My 2990WX

Message boards : Number crunching : My 2990WX
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 7 · Next

AuthorMessage
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1965795 - Posted: 17 Nov 2018, 12:00:25 UTC

I bought a AMD Threadripper 2 2990WX and with the a great deal of help and support from Keith Myers (a guy with a LOT of high end AMD systems) have reached the point where I am trying to find the sweet spot between maximum OverClock speed and system stability.

I think the only symptom I haven't experienced (yet) and hope to never experience is the thing melting down and/or going up in smoke!

Right now I am trying to see if it is possible to run it at 4GHz on a "manual" overclocking w/o having it lock up. It has now run 6+ hours of Seti computation w/o locking up. A new record at 4GHz for this system (I think).

If it runs 24 hours continuously I can stop wondering if I will need to drop the OC back to 3.9GHz which was my last documented stability point.

I have two things I want to do in the near term. I just got a replacement gtx 1060 3GB that I will want to install. And I think I can bump the CL up one step (via the bios memory presets) without messing up anything.

But as Brent Norman has pointed out, I need to wait to do either of those things until the system has run at least 7 days without stopping.

What was the last two counter intuitive changes? I reset the Last Line Control in the bios back to "auto" and in the previous step I had set the CPU V core back to "auto".

As explained to me, normally you need to manually increase the V core voltage to support running the CPU at higher than baseline speeds. And normally you need to set the Last Line Control to a more aggressive position than "auto" in order to keep the V core voltage from dropping (aka: sagging) while the system is under heavy load.

I was trying to follow this advice and the system was crashing when Keith suggested I switch everything back to automatic and only change the "main thing" which is the CPU multiplier. I have already tried switching the V core back to "auto" with the LLC set to Mode4. But it locked up again. :(

However when I switched the LLC back to "auto" and booted the machine, it came up and has been happily crunching Seti tasks w/o interruption since about 11pm last night. Its about 6pm this morning and I am using the 2990WX to post this message.

THANK YOU Keith!!!

Tom
A proud member of the OFA (Old Farts Association).
ID: 1965795 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1965890 - Posted: 17 Nov 2018, 18:34:19 UTC - in response to Message 1965795.  
Last modified: 17 Nov 2018, 18:34:48 UTC

Sigh. This morning at about 6am local time I shut down all the non-seti stuff (browser, leafpad, 3 terminal sessions with 2 doing active things) and went back to bed. When I got up, it was "non-responsive" and required a cold boot to get it running again. I failed to look at the MB diagnostic LED thing to see what it was displaying.

I am going to leave all the "non-seti" stuff running and see if it "makes a difference" again or was it just experimental randomness. While I don't have a problem with leaving all the "other stuff" running, it does bother me if in fact that makes a difference between becoming non-responsive and staying responsive.

While the mouse, keyboard and monitor are all displaying symptoms of "non-connection" (no lights) I am wondering if the cpu/gpu are still crunching along or not. I need to examine the MB Led next time I try to "re-create" this non-responsiveness. It is showing AA while running right now.

Tom
A proud member of the OFA (Old Farts Association).
ID: 1965890 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1965894 - Posted: 17 Nov 2018, 19:14:14 UTC
Last modified: 17 Nov 2018, 19:21:57 UTC

If you are interested in a lot of details on how I got this far it is located at: https://setiathome.berkeley.edu/forum_thread.php?id=83457

I am running a MSI X399 Gaming Pro Carbon AC motherboard with a Amd 2990WX cpu that I got on eBay at a slight discount (new unused with none of the packaging).

I am running a "Enermax Liqtech TR4 II 280 Addressable RGB AIO Liquid CPU Cooler, Support 500W+TDP, Overclocking, AMD Socket TR4" which promptly died and was exchanged on an RMA. Keith reports that this model has a history of dieing early and often. I would have moved up to the 360 model he recommended but my PC Case isn't big enough. I may have to move up anyway :( unless this one lasts longer. Keith recommended a backup air cooler that is now on my wish list at NewEgg. If/when this one dies, I will be able to get back up running at a slower pace while I decide what to do.

I started with DDR4-2666 CL 16-16-16-36 1.2 volts. (4X4GB) Upgraded the ram to DDR-3200 CL 16-18-18-18-38 1.2 volts. (4X8GB) Upon further discussion by Keith it appears I made a mistake in identifying what I needed to upgrade to. I needed to upgrade to a CL 14 kit preferably a "Samsung die B" which is relabel and sold under many brand names. In all cases it needs to be 4 sticks so I can run it as Quad Channel.

I have been having excellent luck running used Gtx 1060 3GB video cards which I get off eBay for $150 and when run with Linux/CUDA91 (Thank you Petri & Tbar) usually process a gpu task in around 3 minutes or less.

I you have any questions about this hardware please let me know.

Thank you,
Tom
A proud member of the OFA (Old Farts Association).
ID: 1965894 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1965895 - Posted: 17 Nov 2018, 19:17:26 UTC - in response to Message 1965890.  

Sigh. This morning at about 6am local time I shut down all the non-seti stuff (browser, leafpad, 3 terminal sessions with 2 doing active things) and went back to bed. When I got up, it was "non-responsive" and required a cold boot to get it running again. I failed to look at the MB diagnostic LED thing to see what it was displaying.

I am going to leave all the "non-seti" stuff running and see if it "makes a difference" again or was it just experimental randomness. While I don't have a problem with leaving all the "other stuff" running, it does bother me if in fact that makes a difference between becoming non-responsive and staying responsive.

While the mouse, keyboard and monitor are all displaying symptoms of "non-connection" (no lights) I am wondering if the cpu/gpu are still crunching along or not. I need to examine the MB Led next time I try to "re-create" this non-responsiveness. It is showing AA while running right now.

Tom


Well it became "non-responsive" again shortly after the above post and while I was in the middle of the post that followed it.

Yes, the LED on the MB went from AA to 00.

Did a cold boot and backed the cpu multiplier down one step to "39.75" and so far, nothing has crashed (I would cross my fingers except I am also trying to type :)

Tom
A proud member of the OFA (Old Farts Association).
ID: 1965895 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1965904 - Posted: 17 Nov 2018, 19:50:58 UTC
Last modified: 17 Nov 2018, 20:09:06 UTC

My goal for this system is to process cpu tasks as fast as I can with the most active cores that I can.

The problem is this cpu has 16 cores with direct access to the memory chips and 16 cores that require a "second hop" to access memory cores. So the CCX (if I am not confused that is AMD's name for that part of the cpu) cores are running an application that can fit into the cpu cache all is fine and dandy otherwise the same app will run significantly slower.

If that wasn't enough apparently the internal communications channels can get congested by needing a lot of those 2 hop memory chip accesses. This slows the production level of the processing of the entire cpu. The only short term fix hardware is memory that is both faster and has a lower access latency. Hence the reference to the CL 14 Samsung Die-B ram in the other message.

The other "fix" is to locate the "sweet spot" between faster processing and using more cores.

My current settings replicate an Intel dual cpu e5-2690v2 that turbo boosts to just short of 3.3GHz. I am not slowing the AMD to 3.3Ghz :)

I am running the Intel at 90% of the cores which is 36 threads. I am running it without an "app_config.xml" so that the gpu's use whatever the system default for driving the cores is. This is usually less than one core per gpu task. Gtx 1030 3GB's only run 1 task at a time under any OS/App combination I have tried (Windows, Windows Lunatics distro & Linux/CUDA91). I am also running the gpu's without a parameter called "-nobs" which when enabled will cause a cpu to be devoted to that gpu tasks care and feeding.

The cpu tasks are mostly processing at about 1 hour 6 minutes to 1 hour 12 minutes with the rare slower speed. I am not sure if I have caught any running faster than that.

The AMD 64 thread machine is also currently running at 36 threads. It has not been running stable enough lately to begin to bring the "estimated time to completion" down below 1 hour and 45 minutes. I am getting more tasks taking about an hour and "several" have been taking as low as 40 minutes.
I am hoping that after I find a stable spot for the cpu speed that I can get some faster than Intel e5 results.

Based on wide community opinion as well as Keith and Brent's input it is going to take weeks to 1) Prove I have actually got this system to a stable point 2) Get a steady baseline of cpu task performance.

I am still wondering what the actual sweet spot for speed of processing vs. maximum cores is going to be. I read an article that was using a repeatable scientific processing test and he got a result of about 46 cores.

The wildcard here is that if the CPU is too heavily loaded the GPU processing slows significantly. So my final mix will be the highest # of cores that will maintain full GPU processing speed.

There are a couple of people on the individual computer LeaderBoard who are grappling with this specific question of speed/number of cpu tasks vs. not slowing the gpu processing down.
A proud member of the OFA (Old Farts Association).
ID: 1965904 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1965905 - Posted: 17 Nov 2018, 20:05:34 UTC - in response to Message 1965904.  

The wildcard here is that if the CPU is too heavily loaded the GPU processing slows significantly. So my final mix will be the highest # of cores that will maintain full GPU processing speed.

There are a couple of people on the individual computer LeaderBoard who are grappling with this specific question of speed/number of cpu tasks vs. not slowing the gpu processing down.


The LeaderBoard for individual systems is here: https://setiathome.berkeley.edu/top_hosts.php

The individual systems are here
https://setiathome.berkeley.edu/show_host_detail.php?hostid=8528566
This system is running something like 26 threads. With SMT (aka: hyper threading) engaged. I think his cpu is clocking at 4GHz.

https://setiathome.berkeley.edu/show_host_detail.php?hostid=8365846
I am not certain how many threads this system is running but he specifically turned off SMT and maybe running the cpu at something like 3.3GHz,

Both of these systems are very sensitive to their GPU processing speeds because that is where their major Seti production is coming from.

Tom
A proud member of the OFA (Old Farts Association).
ID: 1965905 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1965994 - Posted: 18 Nov 2018, 6:14:01 UTC

Came home from work and it was still crunching. Happy. Happy. :)
Tom
A proud member of the OFA (Old Farts Association).
ID: 1965994 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1966640 - Posted: 22 Nov 2018, 20:21:38 UTC - in response to Message 1965994.  

Unfortunately after burning nearly two days of my "weekend" (I have Tuesday/Wednesday off) I believe I have hit a incorrect memory hardware issue I have alluded to previously.

It looks like I can't Over Clock past 3.7GHz until I get the right Samsung die-B memory.

While waiting for my Christmas savings to catch up with the cost of Samsung die-B memory (CL14/3200/4X8GB apparently) I have been analyzing what core count would be the most productive with the Threadripper CPU architecture.

Since half the cpus have direct access to MB memory and the other half access the MB memory through a memory controller chip (1 hop vs. 2 hop's) then if I am running more than 32 threads/tasks the 33rd (and higher) task is going to be "memory bound" in comparison to the cpus with direct access.

Results so far seem to show 36 cores may have an average "time remaining (estimated)" of 1 hour and 2 minutes for the mix of seti cpu tasks coming out last week. Those were mostly blc4 type tasks (I think). I was crunching at least 3.7GHz. Perhaps higher. My research notes are not clear. :(

46 cores with a slower task mix seem to be running about 1 hour 51 to 57 minutes (current estimated time).

I was running 56 cores earlier (about 90% of available cores). I was having trouble with system lockups/crashes and got impatient. But I was getting tasks apparently running upto 3+ hours. And very little running under an hour and 15 minutes.

I need to revisit the 56 core count running 3.7GHz and see what it settles down to.

I read a posting about a 2990WX running Linux on a scientific application. He re-ran it, systematically increasing the # of cores/threads he was using under SMT (aka: Hyper Threading). His "sweet spot" was 46 cores/threads.

That is why I am determined to test 46 cores/threads. It might be the sweet spot for Seti CPU processing on a Linux-based 2990WX.

Tom
A proud member of the OFA (Old Farts Association).
ID: 1966640 · Report as offensive
Kevin Olley

Send message
Joined: 3 Aug 99
Posts: 906
Credit: 261,085,289
RAC: 572
United Kingdom
Message 1966709 - Posted: 23 Nov 2018, 4:31:47 UTC - in response to Message 1965905.  


The individual systems are here
https://setiathome.berkeley.edu/show_host_detail.php?hostid=8528566
This system is running something like 26 threads. With SMT (aka: hyper threading) engaged. I think his cpu is clocking at 4GHz.

Tom


25 threads, 21 cpu + gpu's, and yes it is set at 4GHz

Increasing the number of CPU tasks does cause a slowdown on GPU's and will increase the running time of some CPU tasks.
Kevin


ID: 1966709 · Report as offensive
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1966755 - Posted: 23 Nov 2018, 14:55:01 UTC

LLC = Load Line Calibration. not Last Line Control.

it helps prevent Vcore drop under load which can cause instability of the OC. you have to be careful with it though, it can cause overvolting if you set it too high.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1966755 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1966776 - Posted: 23 Nov 2018, 18:01:15 UTC - in response to Message 1966709.  


The individual systems are here
https://setiathome.berkeley.edu/show_host_detail.php?hostid=8528566
This system is running something like 26 threads. With SMT (aka: hyper threading) engaged. I think his cpu is clocking at 4GHz.

Tom


25 threads, 21 cpu + gpu's, and yes it is set at 4GHz

Increasing the number of CPU tasks does cause a slowdown on GPU's and will increase the running time of some CPU tasks.


So far, if I use -nobs the gpus haven't slowed down at all even at 56 cores. Unfortunately my cpu processing speed drops considerably.

I am currently running 46 cores at 3.7GHz using the builtin OC control plus disabling the "cool&quiet" which is aka as "speedstep". So far, the tasks seem to be running in between 1 hour 20 minutes and something over 2 hours. I want to run it long enough to see if the "estimated run-time" goes below 1 hour 56 minutes.

I am trying to raise the money for the ram I should have bought the first (and 2nd) times. Then I will try for the glory of 4MHz :) Again.

Tom
A proud member of the OFA (Old Farts Association).
ID: 1966776 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1966777 - Posted: 23 Nov 2018, 18:07:25 UTC - in response to Message 1966755.  

LLC = Load Line Calibration. not Last Line Control.

it helps prevent Vcore drop under load which can cause instability of the OC. you have to be careful with it though, it can cause overvolting if you set it too high.


Sigh. It seems my memory goes fuzzy and then I spit out an acronym name that doesn't even make sense.

Thank you for furthering my education. I didn't know that it could also cause "overvolting" as well as deal with the Vcore "voltage sag" under load. Come on savings.

Come on savings for a CL14/3200/4X8GB ram kit.

"I want more patience and I want it RIGHT NOW" ;)

Tom
ps. I just sold my original ram for the price I paid for it. Now all I have to do is raise another 100% of that and I can buy a new ram kit from NewEgg for about $400 :(
A proud member of the OFA (Old Farts Association).
ID: 1966777 · Report as offensive
Profile StFreddy
Avatar

Send message
Joined: 4 Feb 01
Posts: 35
Credit: 14,080,356
RAC: 26
Hungary
Message 1966789 - Posted: 23 Nov 2018, 18:39:34 UTC - in response to Message 1965894.  


I started with DDR4-2666 CL 16-16-16-36 1.2 volts. (4X4GB) Upgraded the ram to DDR-3200 CL 16-18-18-18-38 1.2 volts. (4X8GB) Upon further discussion by Keith it appears I made a mistake in identifying what I needed to upgrade to. I needed to upgrade to a CL 14 kit

as far as I read on different forums, for ddr 3200MHz, 1.35V is recommended on Ryzen systems. Not sure about threadripper, but I think that 1.2V is simply not enough.

On the other hand, you are modifying too much parameters at once. Put your CPU back to its factory default settings and test RAM stability first. Use memtest86. If that's stable, move forward with the CPU.
BUT, keep in mind that your CPU is YERY power hungry. I am not sure if your motherboard can handle that amount of power if you overclock. Just run it on default settings for a few days. After that, increase 100MHz, then test for a few days.

I wouldn't try that 4.0GHz all core overclock. your motherboard's VRM can be a weak point.
ID: 1966789 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1966840 - Posted: 24 Nov 2018, 5:28:18 UTC - in response to Message 1966789.  


I started with DDR4-2666 CL 16-16-16-36 1.2 volts. (4X4GB) Upgraded the ram to DDR-3200 CL 16-18-18-18-38 1.2 volts. (4X8GB) Upon further discussion by Keith, it appears I made a mistake in identifying what I needed to upgrade to. I needed to upgrade to a CL 14 kit

as far as I read on different forums, for ddr 3200MHz, 1.35V is recommended on Ryzen systems. Not sure about threadripper, but I think that 1.2V is simply not enough.

On the other hand, you are modifying too much parameters at once. Put your CPU back to its factory default settings and test RAM stability first. Use memtest86. If that's stable, move forward with the CPU.
BUT, keep in mind that your CPU is YERY power hungry. I am not sure if your motherboard can handle that amount of power if you overclock. Just run it on default settings for a few days. After that, increase 100MHz, then test for a few days.

I wouldn't try that 4.0GHz all core overclock. your motherboard's VRM can be a weak point.


Since I have at least one example of a Seti processing 2990WX running 4GHz I think it is a worthy goal :)

Too many people tell me "I am modifying too many parameters at once" when they have no idea exactly what I have been switching or not switching. I am up to 4 pages of painstaking one change at a time of notes over the last couple of weeks. :( I am approaching more than a 100 one change at a time combinations. I have more of an idea what doesn't work at some bios parameters that I have been stepping one notch at a time than what does work.

I am running with the memory I have until I can get the memory I need. Since it says "1.2 volts" on the side of the chip I am more than a little leery trying anything higher.

Once I have the ram chipset that is actually recommended for this MB/CPU combination I will be perfectly happy to increase the voltage if that is what is on the side of the chip.

Why should I try to test the memory for stability when it's not the target memory?

I can run for days and weeks with the current memory while it is set at the Automatic OC speed of 3.7GHz. Since I have demonstrated exhaustively the instability of any CPU speed above that why try pushing it any further until I have a memory upgrade in hand?

Sure I could run it at its baseline speed with all the ordinary turbo boost engaged. The result is a stable system running at about 3.2-3.3GHz. Since I can do the same at 3.7GHz why should I bother with 3.3GHz?

I will download and look at the "memtest86" but since I don't have the target memory in hand yet, I don't want to stop production at a known working level of speed/stability just to test some non-target memory for stability.

Thank you for pointing me to yet another Linux utility I would want to know about.

Respectually,
Tom
A proud member of the OFA (Old Farts Association).
ID: 1966840 · Report as offensive
Profile betreger Project Donor
Avatar

Send message
Joined: 29 Jun 99
Posts: 11358
Credit: 29,581,041
RAC: 66
United States
Message 1966842 - Posted: 24 Nov 2018, 5:40:18 UTC - in response to Message 1966840.  

Sure I could run it at its baseline speed with all the ordinary turbo boost engaged. The result is a stable system running at about 3.2-3.3GHz. Since I can do the same at 3.7GHz why should I bother with 3.3GHz?

Tom given that is a CPU does it really matter other than intellectual curiosity? A middle to upper range GPU will do so much more work.
ID: 1966842 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1966844 - Posted: 24 Nov 2018, 6:03:34 UTC - in response to Message 1966842.  
Last modified: 24 Nov 2018, 6:13:23 UTC

Sure I could run it at its baseline speed with all the ordinary turbo boost engaged. The result is a stable system running at about 3.2-3.3GHz. Since I can do the same at 3.7GHz why should I bother with 3.3GHz?

Tom given that is a CPU does it really matter other than intellectual curiosity? A middle to upper range GPU will do so much more work.


Sorry. I may have not made my bias clear. I want the most cores running the fastest I can get them to run. If I simply wanted the highest possible RAC within my budget constraints I would be running machines with more GPUs than cpus. Two out of five of my machines are running heavy mixes of mostly biology processing which are purely CPU-based.

I have 3 machines with pretty high core counts. A single e5-2670v1 (8c/16t), a dual CPU e5-2690v2 (10c/20t) and a 2990WX (32c/64t). They are running respectively a GTX 750Ti (about 4-5 minutes per Seti gpu task), a triple load of gtx 1060 3GB's (2-3 minutes per Seti task) and 2 Gtx 1060 3GB's (ditto) with a 3rd on order for the 2990WX.

If I wanted absolutely higher RAC's I wouldn't be farting around with Gtx 1060 3GB's would I?

I am confident a pair of used Gtx 1070's or even 1070Ti's would send me up to the bottom edge of the first page or the top of the 2nd page of the LeaderBoard when combined with 32 CPU core counts. I am resisting that lure since only on the pure Seti machines would it make sense to do that.

For the bang for the buck (about $150 per so far) used Gtx 1060 3GB's are doing just great.

Thank you for bringing up the question.

Respectfully,
Tom
A proud member of the OFA (Old Farts Association).
ID: 1966844 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1966851 - Posted: 24 Nov 2018, 13:07:01 UTC - in response to Message 1966842.  

Sure I could run it at its baseline speed with all the ordinary turbo boost engaged. The result is a stable system running at about 3.2-3.3GHz. Since I can do the same at 3.7GHz why should I bother with 3.3GHz?

Tom given that is a CPU does it really matter other than intellectual curiosity? A middle to upper range GPU will do so much more work.


After a while, it occurred to me that I had a 2nd explanation that might also be useful. This is my "core envy" thread.

https://setiathome.berkeley.edu/forum_thread.php?id=83576

So like any average Setizen I have multiple reasons for doing things "my way" :)

Tom
A proud member of the OFA (Old Farts Association).
ID: 1966851 · Report as offensive
Ian&Steve C.
Avatar

Send message
Joined: 28 Sep 99
Posts: 4267
Credit: 1,282,604,591
RAC: 6,640
United States
Message 1966856 - Posted: 24 Nov 2018, 13:37:24 UTC - in response to Message 1966844.  
Last modified: 24 Nov 2018, 13:38:41 UTC

Memory stability and cpu stability are linked.

You can definitely have memory work with the CPU clocks at one speed, but then cause instability at higher clocks. The memory controller is on the CPU.

I’m not saying it’s definitely your problem. I’m just saying it’s possible and shouldn’t be dismissed.
Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

ID: 1966856 · Report as offensive
Profile Tom M
Volunteer tester

Send message
Joined: 28 Nov 02
Posts: 5124
Credit: 276,046,078
RAC: 462
Message 1966874 - Posted: 24 Nov 2018, 17:55:44 UTC - in response to Message 1966856.  

Memory stability and cpu stability are linked.

You can definitely have memory work with the CPU clocks at one speed, but then cause instability at higher clocks. The memory controller is on the CPU.

I’m not saying it’s definitely your problem. I’m just saying it’s possible and shouldn’t be dismissed.


I promise not to dismiss the issue. I do want to hold up on trying to test out the situation until I have the best recommended ram. I looked at some money sitting in my Paypal account plus the memory I just sold and ordered a new, exactly what b-die recommended memory kit that appears to be exactly what will run the best for a 2990WX on any MB. As soon as it arrives and demonstrates it will run on my current setup I will be selling the previous ram in an attempt to recoup the costs of that ram set.

So that will be my next hardware upgrade. I also have been going back over the MB reviews with the 2990WX and have been collecting comments. Ultimately I may be pushed out of the MB I chose originally (it came highly recommended by a game website and it was the cheapest choice) by the "weak power system" that has been mentioned. Only further testing will tell.

Tom
A proud member of the OFA (Old Farts Association).
ID: 1966874 · Report as offensive
Profile StFreddy
Avatar

Send message
Joined: 4 Feb 01
Posts: 35
Credit: 14,080,356
RAC: 26
Hungary
Message 1966878 - Posted: 24 Nov 2018, 18:30:42 UTC - in response to Message 1966874.  

one more comment on the RAM:
Based on your MB's QVL here: https://www.msi.com/Motherboard/support/X399-GAMING-PRO-CARBON-AC#support-mem-17
you can see anything above 3000MHz needs 1.35V.
ID: 1966878 · Report as offensive
1 · 2 · 3 · 4 . . . 7 · Next

Message boards : Number crunching : My 2990WX


 
©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.