GTX480soc artifactaing out the box (screenshots/temps)

tommybhoy · 22 Mar 2011 at 18:12

Surely you can take it back and exchange it at OC with sale of goods act or something?
An extra £100+ for a quality 850w PSU on top of the price you paid.
Pay the extra and get a 570 or get them to swap it for a 6950>70.

bemaniac · 22 Mar 2011 at 18:21

This is worrying as I was just about to buy one in the morning

Rroff · 22 Mar 2011 at 18:28

Gigabyte seem to have had a hard time getting enough cores for the SoC parts lately (which is probably why the SE version even exists) seems quite a few that aren't making the grade are slipping through the selection process too

Buckster · 22 Mar 2011 at 18:31

juno_first said:
FIXED IT....

Took it back OCUK & they tested it with Corsair/Enemax 800/1000w PSU's & it ran fine out the box: 820 @1025mv.

Obviously my 750w Coolmaster is not good enough! Even with a claimed 80proof & 60amps for the 12v.
I then removed my Asus Zonar (which uses an external molex) & the 480soc is now nearly artifact free at 820, if I increase the voltage to 1138mv it appears stable at 820, but hits 105c with the fans at 100%.

Whats interesting is watching the 12v rail, which drops to 11.71 but when the 3 fans fire up to 100% this drops to 11.65.
Ironic how the graphics card fans could push the power requirements over the edge.

Hmm, so im going to have to get at least a Corsair 850w when I can afford now :/
Even though it only says a 650w is required.
Maybe everyone with issues over this card just need a better PSU ?

mmm this is very very odd - so you're saying (did you watch it ?) it tested fine on an 850W at OCUK ? ran OCCT without any errors ?

and same settings etc- on your 700W PSU error'd in a big way ?

I can't see how if a PSU is not "good enough" for the card - the card would funk out so to speak - as if it was trying to draw too much power from the PSU the PSU would likely shutdown (they are quite sophisticated) and or overheat ?

one of the other guys with similar artefacts had a 1000W PSU ....

Buckster · 22 Mar 2011 at 18:32

Rroff said:
Gigabyte seem to have had a hard time getting enough cores for the SoC parts lately (which is probably why the SE version even exists) seems quite a few that aren't making the grade are slipping through the selection process too

I don't get how it was fine on another PSU though ....

if he PSU was way off - I'd see the issue - but a 700W should be plenty

and I wouldn't expect a lack of power to create artefacts ?

Buckster · 22 Mar 2011 at 18:37

btw at my calcs - worst case my 480 is pulling about 19A in OCCT

which should be fine on any decent 650/700W PSU

to the OP - what mobo do you have - as had a thought today - there isn't an issue with these cards and PCI-E 1.1 is there? power via the PCI-E connector wise ..

bemaniac · 22 Mar 2011 at 21:09

Rroff said:
Gigabyte seem to have had a hard time getting enough cores for the SoC parts lately (which is probably why the SE version even exists) seems quite a few that aren't making the grade are slipping through the selection process too

So if I get the SE it might mean they were ones which totally had no chance of hitting 800 core......hmm Maybe I will go for the SOC but so many issues around. I know now after asking the hx1000 is perfect for even 2 of them but still hmm

juno_first · 22 Mar 2011 at 21:22

My mobo is an old P35 DQ6, which was a high end board 3 years ago, it is x16 PCIe.
Also tried it on another P35 DSr3, both Gigabyte.

Well, after lots of testing & removing the Zonar & increasing the voltage to 1088mv, everything runs fine now at stock clocks.
Just been playing Crysis1 for an hour & never saw any artifacts at max 1080p, also firmark runs fine now. Both these are now maxing out at 85c which is strange as earlier firmark maxed at 105c. Maybe it needed burning in?

Looks like it just needs 1088mv to run at stock with my setup, as default was 1025mv.
I dont know how OCUK ran it at 1025mv, unless a better PSU auto runs the card higher?

Im not enthralled it may not overclock much more if anything, but at least its running at stock now

bemaniac · 22 Mar 2011 at 21:28

Sounds almost like the thermal paste has spread out a bit or something that's weird!

Jono8 · 22 Mar 2011 at 21:53

Seems like sub par performance from an apparantly cherry picked/factory overclocked card. I would probably demand my money back. I thought that aftermarket cooler on it was supposed to make it run loads cooler ??

I dont think Crysis wont push your temps or power draw up as high as Furmark..

On my Asus one it was 0.988mv out of the box. 850 core is stable on 1050mv on mine but it runs a little hotter than i like it to. I run it at 1000mv at 800core 24/7 which means it never gets loud and temps stay around the 80 degree mark under load.

But if it runs all your games without a hiccup, with no artifacts and decent temperatures then i wouldnt worry too much about OCCT and Furmark.

juno_first · 22 Mar 2011 at 21:59

Yeah I bought the card because my other card went in a media rig & I need something with a bit of grunt for Crysis2 NEXT MONTH WHEN ITS ITS IN DX11. Fancy them not want to hurt console sales, even the PC DX9 version looks lots better lol.

I have not done a lot of testing yet.
Maybe there should be a thread of graphics cards with what clocks/voltages have been reached? To help others thinking of buying a particular card.

Buckster · 22 Mar 2011 at 22:01

juno_first said:
My mobo is an old P35 DQ6, which was a high end board 3 years ago, it is x16 PCIe.
Also tried it on another P35 DSr3, both Gigabyte.

Well, after lots of testing & removing the Zonar & increasing the voltage to 1088mv, everything runs fine now at stock clocks.
Just been playing Crysis1 for an hour & never saw any artifacts at max 1080p, also firmark runs fine now. Both these are now maxing out at 85c which is strange as earlier firmark maxed at 105c. Maybe it needed burning in?

Looks like it just needs 1088mv to run at stock with my setup, as default was 1025mv.
I dont know how OCUK ran it at 1025mv, unless a better PSU auto runs the card higher?

Im not enthralled it may not overclock much more if anything, but at least its running at stock now

be careful upping volts - according to Gigabyte forums - increasing volts over stock invalids the warranty !

Rroff · 22 Mar 2011 at 22:04

bemaniac said:
So if I get the SE it might mean they were ones which totally had no chance of hitting 800 core......hmm Maybe I will go for the SOC but so many issues around. I know now after asking the hx1000 is perfect for even 2 of them but still hmm

no... the cherry picked cores for the SoC are supposed to hit a decent margin above the factory clocks, so you could get a core that does stock SoC clocks stable even - but doesn't go much further.

Looks like they've been having trouble getting enough cores that are of that level tho and cutting corners... IMO the SE only exists because too many cores and PCBs didn't make the grade for SoC.

Freddie1980 · 22 Mar 2011 at 22:30

Buckster said:
mmm this is very very odd - so you're saying (did you watch it ?) it tested fine on an 850W at OCUK ? ran OCCT without any errors ?

and same settings etc- on your 700W PSU error'd in a big way ?

I can't see how if a PSU is not "good enough" for the card - the card would funk out so to speak - as if it was trying to draw too much power from the PSU the PSU would likely shutdown (they are quite sophisticated) and or overheat ?

one of the other guys with similar artefacts had a 1000W PSU ....

If the 12v rail isn't stable enough (like dropping out of the ATX spec which I think allows for a few % wobble) it could cause problems for the video card if the power delivery isn't stable.

Buckster · 22 Mar 2011 at 22:36

Freddie1980 said:
If the 12v rail isn't stable enough (like dropping out of the ATX spec which I think allows for a few % wobble) it could cause problems for the video card if the power delivery isn't stable.

yeah you could be right - in a dillema now - as I've no idea if card or PSU

testing in my 380W HTPC not going to happen .... lol

if I buy a 100 quid PSU - say a Corsair 850 - I'd be well unhappy if it was the GPU after all

darn

juno_first · 22 Mar 2011 at 22:52

Buckster said:
yeah you could be right - in a dillema now - as I've no idea if card or PSU

testing in my 380W HTPC not going to happen .... lol

if I buy a 100 quid PSU - say a Corsair 850 - I'd be well unhappy if it was the GPU after all

darn

YEAH, same issue I have. I want to put my Zonar back in & run the card at stock voltage & maybe overclock it to 850Mhz as it says you can on the back of the box.

Oh, was the box to your 480soc not wrapped in celophane & the static bag of the card just have normal tape on it?

Buckster · 22 Mar 2011 at 22:55

I hear your pain

dillema of taking risk RMA'ing and risk of it just being returned at cost to me

just returning under DSR as not suitable (within the 7 days)

or keep using it - under stock speeds (760 is perfectly stable) - and perhaps RMA it to Gigabyte in a few months time

Yurihyuga · 23 Mar 2011 at 09:30

Buckster said:
yeah you could be right - in a dillema now - as I've no idea if card or PSU

testing in my 380W HTPC not going to happen .... lol

if I buy a 100 quid PSU - say a Corsair 850 - I'd be well unhappy if it was the GPU after all

darn

buy the PSU test it and if same problem just return both the card and PSU for a refund. PSU will be easy since it would be under DSR.

Problem is even though your power supply is suitable it could be faulty would make sense since they ran it fine but you can't unless you undervolt/lower clocks.

meggsc · 23 Mar 2011 at 11:39

I RMA'd my card after I was having problems with instability. The result of fault testing by Overclockers was:

"No fault found, the card ran all manufacturers tests for 8 hours stable, no crashing. Ran Furmark, Crysis, 3D mark vantage, Streetfighter 4, Heaven and Aliens Vs Predators."

It has been sent back to me today, after I asked for a credit note the day before they tested it (see Webnote 2) to purchase a better card.

I asked for a refund under the DSRegs and they've very kindly said I can have a refund, but I don't know where I stand with the postage. So this card could set me back an extra £24.68(£10 for testing, 2x Postage £7.34).

I still think there's problem with this card. It should run fine out of the box with the spec I've got.

In addition to my previous webnotes to OC's, I tested the 12V rails on my HX1000W with my old GTX260 (I'd RMA'd the GTX480SOC) and processor running the power supply test in OCCT. It was reading 12.14 in the BIOS, and 12.15 at a Molex with a Fluka meter.

Webnote 1
I received my Gigabyte GTX480SOC yesterday but it crashes when running Nvidia demos, sled, alien & endless city, particularly during intensive scenes. This happens at the SOC's shipped clocks of 820/1640/950. The only way I can get them all to run OK is when I underclock it to reference speeds 700/1400/924.

I thought it could be my PSU to GPU connections, but they are on separate 12V rails and I've swapped them around with no change.

I have plenty of airflow through the case and the CPU is WC'd with two rads right now. (setup is ready for a WC'd GPU eventually).

Nothing is OC'd atm. Typical temps for CPU cores is 21-36 deg. System 33deg
GPU V 1.06 MEM 1.54. I did see 78 degs for the GPU once.

GTA crashes badly at max settings, but it is a fresh install and I've not looked in that yet. Crisis Warhead seems to run OK!

Any ideas. It should really be stable at it's SOC stock speeds, or should I RMA it and start saving for a GTX570-580

Webnote 2

Further to my previous web note, I tried Farcry2, which again like the other games and demos I’ve tried, crashed after a while. I tried two Nvidia drivers. Orginal v258.96 as supplied on the Gigabyte driver CD, and the latest Nvidia driver v266.58.

I ran Prime95-64bit with my old GTX260 plugged in, just to check the cpu cores and RAM, and they seem to be OK after an hour.
I’m not used to my computer being this flakey and I was already a bit weary of the GTX480, but this card looked such a good deal at the time.

Could you issue a credit note so I can purchase a more stable GTX580, a Gainward GeForce GTX 580 reassuringly “GOOD Edition” for example?

HeX · 23 Mar 2011 at 12:05

TBH i've had the same problems with OCUK's RMA on GPU's.

I sent an EVGA GTX260 Super Overclock back (when they were new), because it wasn't stable at it's overclocked settings.

They returned it saying it was fine (yet it'd get a billion errors in OCCT etc).

So I RMA'd it to EVGA instead, who confirmed that it was indeed faulty, and sent me out a replacement card that worked flawlessly at its advertised speed.

I informed OCUK of this and got my testing fee and postage costs refunded.

So just comes down to if you want to RMA it to Gigabyte and then get OCUK to refund you, or send it back under DSR and soak up the costs of shipping and testing fees.