HD5970 locking up in 3D

[jcool]

Hey guys,

got a problem here. Diamond brand HD5970 locks up randomly when I fire up 3DMark (tried 3D06 and Vantage). Sometimes it passes the first GPU test and freezes during the 2nd, but mostly it just hangs at the first. No screen distortion or anything, the rig just locks up.

I set the CPU/ram back to stock clocks, but that made no difference (overclock was 20h prime stable anyways).

When I fire up FUR, it works - BUT when I check with MSI afterburner, I see that it auto-downclocks to 550 core
When I run the renamed FUR.exe it locks up within 2-3s. Even when I manually reduce the core speed to that same 550Mhz - anybody got any clue as to what is going on?

Even more strange is that the card seems to be running milkyway@home fine (distributed computing project that is VERY stressful on the GPU, more than games):



The card is watercooled, but I had the same lockups during 3Dmark with the stock cooler (I thought it may be heat related back then) but now, with the waterblock, it clearly isn't.

Tried both the official beta driver from AMD's homepage and the Cat 9,12.

Other hardware is Core i7 920, Asus RIIE, 6GB DDR3-1600, Corsair HX1000.

Is this some internal defect due to TSMC's crappy 40nm process? Any input appreciated...

[jcool]

Well.. now it froze the desktop while running Milkyway. I guess it really is a bad card. Yay, RMA time...

[cowie]

same with my diamond exactly i sent it back
it did not run in 3d for long at most 1minute not heat or power related either

[jaredpace]

Does it lock up if you change from 1.049v to 1.1v?

[jcool]

Changed vGPU to 1,1 and even 1,2V - to no avail. This is crap.

[cowie]

dude only thing i did not try was a bios flash you could try that but....
i knew i could not have the only card like that same brand and everything

[zalbard]

VDDCI pharse #1 temp = 0C? RMA...

[jcool]

dude only thing i did not try was a bios flash you could try that but....
i knew i could not have the only card like that same brand and everything

Yeah, I may try another bios. But it seems more than a coincidence...

VDDCI pharse #1 temp = 0C? RMA...

So that is not normal? Kinda figured it wasn't
One dead phase maybe?

[Dude_Person]

VDDCI is a memory phase, and it's not dead. Mine does that too, and I have also seen a screen cap from someone else with a 5970 that does that.

[zalbard]

One dead phase maybe?

Yeah, exactly what I'm thinking.
Or a temp probe. In any case, something is not right.

[Monstru]

Am I the only one that considers Diamond something made on the ship between China and Europe?
RMA the card bro, it is probably deffective from the start.

[Dude_Person]

Your VRM isn't dead dude. My card and at least one other I know of do the same thing (0c on VDDCI Phase #1) and they work fine. I don't know what your problem is, but that is not it.

[jcool]

Ok guys, got an update for ya'll.
I drained the loop, removed the card, and put the stock cooler back on. Just for the heck of it, I put it in a GB P55M-UD4 that happened to lie on the table crunching. Guess what.. NO freezes/lockups. I can even run the renamed FUR.exe.
Passed 3D06 and Vantage no problem, got 26500 06 and P21600 Vantage with a Core i5 750 @ 4Ghz and the 5970 @ stock. Sounds about right, no?

Next I put it in my comp (DFI x58) to see whether it may be chipset related. No trouble either here. WTF!

So, it's either a bad RIIE, or some compatibility issue, OR a driver up. Because both the GB P55 and my DFI rig run Vista 64, while I tested the RIIE on WIn 7 Pro 64.
Gonna put the card back into the RIIE rig and try again now.

[Monstru]

Fluck!!! This is indeen interesting. Maybe some TIM or dirt got into RE2's PCI-E slot? It happens sometimes.

[TJ TRICHEESE]

how long is that thing maby the power cannot get from one end of the card to the other
sorry

[cowie]

i used two classy boards to test mine using both xp32 and vista 64.... just a referance.
i wanted to test on other chipset but i did not have one avalible

[jcool]

Well, now it works. I cleared CMOS on the RIIE, loaded setup defaults - all fine. Re-set everything in bios, still fine. Card has been running milkyway for the past 5 hours or so...
Gonna put the waterblock back on tomorrow. Maybe it was an issue with the block, but I doubt it. Pads showed perfect contact on all VRMs, cores, rams and whatnot. Probably just some VERY annoying bios up on behalf of Asus

[cowie]

did other cards run ok on the reII?
i even tried lowgap settings remembering the issues that some x2 cards had on the x48,that still did not work.
i'd still like to know what was up with my card.

[jcool]

Yeah, I've had that RIIE running for months with a 5870, no problem. Totally weird...

[lowdog]

Why don't you update this and state that the problem all along is with the EK 5970 block!

I have the same issue, stock cooler works fine (no lockups) put the EK block on (4 x now) and card flucks up every time with lockups. EK block is phuked!.....something wrong with the design.

[wez]

Why don't you update this and state that the problem all along is with the EK 5970 block!

I have the same issue, stock cooler works fine (no lockups) put the EK block on (4 x now) and card flucks up every time with lockups. EK block is phuked!.....something wrong with the design.

This seem to have been a combo of the EK block, and issues with the rest of the system?


Seen a lot of issues blamed on the 5xxx cards. But that in the end was caused by unstable clocks, bios issues or otherwise buggy systems.

[lowdog]

This seem to have been a combo of the EK block, and issues with the rest of the system?


Seen a lot of issues blamed on the 5xxx cards. But that in the end was caused by unstable clocks, bios issues or otherwise buggy systems.

I think it's nothing more than the block to be honest, it's just not fitting some cards correctly, if the stock cooler is not causing issues yet the issues start when the block is installed then that pretty much speaks volumes in itself.

[jooney]

Today came my EK cooler and everything is fine with my Asus 5970,no problems in 2 or 3D.

Here Furmark with 8XMSAA.

[cowie]

nice good news about this card and water blocks for a change

[Dude_Person]

Nice to see that some of these work. Must be some minor component differences.

@ Jooney. You should run furmark fullscreen and even more importantly, disable MSAA. At 1ghz your VRM Current Draw should be able 80 amps. It's currently about 50amps. That will make a huge temp difference on the VRMs. I'm interested to see how this block performs under some real stress.