bp7178
iCX Member
- Total Posts : 373
- Reward points : 0
- Joined: 4/15/2016
- Status: offline
- Ribbons : 3
Re: Low GFlops in Linpack
Wednesday, November 13, 2019 3:05 AM
(permalink)
I agree.
This is the second problem where using an Asus board fixes the problem I'm having.
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Wednesday, November 13, 2019 9:03 PM
(permalink)
Have you had any issues with the ASMedia USB ports?
|
chinobino.
New Member
- Total Posts : 42
- Reward points : 0
- Joined: 9/20/2008
- Status: offline
- Ribbons : 6
Re: Low GFlops in Linpack
Thursday, November 14, 2019 0:34 PM
(permalink)
☼ Best Answerby Cool GTX Monday, November 18, 2019 6:09 PM
I have the EVGA Z370 Classified K with a 9900K at 4.8 GHz all cores (no AVX offset) with some garbage Kingston DDR4 2400 RAM @ 17-17-17-39 2T, here's my results; Linpack Xtreme 1.1.1 = ~330 GFLOPS Linx v0.9.6 = ~302 GFLOPS Looking at the screenshot from the Korean guy he was getting ~460 GFLOPS running Linx V0.9.5 with a 9900K at 5 GHz and DDR4 4000 @ 17-19-19-38 2T, tRFC 560, tFAW 16. This post; hxxps://www.overclock.net/forum/5-intel-cpus/1711718-9900k-overclock-results-questions-29.html#post27739262 by CptKuolio has some interesting info (i.e. gaining 10 to 15 gflops with tREFI maxed out), 9900K outputting ~530 GFLOPS in Linx v0.9.0 running at 5.2 GHz DDR4 4133 @ 16-17-17-37 2T, tRFC 340, tFAW 16 on an Asus Z390. It seems that Linpack is very sensitive to background processes and memory configuration. Maybe you could try dropping down to DDR 4133/4000 CL17 and seeing if it's a sweet spot. You could also try copying the timings CptKuolio used on his ROG Maximus Formula, not sure how much VDIMM he used though.
post edited by chinobino. - Thursday, November 14, 2019 0:38 PM
|
bp7178
iCX Member
- Total Posts : 373
- Reward points : 0
- Joined: 4/15/2016
- Status: offline
- Ribbons : 3
Re: Low GFlops in Linpack
Thursday, November 14, 2019 5:54 AM
(permalink)
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Thursday, November 14, 2019 10:04 PM
(permalink)
chinobino.
I have the EVGA Z370 Classified K with a 9900K at 4.8 GHz all cores (no AVX offset) with some garbage Kingston DDR4 2400 RAM @ 17-17-17-39 2T, here's my results;
Linpack Xtreme 1.1.1 = ~330 GFLOPS
Linx v0.9.6 = ~302 GFLOPS
Looking at the screenshot from the Korean guy he was getting ~460 GFLOPS running Linx V0.9.5 with a 9900K at 5 GHz and DDR4 4000 @ 17-19-19-38 2T, tRFC 560, tFAW 16.
This post;
hxxps://www.overclock.net/forum/5-intel-cpus/1711718-9900k-overclock-results-questions-29.html#post27739262
by CptKuolio has some interesting info (i.e. gaining 10 to 15 gflops with tREFI maxed out), 9900K outputting ~530 GFLOPS in Linx v0.9.0 running at 5.2 GHz DDR4 4133 @ 16-17-17-37 2T, tRFC 340, tFAW 16 on an Asus Z390.
It seems that Linpack is very sensitive to background processes and memory configuration.
Maybe you could try dropping down to DDR 4133/4000 CL17 and seeing if it's a sweet spot.
You could also try copying the timings CptKuolio used on his ROG Maximus Formula, not sure how much VDIMM he used though.
Finally!!!, someone has an answer. https://ibb.co/cgVGx7n https://ibb.co/LzZZq7c Temps were getting into the 90°C, so I had to open a window and get some nice 5°C air going thru the radiator.
post edited by kelkel1 - Thursday, November 14, 2019 10:46 PM
|
bp7178
iCX Member
- Total Posts : 373
- Reward points : 0
- Joined: 4/15/2016
- Status: offline
- Ribbons : 3
Re: Low GFlops in Linpack
Friday, November 15, 2019 0:35 PM
(permalink)
So a memory configuration problem?
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Friday, November 15, 2019 0:40 PM
(permalink)
bp7178
So a memory configuration problem?
Apparently, although I was using the XMP1 profile and never adjusted any memory setting. After changing tFAW to 16 and tRFC to 470, you see the results. Now, I just have to tweak it.
|
bp7178
iCX Member
- Total Posts : 373
- Reward points : 0
- Joined: 4/15/2016
- Status: offline
- Ribbons : 3
Re: Low GFlops in Linpack
Friday, November 15, 2019 1:58 AM
(permalink)
Forgive my ignorance, I've never changed memory timings outside of enabling XMP.
Are those timings included in the XMP profile or are they set by the BIOS as an auto value?
|
chinobino.
New Member
- Total Posts : 42
- Reward points : 0
- Joined: 9/20/2008
- Status: offline
- Ribbons : 6
Re: Low GFlops in Linpack
Friday, November 15, 2019 11:11 AM
(permalink)
Nice results!
Good to see my suspicions were confirmed as I have just ordered some DDR4 4133 CL19 (should be Samsung B-Die) so I should also be able to replicate the results soon.
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Friday, November 15, 2019 4:43 PM
(permalink)
bp7178
Forgive my ignorance, I've never changed memory timings outside of enabling XMP.
Are those timings included in the XMP profile or are they set by the BIOS as an auto value?
You will need to go into memory configuration is BIOS and manually set them.
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Friday, November 15, 2019 4:44 PM
(permalink)
chinobino.
Nice results!
Good to see my suspicions were confirmed as I have just ordered some DDR4 4133 CL19 (should be Samsung B-Die) so I should also be able to replicate the results soon.
Thanks for the solution, you deserve a blue ribbon.
|
HeavyHemi
Omnipotent Enthusiast
- Total Posts : 13887
- Reward points : 0
- Joined: 11/29/2008
- Location: Western Washington
- Status: offline
- Ribbons : 135
Re: Low GFlops in Linpack
Saturday, November 16, 2019 6:09 PM
(permalink)
kelkel1
chinobino.
Nice results!
Good to see my suspicions were confirmed as I have just ordered some DDR4 4133 CL19 (should be Samsung B-Die) so I should also be able to replicate the results soon.
Thanks for the solution, you deserve a blue ribbon.
'Solution'. If the memory is not being correctly set by the XMP profile for the board, I'd contact EVGA and see if this is a BIOS issue they can resolve or a peculiarity of the XMP profile stored in the memory that is setting memory timings differently.
EVGA X99 FTWK / i7 6850K @ 4.5ghz / RTX 3080Ti FTW Ultra / 32GB Corsair LPX 3600mhz / Samsung 850Pro 256GB / Be Quiet BN516 Straight Power 12-1000w 80 Plus Platinum / Window 10 Pro
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Sunday, November 17, 2019 12:18 AM
(permalink)
HeavyHemi
kelkel1
chinobino.
Nice results!
Good to see my suspicions were confirmed as I have just ordered some DDR4 4133 CL19 (should be Samsung B-Die) so I should also be able to replicate the results soon.
Thanks for the solution, you deserve a blue ribbon.
'Solution'. If the memory is not being correctly set by the XMP profile for the board, I'd contact EVGA and see if this is a BIOS issue they can resolve or a peculiarity of the XMP profile stored in the memory that is setting memory timings differently.
Do you really believe evga will do anything about that? I mean, one or two users having an 'issue' with one application that hardly anyone uses and has no bearing on daily 'normal' use? And, those users having different DIMMS? Seems like a lot of work for evga... I seriously doubt evga will feel the need to address the 'issue'.
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Sunday, November 17, 2019 3:41 PM
(permalink)
Anyway, after adjusting tFAW and tRFC, I ran LinX 0.9.5 and Linpack Extreme 1.1.1 and posted the results in my other thread.
|
bp7178
iCX Member
- Total Posts : 373
- Reward points : 0
- Joined: 4/15/2016
- Status: offline
- Ribbons : 3
Re: Low GFlops in Linpack
Monday, November 18, 2019 2:30 AM
(permalink)
kelkel1
Do you really believe evga will do anything about that? I mean, one or two users having an 'issue' with one application that hardly anyone uses and has no bearing on daily 'normal' use? And, those users having different DIMMS?
Seems like a lot of work for evga... I seriously doubt evga will feel the need to address the 'issue'.
I for one think they should. I made a thread about how I couldn't get Intel Performance Maximizer to work and was sent a test BIOS to see if it resolved the issue and it did. Its not that hard for them to modify BIOS files and fix issues as they come up. The big problem is that its a problem on an EVGA board, but not Asus. Kind of the same thing I ran into testing a 9900KS. On the EVGA board it auto applied way too much voltage and would hit 90c+ on stock settings. This wasn't a problem on an Asus board. I think there's a few issues that need to get addressed with a BIOS patch but I suppose that will depend on how close to EOL the Dark is.
|
Kylearan
iCX Member
- Total Posts : 288
- Reward points : 0
- Joined: 12/26/2013
- Status: offline
- Ribbons : 2
Re: Low GFlops in Linpack
Monday, November 18, 2019 6:12 AM
(permalink)
kelkel1
Anyway, after adjusting tFAW and tRFC, I ran LinX 0.9.5 and Linpack Extreme 1.1.1 and posted the results in my other thread.
I replied there, but your Linpack EX 1.1.1 had mismatched residuals, so the math calculation was not stable (insufficient vcore). The built in residual check doesn't work on Intel processors automatically. You have to override it with /residualcheck The 5 ghz LinX test was stable. https://ibb.co/qMRm13X
|
chinobino.
New Member
- Total Posts : 42
- Reward points : 0
- Joined: 9/20/2008
- Status: offline
- Ribbons : 6
Re: Low GFlops in Linpack
Monday, November 18, 2019 3:27 PM
(permalink)
bp7178
kelkel1
Do you really believe evga will do anything about that? I mean, one or two users having an 'issue' with one application that hardly anyone uses and has no bearing on daily 'normal' use? And, those users having different DIMMS?
Seems like a lot of work for evga... I seriously doubt evga will feel the need to address the 'issue'.
I for one think they should. I made a thread about how I couldn't get Intel Performance Maximizer to work and was sent a test BIOS to see if it resolved the issue and it did. Its not that hard for them to modify BIOS files and fix issues as they come up.
The big problem is that its a problem on an EVGA board, but not Asus. Kind of the same thing I ran into testing a 9900KS. On the EVGA board it auto applied way too much voltage and would hit 90c+ on stock settings. This wasn't a problem on an Asus board.
I think there's a few issues that need to get addressed with a BIOS patch but I suppose that will depend on how close to EOL the Dark is.
There is nothing for EVGA to fix, the XMP profile is being correctly set but if you want the extra performance you have to manually fine tune other settings such as tRFC and tFAW. However, I do agree that there probably needs to be another BIOS release that is fine tuned for the 9900KS (and not just for the Z390 Dark).
|
Cool GTX
EVGA Forum Moderator
- Total Posts : 31353
- Reward points : 0
- Joined: 12/12/2010
- Location: Folding for the Greater Good
- Status: offline
- Ribbons : 123
Re: Low GFlops in Linpack
Monday, November 18, 2019 6:10 PM
(permalink)
Best Practice is to set the RAM Manually ---> for Stable & Best OC
Learn your way around the EVGA Forums, Rules & limits on new accounts Ultimate Self-Starter Thread For New Members
I am a Volunteer Moderator - not an EVGA employee
Older RIG projects RTX Project Nibbler
When someone does not use reason to reach their conclusion in the first place; you can't use reason to convince them otherwise!
|
HeavyHemi
Omnipotent Enthusiast
- Total Posts : 13887
- Reward points : 0
- Joined: 11/29/2008
- Location: Western Washington
- Status: offline
- Ribbons : 135
Re: Low GFlops in Linpack
Tuesday, November 19, 2019 10:22 PM
(permalink)
kelkel1
HeavyHemi
kelkel1
chinobino.
Nice results!
Good to see my suspicions were confirmed as I have just ordered some DDR4 4133 CL19 (should be Samsung B-Die) so I should also be able to replicate the results soon.
Thanks for the solution, you deserve a blue ribbon.
'Solution'. If the memory is not being correctly set by the XMP profile for the board, I'd contact EVGA and see if this is a BIOS issue they can resolve or a peculiarity of the XMP profile stored in the memory that is setting memory timings differently.
Do you really believe evga will do anything about that? I mean, one or two users having an 'issue' with one application that hardly anyone uses and has no bearing on daily 'normal' use? And, those users having different DIMMS?
Seems like a lot of work for evga... I seriously doubt evga will feel the need to address the 'issue'.
Okay, I got EVGA to address the micro code issues for X99 that was preventing overclocking on Broadwell-E. EVGA issued BIOS updates, just in the last few months. EVGA in turn contacted intel so the new ucode updates integrated into the OS should no longer kill overclocking on X99. If you just give up, you won't know. It took me over a month and several threads being rather rude about it. I can link you to them. You should have very similar performance to other boards without needing to alter settings. It's your call.
EVGA X99 FTWK / i7 6850K @ 4.5ghz / RTX 3080Ti FTW Ultra / 32GB Corsair LPX 3600mhz / Samsung 850Pro 256GB / Be Quiet BN516 Straight Power 12-1000w 80 Plus Platinum / Window 10 Pro
|
HeavyHemi
Omnipotent Enthusiast
- Total Posts : 13887
- Reward points : 0
- Joined: 11/29/2008
- Location: Western Washington
- Status: offline
- Ribbons : 135
Re: Low GFlops in Linpack
Tuesday, November 19, 2019 10:29 PM
(permalink)
Cool GTX
Best Practice is to set the RAM Manually ---> for Stable & Best OC
Best practice is to have the RAM function as designed with the performance as designed with no user intervention. Additional performance, ie overclocking and manual settings, is not a best practice and never has been. It is a practice preferred by enthusiasts to eek out more performance. Generally speaking, 100% stability is not the goal, game stable is. Sorry, if I seem a bit harsh, but I think being accurate is important and a lot of times folks on forums get a bit confused thinking the enthusiast practices are the best or even preferred practice. A form of confirmation BIAS. For the average daily user, they are not. Put another way overclockers are a tiny fraction of the PC user base. Their practices are not best practices for the general user base. That is all. 
EVGA X99 FTWK / i7 6850K @ 4.5ghz / RTX 3080Ti FTW Ultra / 32GB Corsair LPX 3600mhz / Samsung 850Pro 256GB / Be Quiet BN516 Straight Power 12-1000w 80 Plus Platinum / Window 10 Pro
|
bp7178
iCX Member
- Total Posts : 373
- Reward points : 0
- Joined: 4/15/2016
- Status: offline
- Ribbons : 3
Re: Low GFlops in Linpack
Wednesday, November 20, 2019 3:09 AM
(permalink)
chinobino.
There is nothing for EVGA to fix, the XMP profile is being correctly set but if you want the extra performance you have to manually fine tune other settings such as tRFC and tFAW.
However, I do agree that there probably needs to be another BIOS release that is fine tuned for the 9900KS (and not just for the Z390 Dark).
Problem. This "fine tuning" is only needed on EVGA boards. 500+ Gflops in Linpak isn't extra performance, its the correct performance for that clock speed, 5ghz.
post edited by bp7178 - Wednesday, November 20, 2019 7:42 PM
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Thursday, November 21, 2019 0:35 PM
(permalink)
bp7178
chinobino.
There is nothing for EVGA to fix, the XMP profile is being correctly set but if you want the extra performance you have to manually fine tune other settings such as tRFC and tFAW.
However, I do agree that there probably needs to be another BIOS release that is fine tuned for the 9900KS (and not just for the Z390 Dark).
Problem. This "fine tuning" is only needed on EVGA boards. 500+ Gflops in Linpak isn't extra performance, its the correct performance for that clock speed, 5ghz.
I checked my DIMMs timing with Thaiphoon Burner, and the Dark is setting the correct XMP1 profile timings, so a BIOS will not change that. However, I do see Asus and Gigabyte boards with 9900K @ 5000MHz doing over 500GFlops, while my Dark will not @ 5200MHz. If the Dark was a sub-$200 board, I could understand, but it is a $500 board.
|
DEJ915
SSC Member
- Total Posts : 544
- Reward points : 0
- Joined: 11/4/2013
- Status: offline
- Ribbons : 11
Re: Low GFlops in Linpack
Thursday, November 21, 2019 0:50 PM
(permalink)
XMP only specifies a few timings, it is up for motherboard BIOS to configure the rest optimally.
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Thursday, November 21, 2019 9:40 PM
(permalink)
DEJ915
XMP only specifies a few timings, it is up for motherboard BIOS to configure the rest optimally.
Do tell.
|
DEJ915
SSC Member
- Total Posts : 544
- Reward points : 0
- Joined: 11/4/2013
- Status: offline
- Ribbons : 11
Re: Low GFlops in Linpack
Thursday, November 21, 2019 11:51 PM
(permalink)
kelkel1
DEJ915
XMP only specifies a few timings, it is up for motherboard BIOS to configure the rest optimally.
Do tell.
Here's the XMP reading from a stick on an MSI board I had handy. If you open up a timing configurator there are a lot more subtimings available. Reading up above it does seem to have the settings they were talking about though (tRFC and tFAW). I'm no memory overclocker though. 
post edited by DEJ915 - Friday, November 22, 2019 0:00 PM
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Friday, November 22, 2019 8:03 PM
(permalink)
DEJ915
kelkel1
DEJ915
XMP only specifies a few timings, it is up for motherboard BIOS to configure the rest optimally.
Do tell.
Here's the XMP reading from a stick on an MSI board I had handy. If you open up a timing configurator there are a lot more subtimings available. Reading up above it does seem to have the settings they were talking about though (tRFC and tFAW).
I'm no memory overclocker though.
Interesting, is that the actual programmed XMP profile, or what the motherboard BIOS sets?
|
Kylearan
iCX Member
- Total Posts : 288
- Reward points : 0
- Joined: 12/26/2013
- Status: offline
- Ribbons : 2
Re: Low GFlops in Linpack
Friday, November 22, 2019 10:13 PM
(permalink)
TFAW at 62? Yikes. That stick will get 350 gflops at 5 ghz also.
That is......really. bad.
TRRD_S (I'm assuming that's TRRD) of 19 is just as bad.
I have tRRD_S at 4 and TFAW should be 4xTRRD_S so I have it at 16.
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Saturday, November 23, 2019 12:06 AM
(permalink)
XMP1 sets my DIMMs at tFAW 55, and tRRD_S at 6.
|
DEJ915
SSC Member
- Total Posts : 544
- Reward points : 0
- Joined: 11/4/2013
- Status: offline
- Ribbons : 11
Re: Low GFlops in Linpack
Saturday, November 23, 2019 9:16 PM
(permalink)
kelkel1
Interesting, is that the actual programmed XMP profile, or what the motherboard BIOS sets?
From XMP, here is one in Asus board. I've not seen vendors publish these additional timing information which kind of sucks I guess if you just want to use XMP.  I need to do more reading up on memory performance increasing.
post edited by DEJ915 - Saturday, November 23, 2019 9:18 PM
|
kelkel1
SSC Member
- Total Posts : 601
- Reward points : 0
- Joined: 5/18/2019
- Status: offline
- Ribbons : 0
Re: Low GFlops in Linpack
Sunday, November 24, 2019 12:55 AM
(permalink)
|