Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - skiny75

Pages: [1] 2
1
It makes at best 1-2C difference at normal RPM (<1500). U14S doesn't exactly need more static pressure because it's thin&spare heatsink (compared to wide & dense heatsinks like U12A or NH-D15).
Additional fans only increase pressure, they don't increase volume of air. You probably already have exhaust fan at back of the case that is few cm distance from U14S, there is no need to have 3 fans in row.

Yes some people mount even high-pressure industrial fans to NH-U14s but it's really bad idea... those kind of fans belong to super-dense heatsinks like Ice Giant ProSiphon or water-cooling radiators.

I run my 3990X with U14S and single-fan at only 800 RPM to make it silent. But 3960X/3970X do benefit from faster speeds.

Threadrippers, or Ryzen CPUs, also don't suffer from lack of heatsink dissipation, but from thermal transfer between the chip and the cooler. Even million fans will not help with temperature much because the chip is hot under the heatsink, the heatsink is ok (just try to touch it, it's not super hot at all).

The only way to lower temperature is to lower the room ambient temperature. It's good idea to work in <22C in general anyway. Or reduce voltage through power thermal limit.

Overheating Threadripper simply downclocks to lower frequency, it doesn't cause any crashes. Over-heating memory... can cause crashes. Those are hard to cool but they actually easily run 70+C on Threadripper boards with XMP profiles in 1.35+ Volts. Good room ambient and strong case airflow are important then.

Thank you for the explanation. 
They discover that the problem was in the CPU , the 3990x, now they will change it.  I was unlucky, after 1 month that i bought, I started to have problems. 

2
What? D15 doesn't even fit Threadripper, the only available cooler is U14s TR4. Since this wasn't really Threadripper related I will clean the thread in upcoming weeks.
It's becoming free technical support.

Sorry Juraj, 
I have the U14s TR4 (1 fan) , but I have seen that many user of 3990x have also the Noctua NF-A15 PWM (singular fan) optional to add. 
 

3
Thanks Guys, 
1 week ago I sent back the workstation and now they are testing every component or at least I hope they are doing it.
In my configuration I had only 1 fan Noctua 14 without the d15.  Anyway the CPU was max around 70c
Do you think that can also a be a problem?   

4
Thanks Juraj,

Yesterday morning with the IT guy, we : 
-Check the Memory and it swap 2 of the pair
- GPU place in different PCi lane x16, Stress software test. He said that the RTX 3080 has some problem, but this looks correct.
- Connection from Power supply to GPU 2 times 8 pin directly
- XMP disabled. Default settings
- CPU temperature 61c during rendering time.

After that, It was working-rendering perfectly.... in the evening it crashed 2 times. 1 in interactive and 1 in rendering normal.  Black screen and restart like last time.
Another problem came out. When I switch off Windows 10, the pc still on...

 I am a bit frustrated ;-(  THanks for any suggestions.


5
Short Update.  The workstation does not switch ON anymore.
I contact the IT to come and fix it  ;-(   

6
I made the check memory after the crash.
The XMP profile is disabled.
I check in the Bios.  I attached the view of the Bios.
How can I   "at default clocks for memory (2133/2400/2667 depending on module" , Sorry but I did not understand.  Do I need to change the frequency of the memory?
If yes, How to do it ? 

This is the result of the Nirsoft program to check the reason of the crash.

Dump File         : 120120-15765-01.dmp
Crash Time        : 01/12/2020 12:56:52
Bug Check String  : MEMORY_MANAGEMENT
Bug Check Code    : 0x0000001a
Parameter 1       : 00000000`00041792
Parameter 2       : fffff23f`fc604400
Parameter 3       : 00000000`00240000
Parameter 4       : 00000000`00000000
Caused By Driver  : ntoskrnl.exe
Caused By Address : ntoskrnl.exe+3f5210
File Description  :
Product Name      :
Company           :
File Version      :
Processor         : x64
Crash Address     : ntoskrnl.exe+3f5210
Stack Address 1   :
Stack Address 2   :
Stack Address 3   :
Computer Name     :
Full Path         : C:\Windows\Minidump\120120-15765-01.dmp
Processors Count  : 128
Major Version     : 15
Minor Version     : 19041
Dump File Size    : 6.510.940
Dump File Time    : 01/12/2020 13:00:16
==================================================

Thanks
 
 

7
Hi All,
unfortunately, the problem was not solved. After 1 day and half same problem with the workstation.
I make a summery.
I use max 2021 and Corona 6.1. After computing the GI, Start rendering and after 1 second black screen and restart.
Many times, after the restart, I need to validate again my Corona licence (I do with login online).

I have a 3990x - Noctua NH-U14S - Gigabyte TRX40 Designare.  128 Corsair Vengeance LPX  (4x32) - Seasonic Prime Ultra Platinum 1300.
I have the RTX 3080.  I installed the last version of the Nvidia Studio driver.

Temperature of the CPU around 60.  I did memory check, was ok.

Any suggestions?  Thanks a lot in advance.


8
Hi Juraj, 
Sorry, for the delay. 
Temperature of CPU was 65-70 and VRm of the motherboard 40 ( if i see correct ).
The IT guy where I bought, said that Windows 10 pro was corrupted. ( I think also Max 2021 had some issue). No issue in the hardware.
He reinstalled from zero, checked all driver compatibility(GPU) Bios Gigabyte updated.
Now it's working again (it's great)  and I cross the fingers.
Thanks a lot for the help and availability

9
HWMonitor for detailed information on temperatures (CPU, individual cores, VRM, chipset and GPU, even SSDs) or even basic RyzenMaster for quick CPU temp.

The frequencies are ok, that's how turbo boost works. It's affected mainly by load (secondary by threads amount utilized).
It can be GPU driver as well if you are using Optix denoiser. The latest Ampere drivers are pretty fresh out of the box, with plenty of issues reported. Try Studio driver, it's more conservative.

Under normal conditions, 3990X should go up to 70C with U14S in well ventilated case and average room ambient (25).
It wouldn't crash if it was hot though. It would just downclock further, but like..a lot further.

Lastly, most common culprit, esp. if it crashes the whole Windows, not just software, could be memory due to unstable profile. Not sure what the particular XMP/DOCP settings are for your Vengeance kit.

Thanks a lot Juraj. I will try to fix it tomorrow.. I think it's a tecnical hardware problem, but I have a small doubt.
This morning I installed the Security Tools for Autodesk( i receive the email that they suggest to install it for the MAx 2021). In the afternoon, I start to have a problem Do you think that it can also be a reason?  Thanks

Hi Juraj.  I tried many options. I have updated GPU Studio driver, Windows 10 pro ( last update). Reset default Bios config.   The system still unstable.  I check in the Bios and I have the XMP Disabled.  System Memory speed  Auto . Infinity Fabric Frequency and Dividers Auto.   Shall I Activate the XMP ?  Thank you in advance.

10
HWMonitor for detailed information on temperatures (CPU, individual cores, VRM, chipset and GPU, even SSDs) or even basic RyzenMaster for quick CPU temp.

The frequencies are ok, that's how turbo boost works. It's affected mainly by load (secondary by threads amount utilized).
It can be GPU driver as well if you are using Optix denoiser. The latest Ampere drivers are pretty fresh out of the box, with plenty of issues reported. Try Studio driver, it's more conservative.

Under normal conditions, 3990X should go up to 70C with U14S in well ventilated case and average room ambient (25).
It wouldn't crash if it was hot though. It would just downclock further, but like..a lot further.

Lastly, most common culprit, esp. if it crashes the whole Windows, not just software, could be memory due to unstable profile. Not sure what the particular XMP/DOCP settings are for your Vengeance kit.

Thanks a lot Juraj. I will try to fix it tomorrow.. I think it's a tecnical hardware problem, but I have a small doubt.
This morning I installed the Security Tools for Autodesk( i receive the email that they suggest to install it for the MAx 2021). In the afternoon, I start to have a problem Do you think that it can also be a reason?  Thanks

11
Hi Guys,

I have bought the new workstation with the 3990X, I am using since 3 weeks. until now was ok. but from today, when I render many times he crash and restart.
I have the Noctua NH-U14S - Gigabyte TRX40 Designare.  128 Corsair Vengeance LPX  (4x32) RTX 2080 Super. 

Do you think that It can be because it's too hot?  How Can I check it? I see that in normal situation the speed is at 3.9Ghz. When I render, It goes to 2.9 at 64cores. 
I did not make any overclocking.

I would be really happy if somebody can help me to understand where it can be the problem. 
Thanks a lot in advance

12





Hey, so:

- Noctua is the right choice, always. Why it absolutely beats BeQuiet and other TR4 coolers is the heatpipe orientation. Noctua runs the heatpipes across the length of the chip, while everyone else across the width.
  Because of this, only Noctua actually covers all the dies on 3990X with heatpipes. All TR4 dedicated coolers cover the heatspreader, but quick access to heatpipes is what really makes the difference. KitGuru reviewed this and difference was bigger than 10C on 3990X.
On 3960/3970X the heatpipes are ok, but since those run hotter it's still bad idea. Noctua is always the correct answer. Always.

Regarding memory, I recently upgraded to 32GB 3600 CL18 DIMMs from Patriot, they use the same Hynix CJR dies that G.Skill & Corsair, but were so much cheaper :- ). I paid 280 Euro for each 64(2x32) Kit. And it works on my Aorus Xtreme TRX40 super well, super stable at XMP.
If you want higher-end kit, then Micron Rev-E Crucial Ballistics offer 32GB DIMMs with 3600/CL16. These are even better than Hynix CJR. (There is no Samsung B-Die with 32GB Dimm capacity for consumers).

Cannot say anything about HyperX Fury, no idea what it is, but since it's binned at 3200/CL16 I would say it's not a greatest quality. The speed itself isn't very important for 3rd gen Threadrippers, but higher-end memory kits have higher chance to run stable in 128-256GB configurations with higher speed. For those who want 256GB 3600 CL16-18, you need the best kits.

Good choice of SSDs, can't go wrong with those, PCI 3.0, but good controller, dram capacity and TLC with SLC Cache.

GPU, sure, any choice you like.

Hey Juraj,  Thanks a lot for the help. 
Next days I will order it. Really curious to try IR.


13
Hi Juraj and all the other guys,

I have read all the posts from 50-78 page. Thanks a lot for all the information, I did not know so much.
I arrived at conclusion to configure this AMD workstation, but before to order I would like to check if everything is fine.

AMD Ryzen Threadripper 3990X
Gigabyte TRX40 DESIGNARE
HyperX FURY - DDR4
128 GB: 4 x 32 GB - DIMM 288-PIN - 3200 MHz / PC4-25600 - CL16 - 1.35 V - unbuffered - non-ECC
GIGABYTE GeForce RTX 2080 SUPER WINDFORCE OC or maybe a RTX2080Ti( I know that you suggest to buy a 2nd hand 1080ti, but for me it's more convenient to buy new(I have warranty and discount on taxes)
Samsung 970 EVO Plus 1TB - Solid state drive
Samsung 860 EVO 2TB - Solid state drive
Seasonic Prime Ultra Platinum 1300 - Power supply
Phanteks Enthoo Pro SE Tempered Glass - Full tower
For the CPU Cooler I propose the Noctua NH-U14S TR4-SP3 as you suggest, the seller pushed me to Be quiet! Dark Rock PRO TR4.  I think i will keep the Noctua. I see that almost everybody here has it.

I would really appreciate if Juraj  (I can immagine you are  busy or on holidays) or somebody else can confirm or suggest this configuration.
Thanks a lot in advance 



14
Thanks a lot TomG.

your answer helps. I will try in this way.


15
Hi Guys,

Sorry, but somebody has the same problem ?  How did you solve it ? 

I have no idea what to do now..

Thanks in advance,

Paolo

Pages: [1] 2