The RX Vega 64 Owners Thread

[jimpsar]

Use a more demanding bench, like timespy Extreme, Firestrike Ultra, or even Extreme and use that as a comparison for performance.

Standard results can vary a bit as its not as demanding and you have other limitations which can affect score. You can already see that some of the highest Vega G scores come from using lower clock speeds in Firestrike Standard. You won't generally see that in the benches listed above.

Thank you Matt,
Will try and report

 

[balky12]

After a weekend of tweaking I have just got:

https://www.3dmark.com/3dm/23882386? GPU Score 24,657.

Although Destiny 2 tends to crash any OC I do pretty quickly

Not sure why drivers aren't approved.... 17.11.4

 

[silversurfer]

I'd love the idea of 16gb HBM2 being some great advantage in at least one game, maybe open world but I've yet to read of anything close to a premier advantageous parallel to the price. Still would be nice to pick one up cheap second hand from some design firm maybe. I once bought a graphics card from a Van Goch dealer in Colarado on ebay, big bargain you never know

If you think 3900 RPM is okay then I'd really consider going to the doctors.

I've got mine currently at 75c core maximum at 2800 RPM.

That's at 1025mv and I'm pretty sure I can get that down more.

You are probably right because my context is the 2900gt era cards. After my 7950 conked out I refused to get anything better till vega was out, probably should have got the 480 but anyhow Im used to fairly loud fan noise and vega I dont rank that bad. 4100rpm now in The division. I dont need it that fast as the temp is 61c 1652/1000 but I honestly cant hear it behind all the background ambient in this game. I never use speakers or I would need the AIO tbh or if I ran at 4k I think that would drive it harder. I oversample GTA5 sometimes and that does push it to the maximum and its audible on loading screens

 

[LoadsaMoney]

HWiNFO32/64 v5.70 Released

Changes:

• Enhanced sensor monitoring on EVGA X299 FTW K.
  
• Added reporting of more AVX-512 features supported.
  
• Added reporting Secure Memory Encryption (AMD) / Total Memory Encryption (Intel) supported.
  
• Added monitoring of Intel GPU (GT) and Media Engine Usage for Skylake and later iGPUs.
  
• Fixed reporting of some EC voltages on ASUS X99-DELUXE II, STRIX X99, X99-A II and RAMPAGE V EDITION 10 series.
  
• Enhanced sensor monitoring on ASRock IMB-19x series.
  
• Enhanced sensor monitoring on ASUS WS X299 PRO.
  
• Enhanced SPD information for NVM (AEP) DIMMs.
  
• Enhanced preliminary support of Intel Ice Lake.
  
• Fixed reporting of temperatures on AMD Vega since Crimson 17.11.2 driver.
  
• Added reporting of CPU per-core turbo ratios (IA/SSE, AVX2, AVX-512) for Intel.
  
• Added several Coffee Lake CPU models.
  
• Improved access to Intel Management Engine.
  
• Improved BCLK measurement on Skylake-X/SP CPUs.
  
• Fixed a rare crash on some AMD Multi-GPU systems.
  
• Fixed reporting of VDDC on some Bonaire/Hawaii GPUs using later AMD drivers.
  
• Fixed sensor monitoring on some MSI X99A series.
  
• Added recognition of AMD Pinnacle Ridge.
  
• Added reporting of CPU TDP and Tj,max for AMD Zeppelin family.
  
• Added an universal method for determining Tctl_offset for AMD Zeppelin family.
  
• Improved reading the real CPU model to avoid fakes for AMD Zen family.
  
• Added monitoring of CPU Cache/Mesh voltage for ASUS RAMPAGE VI APEX/EXTREME (requires new BIOS).
• Improved support of AMD Raven Ridge.
• GPU I2C support is now automatically disabled on AMD Vega series to prevent a possible system crash.
• Automatic disabling of GPU I2C support can be circumvented using a new GPU I2C Support Force option.
• Added several new GPU card models.
• Added preliminary support of Whiskey Lake.
• Added NVIDIA TITAN V.

https://www.hwinfo.com/download.php

 

[LtMatt]

I had to ditch HWINFO64 due to that bug with Vega Loadsa. Gone back to Msi Afterburner and Rivatuner for monitoring and overlay and no more issues since.

 

[henson0115]

I had to ditch HWINFO64 due to that bug with Vega Loadsa. Gone back to Msi Afterburner and Rivatuner for monitoring and overlay and no more issues since.

According to the dev on oc.net he is disabling the vr monitoring by default to stop the crashing and you can opt in if you want.

 

[LtMatt]

According to the dev on oc.net he is disabling the vr monitoring by default to stop the crashing and you can opt in if you want.

Are we talking about the bug which shuts the GPU down, puts the fan to 100% andyouget no GPU signal until you power down the PSU?

Interesting, which version? Tried the latest, this morning and has the same issue. Not sure if it was the version Loadsa posted or the one previous, will have to check my recycle bin when I’m at the computer.

 

[henson0115]

Are we talking about the bug which shuts the GPU down, puts the fan to 100% andyouget no GPU signal until you power down the PSU?

Interesting, which version? Tried the latest, this morning and has the same issue. Not sure if it was the version Loadsa posted or the one previous, will have to check my recycle bin when I’m at the computer.

http://www.overclock.net/t/1634018/official-vega-frontier-rx-vega-owners-thread/5100#post_26492469
Hope it helps. From memory it causes a crash that requires an hard reboot. Looks like it's the same version Loadsa posted.

 

[Ree]

Are we talking about the bug which shuts the GPU down, puts the fan to 100% and you get no GPU signal until you power down the PSU?

this has happened to me a few times this week on boot up as soon as desktop screen has loaded.. i was blaming the new 17.11.4 version as it was only thing i could think of that had changed in my system when it started happening, damn near crapped myself the first time the fan ramped up as my head was right next to pc reaching for pen i dropped lol.

 

[TonyTurbo78]

I had to ditch HWINFO64 due to that bug with Vega Loadsa. Gone back to Msi Afterburner and Rivatuner for monitoring and overlay and no more issues since.

Ah so that's what's causing my system to hard crash, black screen and turn my gpu completely off. So annoying when it happens, not even smashing the reset button brings it back to life. Only thing that works is to totally power off from the case.

Will not use hwinfo to monitor from now on then.

 

[LtMatt]

Yes it was the version Loadsa posted, so would recommend avoiding for now if you use Vega GPUs, it's a really nasty bug. I had HWINFO64 running on startup and it was shutting down the system within seconds of Windows boot, was a pain to get around.

EDIT - Looks like i used the latest Beta which still has the bug, use 5.70 official release and no system lockup.

 

[gupsterg]

Are we talking about the bug which shuts the GPU down, puts the fan to 100% andyouget no GPU signal until you power down the PSU?

Interesting, which version? Tried the latest, this morning and has the same issue. Not sure if it was the version Loadsa posted or the one previous, will have to check my recycle bin when I’m at the computer.

This issue is AMD's fault.

AMD told Martin while back VRM monitoring would work on VEGA like past cards, it has not! v17.11.4 finally started giving VRM temps, etc but is still borked! Martin keeps reporting it but AMD fail to fix properly!

AMD really are failing recently with driver updates IMO.

 

[LtMatt]

This issue is AMD's fault.

AMD told Martin while back VRM monitoring would work on VEGA like past cards, it has not! v17.11.4 finally started giving VRM temps, etc but is still borked! Martin keeps reporting it but AMD fail to fix properly!

AMD really are failing recently with driver updates IMO.

Thanks for the feedback.

I've had VRM temps in HWINFO64 with Vega since 17.8.1, but had to disable it as it caused stuttering and low fps. It was only in the last month or two that it started causing this shut off issue.

I saw his post on oc.net, not sure who his contact is at AMD otherwise I'd follow up with them.

I've tried using prefer AMDs ADL, disabling VRM monitoring in HWINFO64, but still had the same issue. I've had to uninstall HWINFO64 for now which is a shame as i like using it, but no problems since using MSI AB/Rivatuner.

 

[shankly1985]

The last two DLC maps packs for Battlefield 1 are bloody excellent. Its really got me back into BF1.

 

[gupsterg]

Thanks for the feedback.

I've had VRM temps in HWINFO64 with Vega since 17.8.1, but had to disable it as it caused stuttering and low fps. It was only in the last month or two that it started causing this shut off issue.

I saw his post on oc.net, not sure who his contact is at AMD otherwise I'd follow up with them.

I've tried using prefer AMDs ADL, disabling VRM monitoring in HWINFO64, but still had the same issue. I've had to uninstall HWINFO64 for now which is a shame as i like using it, but no problems since using MSI AB/Rivatuner.

Dunno about the contact within AMD that Martin uses, PM him on OCN or on his own forum .

I have not had VRM temps since launch. They only got exposed for me with v17.11.4 driver. I have used HWiNFO with VRM temp info monitored for ~24hrs+ continuous usage. Mostly f@h / gaming and light office usage at the time.

I have just loaded v5.70-3300, this has I2C comms disabled, so I do not get VRM temps. So this should not create a case of GPU "removing" from system, testing now.

I use HWiNFO virtually 24/7, so I can report to Martin if:-

a) the Super IO chip is going freaky on ASUS C6H/ZE boards.
b) VEGA has an issue.

The current belief is VEGA either has driver issue for I2C access or needs a SMU FW update to function correctly for this. VEGA FE has I2C comms disabled via registers on VRM control chip. RX VEGA does not, it was supposed to work on RX VEGA, but has not. So you can't take I2C dumps from even MSI AB via command line. It maybe that VEGA is like Fiji, the SMC (on die controller) is messaging VRM control chip so much that it is basically not releasing bus for other apps to use. So Fiji had I2C disabled in HWiNFO but AMD must have allowed some of the monitoring values to cross over to AMD ADL API as that is what I believe Martin used to give VRM temp, etc info on Fiji. Perhaps this needs to be done for VEGA in driver also.

 

[LtMatt]

Dunno about the contact within AMD that Martin uses, PM him on OCN or on his own forum .

I have not had VRM temps since launch. They only got exposed for me with v17.11.4 driver. I have used HWiNFO with VRM temp info monitored for ~24hrs+ continuous usage. Mostly f@h / gaming and light office usage at the time.

I have just loaded v5.70-3300, this has I2C comms disabled, so I do not get VRM temps. So this should not create a case of GPU "removing" from system, testing now.

I use HWiNFO virtually 24/7, so I can report to Martin if:-

a) the Super IO chip is going freaky on ASUS C6H/ZE boards.
b) VEGA has an issue.

The current belief is VEGA either has driver issue for I2C access or needs a SMU FW update to function correctly for this. VEGA FE has I2C comms disabled via registers on VRM control chip. RX VEGA does not, it was supposed to work on RX VEGA, but has not. So you can't take I2C dumps from even MSI AB via command line. It maybe that VEGA is like Fiji, the SMC (on die controller) is messaging VRM control chip so much that it is basically not releasing bus for other apps to use. So Fiji had I2C disabled in HWiNFO but AMD must have allowed some of the monitoring values to cross over to AMD ADL API as that is what I believe Martin used to give VRM temp, etc info on Fiji. Perhaps this needs to be done for VEGA in driver also.

My bad, looking back on my posts in August it was VRM Voltage monitoring, not temps, on HWINFO64 that was causing the FPS drops and stuttering. Had to disable VRM monitoring to fix in HWINFO64.

Also just realised i tried the latest Beta and not the latest official release of HWINFO64 5.70. I was able to load 5.70 without the instant lock up which is good.

 

[gupsterg]

NP .

Yeah the 0.900/1.25V UPi voltages cause stutter/issues, TBH for me I didn't get any.

With v17.11.4 came more VRM Power output values and VRM temp info, again some had issues, I didn't except 2-3 times GPU got removed from system .

Here is v5.61-3290 on 12.5hrs f@h run with VRM temps but other VRM info hidden. Only done ~2hrs on latest, link. We can still get the extra VRM info if we set "GPU I2C Support Force" in settings when launch app.

 

[StonedPenguin]

I cant find any of these in stock anywhere

 

[gupsterg]

That's the magic of VEGA! mwahahaha!

 

[Steampunk]

I cant find any of these in stock anywhere

AMD has stopped making the reference versions, AIB partners haven't got theirs out for sale except in minuscule quantities here and there, if at all. No Christmas Vega for you! Or me!