XFX Speedster Zero RX 6900XT issue?

Pandur

n00b
Joined
Jun 4, 2019
Messages
36
Points
18
I have one of these gpus and I could really use a sanity check from you experts. I have had some driver crashes, a couple times I have also had rainbow screens on boot. All sorts of "goodies" basically. Games CTD more often than expected.

I have been over everything I can think of, from my 30+ years as an overclocking nerd. I am considering doing a RMA by now. But I haven't had an AMD gpu since basically forever.

Let me start with some system specs:
Ryzen 9 5950X
Asus Crosshair VIII Dark Hero
2x 16GB G.Skill Trident Z Royal (F4-4000C16D-32GTRS)
Corsair MP600 Pro Hydro X
XFX Speedster Zero RX 6900XT
Samsung G7 27" monitor x2 (2560x1440 @ 240Hz)
Be Quiet! Dark Power Pro 12 1500W psu. Bit overkill, but I got tired of Corsair psus dying on me. So here we are.
4x Noctua NF-A14 industrialPPC 3000rpm fans
Corsair Obsidian 750D airflow edition

Custom watercooling loop:
EK D5 pump + res combo
EK 280x45mm radiators x2 (with the Noctua fans obviously)
CPU/VRM fullcover block
MP600 Hydro X
GPU fullcover block

This entire setup, minus the watercooled m.2 ssd, psu and gpu have been running for about a year. Just with a MSi GTX1080ti Seahawk X EK instead for gpu.

I have done the obvious: Reinstall Windows, fresh drivers, stability testing of memory and cpu. Memory runs at 3800 with a 1900 fabric clock and the cpu with PBO will hit 4.5GHz all core. I am confident that these settings are stable. I have also replaced the DP cables since I have had those shiny rainbow events.

CPU will hit about 62C single thread and 80C multi thread while at 100% load. But I never see anywhere near 100% load in normal use, so I am not worried about the worst case temps honestly. Most of the time it hovers around 40-45C while gaming and normal usage. So in hindsight I should've gone with a 5800X or 5900X. But where's the fun in that. For comparison the 1080ti would do about 42C (65C hotspot) at full load while hitting the PWR limit in this loop.

Now for the main topic: What should I expect from a properly pasted RX 6900XT watercooled card? I see about 65-68C GPU temp but with a hot spot delta temp of over 30C. Highest I have seen while testing is 98C which is more than the 100% thermal limit of 95C. So this will hit the GPU Hot Spot thermal limit (as reported by hwinfo in the attached image) within minutes of doing anything other than youtube. The screenshot was taken after just one run of Port Royale, and you can see the water temp is . Midway through, the clock was down to 2.2GHz. I am unsure if it is relevant but you can also see that GPU current/power usage is never anywhere near the max settings either. I would expect more in Port Royale honestly. Which in my head is another indication of throttling.

Water temp is shown in the image and that maxed out at 35C while I was testing today. Over time it will go a few degrees higher, but I got an aggressive fan profile based on the water temp so they ramp up quickly.

I know @Zarathustra got this card and did a repaste earlier. My complication is that where I live, the warranty void if removed sticker is for real. So a repaste is out of the question. :( I am honestly expecting a poor paste job. But an actual defect gpu is probably also on the menu.

Have I missed something obvious? Are the GPU temps normal for AMD? My google-fu indicates that I should expect better but I honestly don't know.

Any comments/observations would be greatly appreciated.

Thanks, and sorry for the long post. Congrats if you made it here!
 

Attachments

  • hwinfo_xfxzero.png
    hwinfo_xfxzero.png
    155.9 KB · Views: 9
Hmm.

I can't explain your issues.

I repasted mine, but honestly, when I did, the original paste wasn't that bad, it was among the better OEM paste jobs I've seen. I did get better temps out of it, (though not as much of an improvement as I originally though) but the original paste job was by no means bad enough that it should have caused problems.

Your GPU temps seem much higher than mine, even though your loop temps are only slightly higher than mine, which makes little sense, unless you are having a flow problem or there is - as you suggest - issues with the block contact or paste (but mine looked pretty OK when I took it off.

Do you know what kind of flow you are getting?
 
So - try the easy things first, right?

Make sure that the card isn't sagging. Ran into that one with an XFX 5700 we'd intended to use for case reviews, and dealt with quite a bit of display corruption and driver crashing, or drivers just not loading in the first place.

I highly suspect that this isn't likely given that your watercooled card should be quite a bit lighter and is newer, but worth checking regardless IMO.
 
Hmm.

I can't explain your issues.

I repasted mine, but honestly, when I did, the original paste wasn't that bad, it was among the better OEM paste jobs I've seen. I did get better temps out of it, (though not as much of an improvement as I originally though) but the original paste job was by no means bad enough that it should have caused problems.

Your GPU temps seem much higher than mine, even though your loop temps are only slightly higher than mine, which makes little sense, unless you are having a flow problem or there is - as you suggest - issues with the block contact or paste (but mine looked pretty OK when I took it off.

Do you know what kind of flow you are getting?

Sadly I don't have a flow meter. I can say that when I flushed it last time, to install the new GPU, it does flow very well out the drain port while running water through with the pump. And it did clear out the air bubbles quickly. But that is not very scientific data. I have been meaning to get a flow meter for some time now. Guess this is a good excuse.

I got a new generation EK D5 pump that I run at a flat 90% PWM. It makes less noise than the fans at that rpm. I have straight or 45 degree angle fittings. I use 3/8" ID tubing. The m.2 waterblock is straight-through. Fullcover cpu/vrm combo is a bit restrictive of course. But same with the gpu fullcover. And I had no such problems with the 1080ti which also got an EK fullcover block.

How was your GPU temps in something like Port Royale?

So - try the easy things first, right?

Make sure that the card isn't sagging. Ran into that one with an XFX 5700 we'd intended to use for case reviews, and dealt with quite a bit of display corruption and driver crashing, or drivers just not loading in the first place.

I highly suspect that this isn't likely given that your watercooled card should be quite a bit lighter and is newer, but worth checking regardless IMO.

This card weighs more than most air cooled ones that I have had my hands on. The backplate alone is a solid slab of metal. The weight distribution is a bit better imho, and the tubing does give some support. Short version, no there's not alot of sag. :)

I could do another re-seat of the card just in case.
 
I didn’t see - no overclocks on the GPU or Bclk/pci bus?

Apart from that I share your suspicion about bad paste job. The safe bet would be an RMA, but the quick method would be DIY
 
I didn’t see - no overclocks on the GPU or Bclk/pci bus?
Correct. I have not touched either.

Stock bios for the gpu. Although I believe that the switch is set to the higher performance of the two out of the box.

Latest motherboard bios (4004).
 
Oh and GPU is installed in the top PCIe x16 slot, the primary one. Link speed is x16 4.0 and Resizable BAR is enabled.
 

Attachments

  • gpuz_xfx.png
    gpuz_xfx.png
    18.5 KB · Views: 3
Those hot spot temps to me seem way to high for a water cooled 6900xt. Are you undervolting it?
 
VBIOS updated?

It doesn't seem like there are any updates available.

Those hot spot temps to me seem way to high for a water cooled 6900xt. Are you undervolting it?

I have not touched anything for this card. This is out of the box performance, or lack of. The card will hit the hot spot thermal limit and start throttling at less than 50% TGP. Which seems insane to me. And the large delta to the reported GPU temp is why I am thinking poor paste job or poor water block mount. And, as I said before, I can't break the warranty void seal without losing the warranty where I live.

I have had no issues with my loop previously. And the CPU is performing as before. Which means that unless I really misunderstood how AMD GPUs work, the GPU is the issue. (This is the real downside of having built computers with only Geforce GPUs for the last decade)

I really want to make sure I check all the boxes before I start the RMA process.

Thanks again for all the good posts guys, I really appreciate your help!
 
Yeah, RMA it. Hot spot temp is ridiculous for a water cooled card. The water coole dcards i see dont get over 60-70c hot spot. You have a hot spot temp like an air cooled card.
 
Become a Patron!
Back
Top