EVGA

Hot!X299 Dark Stability Overclock Notes/Info

Page: < 1234 > Showing page 3 of 4
Author
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/09 10:09:28 (permalink)
Percentage improvements from BIOS Default settings.
 
Time Spy GPU 8% + CPU 40% = 12%
Fire Strike graphics 7% + physics 33% + combined 36% = 16%
Furmark 720 11%
Furmark 2160 10%
Cinebench 33%
Realbench image 19% + encode 32% + opencl 6% + multitask 26% = 30%
 
AIDA64 BIOS Default Cache & Memory
 

 
AIDA64 BIOS Default GPGPU
 

Attached Image(s)

#61
bcavnaugh
The Crunchinator
  • Total Posts : 38112
  • Reward points : 0
  • Joined: 2012/09/18 17:31:18
  • Location: USA Affiliate E5L3CTGE12 Associate 9E88QK5L7811G3H
  • Status: offline
  • Ribbons : 276
Re: X299 Dark Bench Notes/Info 2018/02/09 12:20:41 (permalink)
Are you able to run Aida64 Memory Stress?

Affiliate Code: E5L3CTGE12 & Associate Code: 9E88QK5L7811G3H


 
#62
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/09 13:46:32 (permalink)
bcavnaugh
Are you able to run Aida64 Memory Stress?



Have not tried.
 
Ran full 4-pass, 10+ hour, no error MemTest86 in this config.
 
Does AIDA64 memory test provide info over MemTest86?
 
Even at stock, AIDIA64 interface becomes "Not Responding" when starting stress test.  Uses all spare CPU in Task Manager.  No errors.  Starts updating graphs after 8-9 minutes.  Normal?  Has defined stop or runs forever?
#63
bcavnaugh
The Crunchinator
  • Total Posts : 38112
  • Reward points : 0
  • Joined: 2012/09/18 17:31:18
  • Location: USA Affiliate E5L3CTGE12 Associate 9E88QK5L7811G3H
  • Status: offline
  • Ribbons : 276
Re: X299 Dark Bench Notes/Info 2018/02/09 21:19:11 (permalink)
geninfo
bcavnaugh
Are you able to run Aida64 Memory Stress?



Have not tried.
 
Ran full 4-pass, 10+ hour, no error MemTest86 in this config.
 
Does AIDA64 memory test provide info over MemTest86?
 
Even at stock, AIDIA64 interface becomes "Not Responding" when starting stress test.  Uses all spare CPU in Task Manager.  No errors.  Starts updating graphs after 8-9 minutes.  Normal?  Has defined stop or runs forever?


More of can it run on your computer or does it crash.
It crashes on both my two new computer is to why I was asking.

Affiliate Code: E5L3CTGE12 & Associate Code: 9E88QK5L7811G3H


 
#64
zero5891
New Member
  • Total Posts : 2
  • Reward points : 0
  • Joined: 2016/07/03 07:16:58
  • Status: offline
  • Ribbons : 0
Re: X299 Dark Bench Notes/Info 2018/02/10 03:40:12 (permalink)

using the Ram 4000 clock as the R6A b die preset.
when changing to dark, and it is possible to overclock better?
post edited by zero5891 - 2018/02/10 03:43:05
#65
bdary
Omnipotent Enthusiast
  • Total Posts : 9143
  • Reward points : 0
  • Joined: 2008/04/25 14:08:16
  • Location: Florida
  • Status: offline
  • Ribbons : 96
Re: X299 Dark Bench Notes/Info 2018/02/10 12:01:34 (permalink)
bcavnaugh
geninfo
bcavnaugh
Are you able to run Aida64 Memory Stress?



Have not tried.
 
Ran full 4-pass, 10+ hour, no error MemTest86 in this config.
 
Does AIDA64 memory test provide info over MemTest86?
 
Even at stock, AIDIA64 interface becomes "Not Responding" when starting stress test.  Uses all spare CPU in Task Manager.  No errors.  Starts updating graphs after 8-9 minutes.  Normal?  Has defined stop or runs forever?


More of can it run on your computer or does it crash.
It crashes on both my two new computer is to why I was asking.


After reading this about AIDA64 and running the Cache & Memory benchmark, I decided to run it and see if it still works on my system (haven't ran it in awhile).  It launched, ran, and completed without a problem.  Just an FYI...


 


 
 
 
#66
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/10 13:43:22 (permalink)
zero5891
using the Ram 4000 clock as the R6A b die preset.
when changing to dark, and it is possible to overclock better?



What is "b die"?
 
Dark may clock memory higher than Apex, but I have no evidence it will.  Also depends on CPU memory controller(s).  I clocked 3200-rated RAM to 3466 without changing timing.
 
See https://forums.evga.com/FindPost/2753010 (#5).
post edited by geninfo - 2018/02/10 13:45:24
#67
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/10 13:48:04 (permalink)
bdary
 
After reading this about AIDA64 and running the Cache & Memory benchmark, I decided to run it and see if it still works on my system (haven't ran it in awhile).  It launched, ran, and completed without a problem.  Just an FYI...

 
bcavnaugh means Tools -> System Stability Test -> Stress system memory
#68
bcavnaugh
The Crunchinator
  • Total Posts : 38112
  • Reward points : 0
  • Joined: 2012/09/18 17:31:18
  • Location: USA Affiliate E5L3CTGE12 Associate 9E88QK5L7811G3H
  • Status: offline
  • Ribbons : 276
Re: X299 Dark Bench Notes/Info 2018/02/10 13:50:28 (permalink)
bdary
bcavnaugh
geninfo
bcavnaugh
Are you able to run Aida64 Memory Stress?



Have not tried.
 
Ran full 4-pass, 10+ hour, no error MemTest86 in this config.
 
Does AIDA64 memory test provide info over MemTest86?
 
Even at stock, AIDIA64 interface becomes "Not Responding" when starting stress test.  Uses all spare CPU in Task Manager.  No errors.  Starts updating graphs after 8-9 minutes.  Normal?  Has defined stop or runs forever?


More of can it run on your computer or does it crash.
It crashes on both my two new computer is to why I was asking.


After reading this about AIDA64 and running the Cache & Memory benchmark, I decided to run it and see if it still works on my system (haven't ran it in awhile).  It launched, ran, and completed without a problem.  Just an FYI...


It might be this Beta Version I am running.
It could be the Remote Sensors that was added to Aida64.
Thank you for testing.
post edited by bcavnaugh - 2018/02/10 14:22:33

Affiliate Code: E5L3CTGE12 & Associate Code: 9E88QK5L7811G3H


 
#69
bdary
Omnipotent Enthusiast
  • Total Posts : 9143
  • Reward points : 0
  • Joined: 2008/04/25 14:08:16
  • Location: Florida
  • Status: offline
  • Ribbons : 96
Re: X299 Dark Bench Notes/Info 2018/02/10 14:43:18 (permalink)
geninfo
bdary
 
After reading this about AIDA64 and running the Cache & Memory benchmark, I decided to run it and see if it still works on my system (haven't ran it in awhile).  It launched, ran, and completed without a problem.  Just an FYI...

 
bcavnaugh means Tools -> System Stability Test -> Stress system memory


Ok.  Guess I missed that.  So I just did the system memory stress test now.  Let it run about 4 minutes (quick test) just to see if I experienced any "not responding" type things.  All was good.
 
bcavnaugh
bdary
bcavnaugh
geninfo
bcavnaugh
Are you able to run Aida64 Memory Stress?



Have not tried.
 
Ran full 4-pass, 10+ hour, no error MemTest86 in this config.
 
Does AIDA64 memory test provide info over MemTest86?
 
Even at stock, AIDIA64 interface becomes "Not Responding" when starting stress test.  Uses all spare CPU in Task Manager.  No errors.  Starts updating graphs after 8-9 minutes.  Normal?  Has defined stop or runs forever?


More of can it run on your computer or does it crash.
It crashes on both my two new computer is to why I was asking.


After reading this about AIDA64 and running the Cache & Memory benchmark, I decided to run it and see if it still works on my system (haven't ran it in awhile).  It launched, ran, and completed without a problem.  Just an FYI...


It might be this Beta Version I am running.
It could be the Remote Sensors that was added to Aida64.
Thank you for testing.


You're welcome.  I also ran the memory stress test (about 4 min.) with no issue's.  Maybe it is the latest BIOS version?  I haven't upgraded to it yet.  Still running latest stable version here.


 


 
 
 
#70
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/10 17:20:41 (permalink)
bcavnaugh
 
It might be this Beta Version I am running.
It could be the Remote Sensors that was added to Aida64.
Thank you for testing.



How long do your new computers run AIDA64 memory stress test before crash?  Just AIDA64 crash?  Or whole machine?
#71
bcavnaugh
The Crunchinator
  • Total Posts : 38112
  • Reward points : 0
  • Joined: 2012/09/18 17:31:18
  • Location: USA Affiliate E5L3CTGE12 Associate 9E88QK5L7811G3H
  • Status: offline
  • Ribbons : 276
Re: X299 Dark Bench Notes/Info 2018/02/10 20:12:02 (permalink)
It never really starts, it hits 100% and then hangs, the file is aida_bench64.dll locks up.
After about 5 minutes I kill the app but it hangs on the aida_bench64.dll I have to kill this dll to end it.
This has been report as an issue. Seeing this also on my X99 rig with the meltdown spectre Bois, have not had time to go back yet.
Thanks for reviewing.
 

Affiliate Code: E5L3CTGE12 & Associate Code: 9E88QK5L7811G3H


 
#72
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/13 20:47:59 (permalink)
Ran into problem maintaining overclock moving into case with cooler temps.
Asked for help in overclocking forum.
Appreciate anyone who takes a look.
#73
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/16 07:29:28 (permalink)
Dark BIOS 1.05 not stable with memory overclock.
 
Installed BIOS 1.05 on BIOS slot 1.
Attempts to use XMP profile and memory overclocks working with BIOS 1.06 on BIOS slot 2 failed.
XMP profile overclock crashed in Time Spy demo immediately.
Over XMP profile overclock required CMOS reset to get into BIOS!
Fairly sure XMP profile worked on BIOS 1.04.
 
No problems in 1.06, but has Spectre fix Intel no longer recommends.
 
Consider not using 1.05 if having issues when overclocking memory.
 
 
 
#74
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/16 08:12:52 (permalink)
Mesh Ratio overclocking troublesome; CPU frequency and memory overclock more important.
 
Increasing Mesh Ratio from Auto (30) to 32 created heat and instability. #54
Never built good evidence for stable overclock with Mesh Ratio 32.
Mesh Ratio Auto (30) already an overclock (over 24?).
 
Refocused on memory overclocking without 32X Mesh Ratio.
Memory kit rated at 3200 16-18-18-36.
3466 16-18-18-36 worked with no other changes.
3600 16-18-18-36 VDIMM1/2 1.4V in Prime95 4096K, 60,000 MB RAM, 5 minute cycles failed with rounding error.
3600 16-18-18-36 VDIMM1/2 1.45V in same test passed.
 
Setting Uncore to 250 now lowers actual Uncore by about 50mV, to about 1.2V.
Uncore 250 used with 3600 16-18-18-36 memory overclock.
 
Best Time Spy CPU score with mesh ratio 32 and 3466 16-18-18-36: 13,217
Best Time Spy CPU score with mesh ratio Auto (30) and 3600 16-18-18-36: 13,316
AIDA64 5% increase in memory read, and slight improvement in latency.
 
Better thermals and stability with just change to VDIMM instead of several CPU voltage changes for mesh ratio.
 
Candidate stable configuration now:
Fractal Define R6 Case
Titan V (Driver 390.77)
BIOS 1.06 (on BIOS2)
Vcore: Override 1.175
CPU Multiplier: 45x
AVX2 Negative Offset: -5 (Untested)
AVX3 Negative Offset: -7 (Untested)
Memory: XMP Profile 1, Frequency 3600, VDIMM 1.45
VCCIO: 1.166
Uncore: 250
Fast Boot: Off
Force Memory Retraining: Enabled
 
VCore at 1.175 to see if WHEA errors will go away without Mesh Ratio overclock. #57
 
EDIT: VCore originally listed as 1.75. No! Bad! Do not do. :)
 
AIDA64 Cache & Memory At 3600 16-18-18-36 Memory OC

post edited by geninfo - 2018/02/19 07:25:19

Attached Image(s)

#75
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/19 07:22:04 (permalink)
Consider SPECwpc benchmark for testing  workstation builds on Dark.
 
Tests wide variety of professional workstation applications, CPU, and GPU.
Different set of tests than usual gaming benchmarks.
Good for spot checking stability of high end desk top (HEDT) workstation builds?
 
Dark fully stable on SPECwpc benchmark in candidate final stable configuration described above.
Tested with Standard NVMe driver (see below).
 
SPECwpc Results (Candidate Final Stable Configuration)
 

 
SPECwpc Monitoring Graphs (188 Minute Run)
 

 
SPECwpc Additional Monitoring Graphs (188 Minute Run)

 
 
 
post edited by geninfo - 2018/02/21 10:13:38

Attached Image(s)

#76
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/19 21:14:47 (permalink)
Use Samsung NVMe Driver if you use Samsung 960 Pro (1TB) on Dark; standard Windows NVMe driver has terrible write performance!
 
Tried diskspd for I/O testing.
diskspd has command line interface; good text reports.
Ran many quick tests with Standard Windows NVMe driver.
Installed Samsung NVMe driver; retested.
 
Results below show average MB/s and IO/s for different tests.
All tests random read/write.
Samsung 960 Pro 1TB firmware 2B6QCXP7 at time of testing.
 
Horizontal axis from top down:
Percentage Write = 0%/100%
Block Size = 4K
IO queue depth = 1/32
Threads = 1/4
Driver = Samsung Driver (SAM)/Standard Windows NVME Driver from Microsoft (MS)
 
Random Read/Write Performance By Driver, Average MB/s
 

 
Random Read/Write Performance By Driver, Average IO/s

 
Samsung driver better in all tests to 99% statistical confidence.
Read speeds difference not large.
Write speeds difference huge.
So huge, uninstalled Samsung driver to double check with MS driver.  Confirmed.
 
Update
Recommendation remains; results generally hold.
SSD testing may require more subtlety.
MS driver has strange results when testing large sequential reads (not covered above).
Test files created with diskspd read back at higher speeds, even after restart, until used for write tests.
May also happen with Samsung driver (untested).
Suspect internal SSD optimization.
Meaningful tests need full specification including creation of test file plus subsequent operations on it.
Raw tests, instead of file systems tests, preferred by review sites,
But raw tests not representative of true perf with file system and file history.
post edited by geninfo - 2018/02/21 03:43:57

Attached Image(s)

#77
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/21 04:08:23 (permalink)
Samsung 960 Pro (1 TB) firmware may affect performance slightly; still update it.
 
Old firmware 2B6QCXP7.
New firmware 4B6QCXP7.
Retested 4K random read, 1 thread, 1 IO depth.
Very minor difference in MBps and IOps at 99% statistical confidence.
MS driver average (five run) ~-2% both MBps and IOps.
Samsung driver average (five run) ~-2% MBps and ~-1% IOps.
Review sites rarely publish firmware data; check when comparing results.
Caveats on SSD testing subtlety remain.
post edited by geninfo - 2019/11/06 17:26:30
#78
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/02/22 19:44:57 (permalink)
Can we really use all those cores? Yes!
 
Editing session with Affinity Photo.
Real use, not benchmark.
Affinity Photo full multicore application.
High resolution file, complicated layers flipping on/off.
Movement smooth; no lag on visual changes.
Much better response than previous machine (6700K 4.6Ghz OC).
 
Affinity Photo Cores In Use

Attached Image(s)

#79
GGTV-Jon
FTW Member
  • Total Posts : 1811
  • Reward points : 0
  • Joined: 2017/11/25 14:11:43
  • Location: WA, USA
  • Status: offline
  • Ribbons : 18
Re: X299 Dark Bench Notes/Info 2018/03/04 00:32:26 (permalink)
@geninfo you may hate me for this
Fuel for thought / the fire for memory tweaking
 
Not that I am anywhere close parts wise to having a full system to boot yet - but doing reading / research
 
Part of this was kicked off by Buildziod's recent video "DDR4 Overclocking/Buying absolute basics" https://youtu.be/Ke_JgL8gxsA
Happy Note - Buildzoid now has an X299 Dark in his possession - looking forward to some OC videos from him with it
That sent me digging to find out who made the chips on the memory I bought - G.Skill F4-3733C17Q-32GTZR which was listed for the Dark on both EVGA and G.Skill. Last nights digging = Samsung
 
As for overclocking the memory I was poking around the "OC Guide for EVGA X299 DARK motherboard" at https://xdevs.com/guide/e299ocg/#fanpump hoping to see any bios screen shots of the memory section. Did a google search for X299 Memory Training to get a better idea as to it use and implementation and found this gem of an article that was made back in May of 2017
 
"Comprehensive Memory Overclocking Guide" https://linustechtips.com/main/topic/773966-comprehensive-memory-overclocking-guide/
 
By the sounds of it geninfo if you are using memory settings pretty much as the XPM profile sets them there may be room for your AIDA64 scores to improve.
#80
GGTV-Jon
FTW Member
  • Total Posts : 1811
  • Reward points : 0
  • Joined: 2017/11/25 14:11:43
  • Location: WA, USA
  • Status: offline
  • Ribbons : 18
Re: X299 Dark Bench Notes/Info 2018/03/04 01:37:35 (permalink)
geninfo
zero5891
using the Ram 4000 clock as the R6A b die preset.
when changing to dark, and it is possible to overclock better?



What is "b die"?
 
Dark may clock memory higher than Apex, but I have no evidence it will.  Also depends on CPU memory controller(s).  I clocked 3200-rated RAM to 3466 without changing timing.
 
See https://forums.evga.com/FindPost/2753010 (#5).




@geninfo not sure if you ever figured out what he was referring to with the "b" die part
 
In the video from Buildzoid I linked above it is somewhat explained
 
Link will start at 2:30 - meat of it is at 2:46ish - watch through 3:40is
https://youtu.be/Ke_JgL8gxsA?t=2m30s
 
 
At this time I still do not know the exact specs like what was mentioned in the video for the "e" die for the "b" die
 
While typing out this post I found this - https://www.reddit.com/r/..._bdie_confusion_eg_on/
I was able to find specs that should match my memory, did not see your corsairs. Perhaps run the program they have listed to view the nitty gritty on your chips
 
Edit note:
After re watching sections of Buildzoids video and looking at the reddit page I just linked too b die just means they are 8gb units, e die mentioned in the video are 4gb units
post edited by GGTV-Jon - 2018/03/04 02:56:40
#81
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/03/04 18:25:18 (permalink)
CLC 280 with extra fans reduces noise and temp.
 
CLC 280 front intake; two included fans pushing through radiator.
Added two fans pulling through radiator.
Wired each front/back fan pair to same fan header (CPU1 and CPU2) with 4-pin PWM splitter cable.
Added rear exhaust case fan and bottom intake case fan.
Total 5 case fans, 1 bottom intake, 4 exhaust (3 top, 1 rear).
 
Result
Reran previous test of Prime 56K FFT-in-place 5 minute test cycle for 15 minutes.
Similar liquid and ambient temp at start.
After 15 minutes, comparisons before and after pull fans added:
-5C CPU temp (87C -> 82C)
-6C coolant temp (45C -> 39C)
-5C VRM temp (56C -> 51C)
 
Overall sound at 100% fan slightly lower even though 2 more fans.
Sound changed tone to low howl, not loud white noise.
 
Install
EVGA fans come with short case screws.
EVGA fans do not come with long radiator attachment screws.
EVGA fans come with 8x rubber anti-vibration washers each.
EVGA CLC comes with two fans, 8 long radiator attachment screws, 8 metal washers.
 
Use 4 radiator screws for push fan and case radiator attachment in offset pattern:
X
  X
X
  X
 
Use other 4 radiator screws for pull fans in mirrored offset pattern on back.
 
Use rubber washers on front like:
Screw Head -> Metal Washer -> Rubber Washer -> Case -> Rubber Washer -> Fan -> Radiator
 
User rubber washers on back like:
Screw Head -> Metal Washer -> Rubber Washer -> Fan -> Rubber Washer -> Radiator
 
Use rubber washer sticky pad remnants on case and radiator over screw holes not filled.
X = screw, P = pad:
XP
PX
XP
PX
 
Combine
Delid (estimates 15-25C)
Liquid metal cooler paste mod (11C) - see #33 earlier in thread
Push/pull fans (5C)
Overall 31-41C thermal control gain.
= 7980XE 4.5Ghz 18 Cores at 82C extreme load with AIO no custom loop or exotic.
post edited by geninfo - 2018/04/13 17:20:12
#82
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/03/04 18:43:30 (permalink)
GGTV-Jon
Part of this was kicked off by Buildziod's recent video "DDR4 Overclocking/Buying absolute basics" https://youtu.be/Ke_JgL8gxsA
Happy Note - Buildzoid now has an X299 Dark in his possession - looking forward to some OC videos from him with it
That sent me digging to find out who made the chips on the memory I bought - G.Skill F4-3733C17Q-32GTZR which was listed for the Dark on both EVGA and G.Skill. Last nights digging = Samsung
 
As for overclocking the memory I was poking around the "OC Guide for EVGA X299 DARK motherboard" at https://xdevs.com/guide/e299ocg/#fanpump hoping to see any bios screen shots of the memory section. Did a google search for X299 Memory Training to get a better idea as to it use and implementation and found this gem of an article that was made back in May of 2017
 "Comprehensive Memory Overclocking Guide" https://linustechtips.com/main/topic/773966-comprehensive-memory-overclocking-guide/ By the sounds of it geninfo if you are using memory settings pretty much as the XPM profile sets them there may be room for your AIDA64 scores to improve.



Buildzoid took 25 minutes to say this:
https://imgur.com/MhqKfkf (https://www.reddit.com/r/buildapc/comments/5d009u/i_made_a_simple_table_to_help_you_pick_out_what/)
 
I have Corsair Dominator 3200 16-18-18-36 overclocked stable to 3466 at 1.35V, same timings.
At 3600 at 1.45V it fails Memtest at one address on Test 10 (bit fade ) and Test 13 (hammer).
 
G.Skill F4-3200C14Q-64GVK looks like best performance latency/frequency combined.
Also on supported memory list for Dark.
post edited by geninfo - 2018/03/04 18:46:59
#83
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/03/04 18:46:23 (permalink)
GGTV-Jon
@geninfo not sure if you ever figured out what he was referring to with the "b" die part
 
In the video from Buildzoid I linked above it is somewhat explained
 
Link will start at 2:30 - meat of it is at 2:46ish - watch through 3:40is
https://youtu.be/Ke_JgL8gxsA?t=2m30s
 
 
At this time I still do not know the exact specs like what was mentioned in the video for the "e" die for the "b" die
 
While typing out this post I found this - https://www.reddit.com/r/..._bdie_confusion_eg_on/
I was able to find specs that should match my memory, did not see your corsairs. Perhaps run the program they have listed to view the nitty gritty on your chips
 
Edit note:
After re watching sections of Buildzoids video and looking at the reddit page I just linked too b die just means they are 8gb units, e die mentioned in the video are 4gb units




I don't think b-die refers to fixed sizes.
I think b-die refers to tier of memory based on manufacture process.
Buildzoid and other information suggests 3200 RAM with CAS 14 indicates Samsung b-die.
Corsair Dominator 3200 RAM with CAS 16 probably not Samsung b-die.  Hynix?
#84
GGTV-Jon
FTW Member
  • Total Posts : 1811
  • Reward points : 0
  • Joined: 2017/11/25 14:11:43
  • Location: WA, USA
  • Status: offline
  • Ribbons : 18
Re: X299 Dark Bench Notes/Info 2018/03/04 18:50:04 (permalink)
That thaiphoon burner will show you that
#85
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/03/04 18:56:40 (permalink)
GGTV-Jon
That thaiphoon burner will show you that



Your message cutoff?
LOL "taiphoon burner" refers to CLC 280 mod? 
#86
GGTV-Jon
FTW Member
  • Total Posts : 1811
  • Reward points : 0
  • Joined: 2017/11/25 14:11:43
  • Location: WA, USA
  • Status: offline
  • Ribbons : 18
Re: X299 Dark Bench Notes/Info 2018/03/04 19:04:02 (permalink)
Was posting from my phone there, made me have to get on the computer :P
 
So, what I was trying to say is that the Thaiphoon Burner (http://www.softnology.biz/files.html) that was mentioned in the reddit link I posted above will show you the deep down info on your memory sticks
 
Example I am grabbing from the database on that reddit article 

 
 
 
Edit note - the image above was not intentional - random coincidence that it matches the G.Skill F4-3200C14Q-64GVK you posted above in speed and CAS
post edited by GGTV-Jon - 2018/03/05 01:26:38
#87
cateno
SSC Member
  • Total Posts : 974
  • Reward points : 0
  • Joined: 2009/10/18 01:32:46
  • Location: Belgium
  • Status: offline
  • Ribbons : 3
Re: X299 Dark Bench Notes/Info 2018/03/08 21:11:47 (permalink)

X299 DARK  7900X 7740x soon 7980X
X299 APEX
X299 rampage Extreme
 
E762 W3520@4.2  E679 2600k
E760 X980 
E761 I920
E770 I950
and SR2's  and SRX's
Z87 classified , MVIE built with 4770K
Z97 classified , ftw Z97 ,  MFVII  only wait for gigabyte LN2 
X99 series.... classified , FTW , RE5 , oc formula, profesional , X11 , soc force , msi gaming9ack

X79 classified, dark and RIVE, RIVBE,, xtreme11
and many classified a panel from EVGA  no just one sample

and other  DFI giga etc  

#88
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/03/25 12:53:18 (permalink)
Titan V requires companion video card for intended compute role.
 
Explain
Titan V has 2x DisplayPort and 2x HDMI output.
Titan V can operate in WDDM mode.
Titan V can operate in TCC mode.
 
WDDM mode is Windows Display Driver Model.
TCC mode is Tesla Compute Cluster mode.

TCC mode operates under Windows Driver Model (WDM).
 
Titan V must operate in WDDM mode if any display physically connected.
Titan V in WDDM mode becomes Windows desktop and render zombie.
Titan V in TCC mode remains companion internal computer accessible via Windows driver.
 
CUDA legacy debugger can work with some video cards in WDDM mode.
CUDA legacy debugger only works with Pascal architecture and earlier.
Titan V Volta architecture later than Pascal.
CUDA next-gen debugger can work with Volta.
CUDA next-gen debugger only works with cards in TCC mode.
 
nvidia-smi provides info and mechanism to change modes (if no display connected).
 
Result
Titan V must operate in TCC mode to support CUDA debugging.
Titan V must operate in TCC mode to get maximum compute (including tensor core) performance.
Titan V in WDDM mode can still render and debug graphics rendering.

Refer
CUDA Driver Models 
Launching CUDA Debugger
Tesla Compute Cluster (TCC) 
Titan V TCC vs. WDDM Performance 
nvidia-smi 

Imply
This configuration not useless after all!
post edited by geninfo - 2019/11/18 09:59:27
#89
geninfo
Superclocked Member
  • Total Posts : 191
  • Reward points : 0
  • Joined: 2017/12/18 11:22:15
  • Status: offline
  • Ribbons : 2
Re: X299 Dark Bench Notes/Info 2018/04/13 08:46:42 (permalink)
AMD Radeon WX Pro 7100 graphics card works on X299 Dark with Titan V. 
NVIDIA's Titan strategy helps AMD sell workstation cards. :)
 
Enable
4x DisplayPort to monitors.
NVIDIA Titan V in TCC compute mode no displays attached (tested, see #89).
CUDA next-gen debugger (enables simultaneous breakpoints in CPU and GPU code, tested).
10-bit output for OpenGL surfaces (like Photoshop, not tested).
Windows Mixed Reality (Cliff House smooth) with active DisplayPort 1.2 to HDMI 2.0 adapter (tested).
Switching to Titan V if not running CUDA apps and need high render performance (not tested with 7100 active).
Cross test with OpenCL and AMD technologies if desired (not tested).
HDR output to HDR monitor (tested).
 
Config
AMD Radeon WX Pro 7100 in PE1 (x16 PCIe CPU).
NVIDIA Titan V in PE4 (x16 PCIe CPU).
Plugable DisplayPort to HDMI 2.0 Active Adapter (~$20 USD).
 
Difficult
Had freezing problems installing AMD Radeon WX Pro drivers.
win10-64bit-radeon-pro-software-enterprise-18.q1.1-march22.exe
May have helped (uncertain resolution):
Uninstalling NVIDIA graphics/audio/physx drivers (but NVIDIA Control Panel remained - how to get rid of?) and reinstalling after AMD drivers installed.
Attaching one (of three) monitor to WX Pro 7100 DisplayPort closest to PCIe pins.
Launching driver executable from local disk instead of network share.
Not moving mouse during AMD install until prompted.
Express versus custom install.
Running "C:\Program Files\AMD\CIM\BIN64\AMDCleanupUtility.exe" after partial installation.
 
Recommend
Connect primary monitor to WX Pro 7100 port furthest away from PCIe pins.
WX Pro 7100 shows BIOS and start screen on monitor on that port plus port next to it.
For brand new OS, install AMD Radeon WX Pro 7100 and drivers first.
Then install Titan V and NVIDIA drivers. 
Consider WX Pro 9100 if top performance needed on primary GPU in addition to Titan V.
But WX Pro 9100 $1600 USD versus WX Pro 7100 $570 USD.
 
Oops
NVIDIA chose Geforce instead of Quadro for Titan V.
Quadro drivers and Geforce drives cannot cohabitate.
Forces purchase of AMD workstation card for primary GPU in workstation with Titan V.
Could use another Geforce if not wanting workstation card, but Titan V ~= workstation.
post edited by geninfo - 2018/04/13 17:10:32
#90
Page: < 1234 > Showing page 3 of 4
Jump to:
  • Back to Mobile