea of the CPU temps at stock and establish that the CPU is functioning correctly.
- Also check out this guide from Tin, https://xdevs.com/guide/e299ocg/
My method for testing CPU stability... - Prime95 Small FTT - 1 Hour
- Prime95 Small FTT AVX-512 Disabled - 1 Hour
- Prime95 Small FTT AVX Fully Disabled - 30 Minutes
- Prime95 Blend AVX Fully Disabled - 1 Hour
- 3DMark Fire Strike Extreme Physics Test Only - 1 Hour
- Realbench Stress Test 32GB of Memory - 1 Hour
This is my method for testing stability depending on your use case you may need to increase the amount of testing that is done, but I would run this at a minimum before saying that an overclock is stable. Yes, I would run all of these tests with the CPU on default settings as well to establish a baseline and make sure that CPU is functioning correctly. When running an AVX offset it is CRITICAL that you run all of the Prime95 tests mentioned above, its the only way to fully test the CPU. Link To HWInfo65 Software: https://www.hwinfo.com/download/ Link To My HWInfo64 Settings: https://drive.google.com/file/d/1ksNt-sqvVVjtDhQ35kqgXbrbpZgxU4Wu/view?usp=sharing
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2 
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 09:34:17
(permalink)
jasoncodispoti thanks for your response. I was actually using HWinfo64. I didn't realize there was a difference(rookie here) lol. Also not quite sure how to utilize the AVX offsets. and do you think my CPU reaching 96C could of done some permanent damage? is there a way to tell?
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2019/06/08 09:40:57
(permalink)
Eddien2990
jasoncodispoti thanks for your response. I was actually using HWinfo64. I didn't realize there was a difference(rookie here) lol. Also not quite sure how to utilize the AVX offsets. and do you think my CPU reaching 96C could of done some permanent damage? is there a way to tell?
Regarding the high temperature anything is possible, but most likely you are OK. Which BIOS version are you using? Assuming that you have not changed the TJunction value in the BIOS already Inside of HWINfo64 right click on any of the CPU Core Temperature readings and select "adjust tj max" what number does it have set by default, this is what the BIOS is setting it to? Let me grab some lunch and I will make another post regarding AVX... lol.
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 09:42:41
(permalink)
Just changed it in the BIOS. Set tjmax at 92
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 09:43:50
(permalink)
Using bios version 1.17. The latest update
|
Cool GTX
EVGA Forum Moderator
- Total Posts : 30963
- Reward points : 0
- Joined: 2010/12/12 14:22:25
- Location: Folding for the Greater Good
- Status: offline
- Ribbons : 122
Re: EVGA X299 DARK NOOB
2019/06/08 11:08:22
(permalink)
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2019/06/08 11:22:39
(permalink)
Cool GTX
This has a lot of good information https://xdevs.com/guide/e299ocg/
@ jasoncodispoti
Thanks for helping the OP
Nice detailed list 
BR awarded
No problem, thanks!
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 11:25:47
(permalink)
I’m reading that link. Just need to fully understand it all.
|
Cool GTX
EVGA Forum Moderator
- Total Posts : 30963
- Reward points : 0
- Joined: 2010/12/12 14:22:25
- Location: Folding for the Greater Good
- Status: offline
- Ribbons : 122
Re: EVGA X299 DARK NOOB
2019/06/08 11:27:42
(permalink)
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2019/06/08 11:40:34
(permalink)
Regarding AVX...
For the purpose of this conversation lets just say that the CPU can interpret 4 types of instructions; AVX-512, AVX2, AVX, and Non-AVX. The X299 Dark allows you to set what is referred to as an AVX offset for AVX-512 and AVX/AVX2. The offset for AVX/AVX2 instructions is combined into one setting within the BIOS. If for example you set the multiplier for the CPU to 45 the resulting frequency would be 4.5GHz, the max frequency the CPU would run at when processing Non-AVX instructions. Due to the increase in power consumption of both AVX-512 and AVX/AVX2 instructions it is typically not possible to run the CPU at 4.5GHz when processing those instruction types due to heat, this is where an AVX offset comes into play. Setting an AVX-512 offset of 5 would result in a maximum CPU frequency of 4.00GHz and setting an AVX/AVX2 offset of 2 would result in a maximum CPU frequency of 4.3GHz. By decreasing the max CPU frequency when processing AVX instructions you can decrease the power consumption and subsequently the temperature of the CPU. At a basic level this allows you to overclock for each type of instruction set that the CPU can interpret.
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 11:52:53
(permalink)
Understood. Thanks for that explanation. So being that overclocking has many variables/settings. For someone like me that has never done this how do I go about starting? Keeping temps low. And getting results that won’t fry my cpu with longevity lol
post edited by Eddien2990 - 2019/06/08 12:22:43
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2019/06/08 12:14:44
(permalink)
Eddien2990
Understood. Thanks for that explanation. So being that overclocking are many variables/settings. For someone like me that has never done this how do I go about starting? Keeping temps low. And getting results that won’t fry my cpu with longevity lol
Trial and error... Just slowly increase the CPU multiplier until the CPU either gets too hot or is not stable, dont forget to increase the AVX Offsets as needed. Below is some voltage information that you can use for a good starting point. Voltage is what kills CPU's its really the only thing that you need to "worry" about when changing settings pretty much everything else can be reversed. Starting Voltage Settings: - CPU VIN - 1.800v; You should be able to take this safely up to 2.000v, but it should not be needed. As a general rule of thumb dont go below 1.800v either as the CPU may start to phantom throttle (very hard to know that its doing this).
- VCORE - Override 1.200v; This is basically default voltage for the CPU, but should allow for some overclocking before you have to change it. For your CPU since its a 10 core CPU I would say max VCORE is 1.300v to 1.350v. Increasing VCORE voltage will help stabilize the CPU when overclocking.
- VMESH - Override 1.100v; I believe that you can go higher than this, but the returns are usually fairly minimal and not worth the extra heat.
- VSA - 0.800v; Default
- VCCIO - 1.000v; Default
- PCH 1.0v - 1.000v; Default
If you end up applying settings that make it so the PC want turn on or boot, just hold down the reset button on the rear IO panel. Doing this will reset all settings in the BIOS back to default so be sure to save your BIOS settings to a USB drive so you load them back up.
post edited by jasoncodispoti - 2019/06/08 12:17:22
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 12:22:10
(permalink)
So being that Prime95 got my CPU up to 96-97c on stock. Would any undervolting be needed? Also what would be a maximum voltage for my CPU. What should I not exceed?
post edited by Eddien2990 - 2019/06/08 12:32:00
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2019/06/08 12:34:38
(permalink)
Eddien2990
So being that Prime95 got my CPU up to 96-97c on stock. Would any undervolting be needed? Also what would be a maximum voltage for my CPU. What should I not exceed?
Possibly, but that can get a bit complex... For now I would suggest keeping things simply until you get the hang of it. The voltage settings that I suggested above are safe and should give you some room to overclock before needing to adjust them, but you will require an AVX offset to keep those temps under control. Start with an AVX-512 offset of 4 and an AVX/AVX2 offset of 2, if you have the temp headroom you can decrease those values, but you may find that you need to increase them as well. Also I would not suggest using Adaptive voltage stick with override for now. Also try searching for Intel i9-7900x overclocking guides... the information in those guides will be very similar to your CPU.
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 12:42:30
(permalink)
In the BIOS is has the option of AVX2 and AVX3. Both set at 3. AVX3 would be AVX-512?
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2019/06/08 13:24:21
(permalink)
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
Eddien2990
New Member
- Total Posts : 9
- Reward points : 0
- Joined: 2019/05/20 09:46:16
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2019/06/08 14:12:57
(permalink)
Is there anyway you can PM me? I can’t do it. Forum doesn’t let me
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2021/01/19 16:39:33
(permalink)
INteresting post, i find this helpful. Avx3 would be the one that Cinabench punishes the CPU, with heat?
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: EVGA X299 DARK NOOB
2021/01/19 17:22:43
(permalink)
dmann304
Interesting post, i find this helpful. Avx3 would be the one that Cinabench punishes the CPU, with heat?
Both AVX2 and AVX3 create more heat than non-AVX code, AVX3 more than AVX2 -IF- you use same offset. -HOWEVER- you can make all of them run at same temperature if you tune offsets while observing temperature under workload. To find which code some workload (program) uses set AVX2 and 3 at different offsets and observe speed during execution. For example if you set your CPU to 43x, AVX2 to 3, and AVX3 to 5 hitting 4.3 GHz means that program is not using AVX code, hitting 4 GHz would mean it is using AVX2, and 3.8 GHz means it is using AVX3 code.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2021/01/19 18:32:32
(permalink)
I know that, but i notice when i set an 8 - 8 offsets, i still seen 48 on cpu speed, during workload. SO its not downclocking
during a stressful instruction. of the 2 avx's, which one is more concerning.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: EVGA X299 DARK NOOB
2021/01/19 18:38:23
(permalink)
dmann304
I know that, but i notice when i set an 8 - 8 offsets, i still seen 48 on cpu speed, during workload. SO its not downclocking during a stressful instruction.
Hmmm ... that is interesting ... what are you using to check speed of CPU when running Cinebench and which version of Cinebench you are running? dmann304
of the 2 avx's, which one is more concerning.
Out of two AVX3 will always result in more heat if everything else (test type and offsets) is identical.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2021/01/19 20:03:32
(permalink)
R15 and R20. CPU.ID is in taskbar. shows still 48, sometimes i do see a 43, but i have both set to 2 offsets, for 46 under loads.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: EVGA X299 DARK NOOB
2021/01/20 01:42:50
(permalink)
dmann304
R15 and R20. CPU.ID is in taskbar. shows still 48, sometimes i do see a 43, but i have both set to 2 offsets, for 46 under loads.
I thought you said your offsets were at 8, not 2? In any case … I tried to reproduce your findings and have spent last 3 hours sorting through what I found. When running R20 (I didn’t try R15 as R15 doesn’t use AVX) I too didn’t observe any downclocks through Task Manager, nor HWInfo64 or XTU. While searching why that might be the case I’ve come across this post https://www.techpowerup.c...re.213237/post-4256702 by person that has X299 Dark and i9-10900X (just like me) reporting that he too didn’t experience downclock and wondering is R20 using AVX2/AVX512 at all. So on the chance my R20 download is not good I tried R23 from Microsoft Store and didn’t observe any downclocks with it either. Searching for possible problems with R23 and AVX-512 led me to this post https://twitter.com/iancutress/status/1327358373373898752?lang=en that quotes Maxon as saying they are not using AVX512 in R23 which put whole using of Cinebench as reliable indicator AVX2/AVX3 offsets are doing as they should under the question mark. So I’ve turned my head to tool that I know definitely uses AVX2 and AVX3: AVX2 and AVX3 stress test part of XTU. Running them indicated some throttling down by offsets seems to be happening but, if my memory serves me well, when I first did such tests while ago it used to be much easier to spot them. Only way I could definitely confirm they are happening in XTU was by observing ‘minimum’ (rather than ‘current’) frequency value in HWInfo64 with faster than default reading interval. Also this time around it seemed as if it was taking longer for frequency to drop and faster to recover. Why that might be the case? Only reasons I can come up with is that something changed between first time I tested system and now and as a result either downclocking is shorter in duration or monitoring is not fast enough to detect it or combo of both. I can’t remember with certainty which version of BIOS I initially used (IIRC 1.21) and Windows was build 1909 but I don’t think it would be worth my time trying to figure it out as long as AVX* code works and maintains same performance and temperatures as before.
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2021/01/20 07:25:40
(permalink)
ZoranC
I thought you said your offsets were at 8, not 2?
In any case … I tried to reproduce your findings and have spent last 3 hours sorting through what I found.
When running R20 (I didn’t try R15 as R15 doesn’t use AVX) I too didn’t observe any downclocks through Task Manager, nor HWInfo64 or XTU. While searching why that might be the case I’ve come across this post
https://www.techpowerup.c...re.213237/post-4256702
by person that has X299 Dark and i9-10900X (just like me) reporting that he too didn’t experience downclock and wondering is R20 using AVX2/AVX512 at all.
So on the chance my R20 download is not good I tried R23 from Microsoft Store and didn’t observe any downclocks with it either. Searching for possible problems with R23 and AVX-512 led me to this post
https://twitter.com/iancutress/status/1327358373373898752?lang=en
that quotes Maxon as saying they are not using AVX512 in R23 which put whole using of Cinebench as reliable indicator AVX2/AVX3 offsets are doing as they should under the question mark.
So I’ve turned my head to tool that I know definitely uses AVX2 and AVX3: AVX2 and AVX3 stress test part of XTU. Running them indicated some throttling down by offsets seems to be happening but, if my memory serves me well, when I first did such tests while ago it used to be much easier to spot them.
Only way I could definitely confirm they are happening in XTU was by observing ‘minimum’ (rather than ‘current’) frequency value in HWInfo64 with faster than default reading interval.
Also this time around it seemed as if it was taking longer for frequency to drop and faster to recover.
Why that might be the case? Only reasons I can come up with is that something changed between first time I tested system and now and as a result either downclocking is shorter in duration or monitoring is not fast enough to detect it or combo of both.
I can’t remember with certainty which version of BIOS I initially used (IIRC 1.21) and Windows was build 1909 but I don’t think it would be worth my time trying to figure it out as long as AVX* code works and maintains same performance and temperatures as before.
I have seen YouTubers indicate in passing that Cinebench R20 uses AVX instructions... that being said I question the implementation of it. First thing to get out of the way its highly unlikely that Cinebench would use AVX-512 instructions in any from since the only consumer CPU's that I know of that support it is the X299 platform. To be clear no AMD CPU supports AVX-512 and up until recently AMD's implementation of AVX2 was very poor. Side note, I believe that the new CPU's Intel announced are going to support AVX-512 as well not 100% sure on that though. Really the only application's that I know of that has implemented AVX-512 instructions is Handbrake and (I think) some of the Adobe applications. Using HWINFO64 with a polling time of 50ms I have NOT seen any indication that Cinebench R20 uses AVX-512 instructions and if it is its doing so for an extremely short period of time. Again using HWINFO64 with a polling time of 50ms I have been able to confirm that Cinebench R20 appears to be using AVX2 instructions, but its is for a VERY short period of time and using a normal polling time of say 2,000ms does not detect the CPU frequency change. At this point I would say that it is safe to assume that Cinebench R15/R20 is not using AVX-512 instructions and that it MAY be using AVX2 instructions, but at a negligible level. I would NOT recommend using Cinebench in any capacity for stress testing an overclock due to the lack of clarity on AVX usage and the lack of error detection. I do however feel that Cinebench is a good performance benchmark tool and would recommend it for that usage. As I have stated before I would use the paid version of OCCT for all CPU stability testing and it gives you the ability to select the instruction set that you want to use SSE/AVX/AVX2/AVX-512.
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
jasoncodispoti
SSC Member
- Total Posts : 628
- Reward points : 0
- Joined: 2013/03/18 14:40:28
- Location: Nashville, TN
- Status: offline
- Ribbons : 14
Re: EVGA X299 DARK NOOB
2021/01/20 07:30:37
(permalink)
Also keep in mind that up until BIOS v1.27 for the X299 Dark AVX-512 instructions have bene very broken because of a microcode change that Intel made a while back, so I would not be surprised to see different results in regards to AVX-512's behavior, performance, and stability with BIOS v1.27. I know that my previous overclocks were not stable with the new BIOS and I have been re-doing all my overclocks since release of v1.27...
Intel Core i7-9800x | EVGA X299 Dark| EVGA RTX 2080ti FTW3 Ultra HydroCopper | 32GB Corsair Dominator Platinum | EKWB Liquid Cooled Loop | Phanteks Enthoo Primo SE | EVGA SuperNOVA 1000 P2
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2021/01/20 08:50:40
(permalink)
That is an interesting find, i might stop using R15 or 20 for stress anymore, since it behaves that way. Is the Timespy stress test valid for oc? besides OCCT? and i run games like: RDR 2, and watch dogs 4, and those games are the most taxing, and Cyberpunk for hours, and no issues, so my OC must be working good, since i do mostly games and a few runs of Handbrake. As for XTU, the strangest thing is a can't get it installed again on my setup. Error when running setup, and i done all the things that were suggested: like, deleting the folders, the xtu drivers, and even registry keys, and that damn thing wont install. even with apps loaded: like Eleet X, turboboost max 3 applet icon, and even Precision X1. So i am stuck with Aia 64 paid version which runs good, but even though temps aren't out of control shows in red throttling in the lower green box which is bull****, cuz i on't have heat thermal throttling going on.
post edited by dmann304 - 2021/01/20 09:01:56
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: EVGA X299 DARK NOOB
2021/01/20 11:05:39
(permalink)
jasoncodispoti
I have seen YouTubers indicate in passing that Cinebench R20 uses AVX instructions...
I too have seen articles around, like this one https://www.pcworld.com/article/3356179/meet-the-new-tougher-cinebench-r20-benchmark-we-test-it-on-xeon-and-threadripper.html , claiming it uses both AVX2 and AVX512, which is what made me start digging deeper. Using HWINFO64 with a polling time of 50ms I have NOT seen any indication that Cinebench R20 uses AVX-512 instructions and if it is its doing so for an extremely short period of time. Again using HWINFO64 with a polling time of 50ms I have been able to confirm that Cinebench R20 appears to be using AVX2 instructions, but its is for a VERY short period of time and using a normal polling time of say 2,000ms does not detect the CPU frequency change. At this point I would say that it is safe to assume that Cinebench R15/R20 is not using AVX-512 instructions and that it MAY be using AVX2 instructions, but at a negligible level. Same findings as me then. Watching R20 (not trying R15 which officially doesn’t support them) with fast polling HWInfo64 detected no downclocks that could be due to AVX-512 and very little (seemingly just on single core) downclocks that could be due to AVX2 and that were very brief. Using default polling interval was not fast enough to catch them. That combined with a) Quote in that Twitter thread of Maxon that they are not using AVX512 in R23 and that Intel has disabled support for it in Embtree (which Maxon uses), b) Pete Brubaker saying in that same thread “Yes there is a downclock, but it's not from using the instruction set. It's with using 512B registers. 2) You can use the instructions on 256B registers with no downclock. 3) Newer architectures (ICL+) don't have a significant downclock.” leads me to believe that if R20 has ever used AVX2/AVX512 that might have been dropped in later builds, and one can’t look for downclock when running it to check on offsets. In other words: Use it as indicator of system’s performance? Maybe. Use it as an indicator of AVX2/AVX512 offsets doing their job? Nope. jasoncodispoti
As I have stated before I would use the paid version of OCCT for all CPU stability testing and it gives you the ability to select the instruction set that you want to use SSE/AVX/AVX2/AVX-512.
Amen! OCCT is one of tools I find good. I also use XTU that also offers instruction set specific stress tests.
|
ZoranC
FTW Member
- Total Posts : 1099
- Reward points : 0
- Joined: 2011/05/24 17:22:15
- Status: offline
- Ribbons : 16
Re: EVGA X299 DARK NOOB
2021/01/20 11:07:37
(permalink)
jasoncodispoti
Also keep in mind that up until BIOS v1.27 for the X299 Dark AVX-512 instructions have bene very broken because of a microcode change that Intel made a while back, so I would not be surprised to see different results in regards to AVX-512's behavior, performance, and stability with BIOS v1.27. I know that my previous overclocks were not stable with the new BIOS and I have been re-doing all my overclocks since release of v1.27...
1.27? Do you mean 1.25? Cause 1.25 is latest release I see.
|
dmann304
iCX Member
- Total Posts : 410
- Reward points : 0
- Joined: 2008/07/16 22:44:20
- Status: offline
- Ribbons : 0
Re: EVGA X299 DARK NOOB
2021/01/21 13:04:33
(permalink)
Hey guys, someone who has the best experience with Cascade Lake X chips, I decided to stay at 48, with 3 avx offset. but was running
voltages i didn't need to. So i decided to set Mesh at 30 with the recommended voltage for that: 1.116 according to Intel, and was running
a 1.9v VCCIN, and CPU vCore at 1.310. But the consenses around here is: lower is better, so i played with VCCIN to default 1.8v, and CPU
vcore down to 1.27, and wow, boots fine, and passes tests, so i was loading to much voltage in to the CPU, so I am thinking all the issues i was
having, was the Mesh being to high or memory settings.
|