Asus Matrix r9 280X CF = nothing but the issues

[lehtv]

I don't really agree with that.

What part in their test process is flawed?

I can dig up 280X reviews showing past 500W on a single card

Please do. Mind you, a 8+8 pin card can't physically draw more than 150+150+75 = 375W of power. 500W is possible only with a triple 8-pin card.

you can be selective with the reviews

Excuse me? Go throwing accusations elsewhere.

 

[railven]

Will try that too, right now I'm reinstall the whole system.
Could you tell me what are your idle/full usage temps with this setup?

Sure, since I'm a stickler for heat. Here is the break down of my cards:

GPU1:
Reference Radeon HD 7970 + Accelero Heatsink

GPU2:
Sapphire DualX 7970 Non-Ghz Edition

GPU1 is a 3 slot design, GPU2 is about 2.5ish.

Originally with GPU1 in PCIE0 and GPU2 in PCIE1, I'd have the issue Blackened described where GPU1 was so big it was basically resting on GPU2. Temps for GPU2 were insanely high, easily hitting 70C+ on full load. The heat from GPU1 was just being dumped all over it. This also caused GPU1 to get high too since the heat from GPU2 was also being sucked into the fans.

So I swamped them. GPU1 is now on my bottom PCIE slot with about an inch clearance between itself and the PSU, and GPU2 is on the top PCIE slot with about an inch between it and GPU1.

Temps are right and clocks right now:
GPU1: 300/150, Fan @ 20%, 30C, 0.806v, Load 0%
GPU2: 300/150, Fan @ 25%, 30C, 0.96v, Load 0%

Let me load up a game, using my MMO Profile (where I'm more CPU bottlenecked thus no need to clock cards high)
GPU1: 925/1375, Fan @ 25%, 57C, 1.125v, Load 98%
GPU2: 925/1375, Fan @ 50%, 60C, 1.125v, Load 97%
FPS: 131

Now to use my "high performance profile"
GPU1: 1125/1575, Fan @ 25%, 69C, 1.175v, Load 95%
GPU2: 1125/1575, Fan @ 50%, 70C, 1.175v, Load 94%
FPS: 154

My system hard locked about 1 minute into my high performance profile. HAHA. If I had to guess my PSU gave the ghost. It wasn't temp.

I'm in the market for a new PSU but with my rebuild coming up in 5 months just going to wait.

 

[Arkadrel]

I switched from Nvidia to ATI yesterday and I wasn't even able to play any game comfortable since then

That smily face, after the fact that you report you cannot play, probably means its a bold faced lie.
People arnt normally happy about but something like that.

Then there is the issue of you haveing 10 posts, on this account.

Throll thread much?

List all your hardware, resolution, and if you are an nvidia employee lol.

Warning issued for thread crapping.
-- stahlhart

 

[KingFatty]

Nevermind, I see you already swapped cards. Uh, hmm, maybe try laying the computer on its side and open the case, so the heat might be less likely to accumulate around the upper card?

 

[blackened23]

What part in their test process is flawed?

Their test is GPU power consumption only. It is not entire system power draw. Further, you can view various overclocked and overvolted 7970 reviews, i'll look around later but there are several on HardOCP showing near 500W system power usage with a heavily over-volted Tahiti. Additionally, being that the test you linked is for GPU power only and excludes the system, you can easily see how the wiggle room with a 750W PSU becomes negligible once you do over-volt. If a card by itself uses 300W and then another uses 300W, and these are lenient figures because Tahiti, as mentioned, uses a ton of juice when over-volted - you can easily pass 750W. Heck, my 680 lightnings in SLI measured over 800W at that wall from being over volted. This was measured at the wall past 1400mhz and with over-voltage.

It all depends on whether the matrix cards are factory over-volted. I suspect they are, but i'm not sure.

Here's an example from HardOCP:

Let's say a second card is using 300W in the same system. This will pass the 750W used by the OP. Just to be clear: you absolutely can and will pass 750W with heavily overvolted GPUs in mGPU configuration. I have done so myself as measured at the wall. ESPECIALLY if you're also Oc'ing a CPU in tandem with a GPU overclock.

 

[stahlhart]

The next post in this thread that isn't intended to provide the OP with assistance with their GPU issue is going to earn the source of it a vacation. Knock off the thread crapping now.
-- stahlhart

 

[b3ka]

Just finished testing in clean OS. Same thing, going around 80 deegres and drop to ~70% with choppy performance.

 

[Fastx]

Just some thoughts/things to try fwiw I just skimmed though this thread so just some quick things mentioned below. I am using Trixx per below steps.


Disable ULPS and restart computer.

Make sure after restart powerboard set to +20%

Have box checked for synchronize cards in multi-GPU config

Check your vram temps under load if you have vram sensors on these cards in CF this could cause throttling.

You could test each card separately (one card in at a time ) playing games (monitor usage, core clocks, temps,) and bench each card under full load in each pci slot separately with only one card in the system to check each card.

AMD fwiw recommends 850 watt power supply min for the 7970 GE not showing anything for the 280x.

 

[Imouto]

Are you using the CFX bridge? 290/290X don't need them but these rebadges do.

Oh, and if you're using it ensure that you're using only one.

 

[cmdrdredd]

Maybe your system airflow is poorly laid out? It's possible that your case is trapping heat.

What case are you using BTW? Other than a driver issue currently that needs fixing I can't think of another reason this should happen.

 

[Elfear]

OP - Did you plug in the auxiliary 4-pin molex connector near the top of your first PCI-E slot? It's meant to give extra power to the GPUs. Could be the cause of your instability if it's not plugged in.

 

[lehtv]

[...] you can view various overclocked and overvolted 7970 reviews, i'll look around later but there are several on HardOCP showing near 500W system power usage with a heavily over-volted Tahiti. Additionally, being that the test you linked is for GPU power only and excludes the system, you can easily see how the wiggle room with a 750W PSU becomes negligible once you do over-volt.

If a card by itself uses 300W and then another uses 300W, and these are lenient figures because Tahiti, as mentioned, uses a ton of juice when over-volted - you can easily pass 750W.[...]

Yet the OP's cards are not overvolted beyond the the Matrix factory setting, so that point is completely moot. All that need to be considered are the power consumption measurements in reviews at factory settings.

Here's an example from HardOCP:

[...]

Let's say a second card is using 300W in the same system. This will pass the 750W used by the OP.

No it won't. First off, I think the HardOCP's 7970 Platinum power consumption figures are MUCH higher than they would be for 280X Platinum, as per the hexus.net graph posted. But even if we assume they aren't:

System load with video card = 408W
Add another video card = 300W
= 708W

While this is already below the 750W mark, also remember this is a AC watts, not DC watts. Adjusted for efficiency using HardOCP's test unit (Enermax 1350W), the load wattage becomes 708W * 0.9 = 640W, well below the OP's Seasonic 750W unit's maximum capacity. So as long as the OP doesn't overclock the cards beyond their factory setting, we can safely rule out the PSU as being in any way the cause of the issue he's experiencing, unless of course the PSU is faulty.

 

[railven]

OP - Did you plug in the auxiliary 4-pin molex connector near the top of your first PCI-E slot? It's meant to give extra power to the GPUs. Could be the cause of your instability if it's not plugged in.

<snip>

I've never seen a board with that kind of a connector (ie auxiliary power for GPUs) so i checked the manual and found some interesting things:

ftp://66.226.78.21/manual/Z77 WS.pdf

For starters, the PCIE setup for CFX/SLI. It isn't pick and choose, it seems to be PCIE1 and PCIE5. Page 22 in the manual.

And that connector your noticed in the pic, states to connect a power connect when using two or more GPUs. Page 48 in the manual.

Good catch Elfear. OP I'd double check to make sure you have your cards setup on the right lanes and that auxiliary power connected.

 

[blackened23]

That's....interesting as to how ASrock splits their dual GPU slots. To be clear, nearly every motherboard on the market does have specific slots required to be used for 2 way SLI, it will be specified in the manual - if you don't use these specific slots i'm assuming that the speed will be screwed up (you might get x4, or PCIE2.0, for instance). Anyway, i've always used asus motherboards and - like all non workstation motherboards, it explicitly requires 2 specific slots for 2 way SLI.

I guess that ASrock motherboard has more space between 3 slot GPUs which isn't the norm. I know on Asus that I use and some MSI/GBT motherboards that buddies use, that the 2 required slots are a little closer to each other - making 3 slot cards utterly worthless for SLI/CF.

I guess this is kind of a tangent comment, but I didn't know that about ASrock motherboards. Perhaps 3 slot can work afterall on that board. I have a hard time seeing it though, 3 slot cards are truly monsters in size and i've tried so many times in the past to get them to work, with it being a complete exercise in frustration. Then again I didn't use any motherboard like this one. Is this a workstation type motherboard with the PLX chip?

 

[Mastercommander]

List all your hardware, resolution, and if you are an nvidia employee lol.

Warning issued for thread crapping.
-- stahlhart

AMD has its problems, deal with it

 

[Mastercommander]

I switched recently and had the same experience.

Turn off Ultra Low Power State in the registry

Set increased power limit to the max

Make sure to have only one overclocking program running

 

[VulgarDisplay]

AMD has its problems, deal with it

I honestly responded to get the extra response needed to start trouble shooting. I tongue in cheek commented on the unnecessary part of the OP commenting on how he went from No problems nvidia to all problems AMD.

I know both vendors have issues. Either way. I'm probably about to get a vacation for OT posting. But on that topic, the forum already has a moderator so you had no reason to post that.

 

[tential]

I switched recently and had the same experience.

Turn off Ultra Low Power State in the registry

Set increased power limit to the max

Make sure to have only one overclocking program running

HOW?!
Where is this setting?

I forgot what setting I needed to change to fix HDMI Audio and this is it. You're right, AMD does has it's problems and Ultra Low Power State destroys Audio through HDMI I just don't get why they won't fix it. I read threads on it and it goes back for AWHILE now. (Edit: To be clear to any potential buyers, this problem problem won't be relevant to you since most people game with headsets and don't use their setup as an HTPC gaming system. It also doesn't exhibit itself in gaming. Mostly Youtube, and VLC where 2D clocks aren't high enough to process audio (I don't get this at all.....) WMC works fine though and WMP)

True Audio announcement rustled me hard because you want to focus on Audio but wont fix an issue that's been around for 2 generations of cards (Now 3)?

 

[b3ka]

Just some thoughts/things to try fwiw I just skimmed though this thread so just some quick things mentioned below. I am using Trixx per below steps.


Disable ULPS and restart computer.

Make sure after restart powerboard set to +20%

Have box checked for synchronize cards in multi-GPU config

Check your vram temps under load if you have vram sensors on these cards in CF this could cause throttling.

You could test each card separately (one card in at a time ) playing games (monitor usage, core clocks, temps,) and bench each card under full load in each pci slot separately with only one card in the system to check each card.

AMD fwiw recommends 850 watt power supply min for the 7970 GE not showing anything for the 280x.

ULPS disabled, +20% power set, will check if VRAM temps are sensors are present. I forgot about that extra connector on motherboard, will check that too.

I switched recently and had the same experience.

Turn off Ultra Low Power State in the registry

Set increased power limit to the max

Make sure to have only one overclocking program running

That is interesting, how do I turn off ultra low power state? Won't it get the cards hot in normal usage when the power isn't needed?

Right now i'm testing on 100% fan speed to see if it's temps that cause the issue
TESTED 100% fan don't do much as temps are still going up to 80-81 for 1st one and 70 for 2nd. The issue happens too.

 

[tulx]

You bought TRIPLE SLOT cards to crossfire with? That was your first mistake. I don't care what kind of case you have, your temps will get too high and you won't be able to maintain temperatures no matter what kind of GPU you have (nv or AMD).

You can't crossfire or SLI triple slot cards. I've tried it.

Works flawlessly for me. I use two XFX 5870ies with Accelero aftermarket coolers.

 

[RussianSensation]

Also, aside from this, your PSU is not powerful enough for these cards as they are factory overclocked. Tahiti uses a ton more power when overclocked and the Matrix is 1100 factory OC'ed IIRC? One card alone can do near 500W power draw with 1.3V on the Tahiti, let alone two.

The PSU should be enough. Your power consumption figures are way off on this one blackened.

1100 / 1650 MHz @ 1.188V : 200W
1100 / 1650 MHz @ 1.256V : 227W (default)
1200 / 1700 MHz @ 1.256V : 251W
1225 / 1700 MHz @ 1.275V : 281W
1225 / 1700 MHz @ 1.256V & LLC 25% : 294W
http://www.hardware.fr/articles/853-6/asus-hd-7970-matrix-platinum-test.html

2 of his cards are then using 460W of power at stock clocks.

OP, what games specifically are you playing? Did you install CAP profiles?

 

[b3ka]

I'm doing tests on BF3/Valley Benchmark.
I'm currently waiting for other PSU to test.
BTW. Found a person with same issue on other board but the guy used MSI card and they offered new firmware to fix the issue. I will attach usage screenshots when I'm ready with PSU.

 

[Fastx]

ULPS disabled, +20% power set, will check if VRAM temps are sensors are present. I forgot about that extra connector on motherboard, will check that too.

That is interesting, how do I turn off ultra low power state? Won't it get the cards hot in normal usage when the power isn't needed?

Right now i'm testing on 100% fan speed to see if it's temps that cause the issue
TESTED 100% fan don't do much as temps are still going up to 80-81 for 1st one and 70 for 2nd. The issue happens too.

Just so you know Ultra Low Power State = ULPS which you said you already have it disabled above. It will not make the card (s) run hot in normal usage disabled. Ultra low power state (ULPS) will just power down the slave card in CFX to a lower state than normal at idle. When gaming it is recommended to disable ULPS to avoid any of poss reduced power/ throttling.

 

[SolMiester]

Thanks for your input, I've got 14 days to return, still have time to upgrade PSU and check if this is the issue. Anyone else with CF can say something about the issue?

You could get a Raven 03 case switch will resolve the heat issue of the top cards by the rotated motherboard...

 

[DigDog]

You could get a Raven 03 case switch will resolve the heat issue of the top cards by the rotated motherboard...

or even simpler, lay the case on its side ...