ASUS Matrix is a joke OC'er

[Devil_Dog]

Just ran one more quick one @ 945. Seems like I still have some room with these on stock cooling. Screenie in link.

http://hwbot.org/submission/2199748_

 

[Kohta]

Just ran one more quick one @ 945. Seems like I still have some room with these on stock cooling.

thats a great clock and all but is 3DMark11 all you use for a stability test? I ask because iv'e ran both the latest 3DMark's, passed them several times and still crashed or had artifacts playing Final Fantasy XIV (This is my main game) This is also my most demanding game so it serves as a great stability checker, a couple hours of FFXIV will have your card hotter than any of these benchmarks and tests ever will. Whatever 3dMark 11 gets you up to, add +15C-20C.

This is how i test stability:

1. Raise clocks/voltage
2. Quick test running 3DMark Vantage - Followed by 3DMark 11
3. Heaven Benchmark for 1 hour on 1920x1080 - Normal Tessalation - 16xAF - 8XAA
4. Endless City Demo 800x600 & Final Fantasy XIV Benchmark 1080p looping for 30 minutes.
5. and lastly, Final Fantasy XIV 1080p benchmark looping, 800x600 Endless city Auto pilot and 800x600 haven benchmark normal tessalation - 16XAF - 4xAA all at once for 1 - 2 hours.

I over nighted another matrix packing this other one up, if i get it and the same issue, then i will gladly say user error, but i think it's a string of crappy chips, i replaced the TIM on my CPU from AS5 to IC Diamond, that's some amazing stuff, my temps are down 8C across the board, and which ever card works for me will get some of this stuff, it's stupid thick so it should be perfect.

Unrelated

 

[Theocnoob]

Have you tried with another PSU? Are you sure the cards are sitting properly in your slot? I had a 300 case that was slightly warped and cards didn't sit quite right as a result.

My DCUII, which is the predecessor of the matrix, does 1000Mhz at 1.15

 

[Kohta]

Have you tried with another PSU? Are you sure the cards are sitting properly in your slot? I had a 300 case that was slightly warped and cards didn't sit quite right as a result.

My DCUII, which is the predecessor of the matrix, does 1000Mhz at 1.15

yeah that was the first thing i checked, it's perfectly in the slot i had my Noctua NH-D14 off today so i triple checked it while i was able to see it clearly, iv'e tried 2 different PSUs, right now i have it on my Corsair Enthusiast 850watt, it was working on this PSU just fine when i first got it. Yesterday evening i hooked up the Corsair professional 1200w and it didn't make any difference, i can't believe at one point i wanted 2 of these, if i can get just one of them to OC to 900-920mhz i wouldn't need/want a second one.

 

[Theocnoob]

yeah that was the first thing i checked, it's perfectly in the slot i had my Noctua NH-D14 off today so i triple checked it while i was able to see it clearly, iv'e tried 2 different PSUs, right now i have it on my Corsair Enthusiast 850watt, it was working on this PSU just fine when i first got it. Yesterday evening i hooked up the Corsair professional 1200w and it didn't make any difference, i can't believe at one point i wanted 2 of these, if i can get just one of them to OC to 900-920mhz i wouldn't need/want a second one.

I didn't think you would have issues with TX/HX PSUs but it was worth a shot.
How high are the temperatures getting? Do you have the fans locked on medium/high? Or on a profile to start on low-med and ramp up? I have to keep mine cooler than the 'stock fan profile' would allow for it to be happy in the 900s Mhz. It gets to like 80 something on normal fans but if I set the fans to start on like medium and go to high if it hits like 60, that works well. It rarely exceeds 55 even at high vcore. The matrix has bigger fans so you won't have to set them as high. That should solve any temperature issues. Once you get one that you know you want to keep, you might want to reseat the cooler. I thought their TIM was crapp. I replaced with MX4 and it took a good five degrees or so off my load and idle. That's pretty good. You're on the latest drivers from Nvidia?

I think it's odd that you are having issues with these cards and I'm sort of tempted (once you get the temps sorted) to wonder if there is an issue elsewhere in your system that is causing this somehow. You've tried the card in another slot right? It'll only pull 8X in the second slot but it's good to troubleshoot.

 

[Kohta]

I didn't think you would have issues with TX/HX PSUs but it was worth a shot.
How high are the temperatures getting? Do you have the fans locked on medium/high? Or on a profile to start on low-med and ramp up? I have to keep mine cooler than the 'stock fan profile' would allow for it to be happy in the 900s Mhz. It gets to like 80 something on normal fans but if I set the fans to start on like medium and go to high if it hits like 60, that works well. It rarely exceeds 55 even at high vcore. The matrix has bigger fans so you won't have to set them as high. That should solve any temperature issues. Once you get one that you know you want to keep, you might want to reseat the cooler. I thought their TIM was crapp. I replaced with MX4 and it took a good five degrees or so off my load and idle. That's pretty good. You're on the latest drivers from Nvidia?

I think it's odd that you are having issues with these cards and I'm sort of tempted (once you get the temps sorted) to wonder if there is an issue elsewhere in your system that is causing this somehow. You've tried the card in another slot right? It'll only pull 8X in the second slot but it's good to troubleshoot.

All good suggestions, the reason i haven't replaced the TIM is because IC Diamond cost $25 for a decent sized tube, because of it's viscosity it takes twice as much over AS5, it's probably thicker than MX4, i haven't owned any in a long while and i think it was called TX3 or TX4 back then but it was thick, but not this thick, this stuff is like slightly moist play dough.

I have it set to 40% fan speed as the lowest/idle speed since that's not any louder than my Antec 1200 case fans running, here's a screen cap of my fan profile i made, it typically never gets past 63C, and i make sure it doesn't pass 75C just for the sake of benchmark testing, i have to admit though, the fans are "too effective" when it hits the 60C threshold it just bobbles the fans back and forth in most cases, up-down-up-down, iv'e tried eliminating it but if you look at the 50C fan speed, it will climb above my market at that fan speed and hit 60C, if i set the 50C fan speed to 60% (only 3% more) it will just bobble again, i wish it had a speed curve to it, instead of going straight from A-B speed, I know i can do that in MSI afterburner but i'm a bit OCD i want that kind of thing on the BIOS, not having to wonder if i closed MSI after burner by accedent or what-not, games like FFXIV you can't just ALT+TAB to look at background/desktop stuff, it will crash the application, it's ment to to that by design to fend off hackers/botters.

Lastly, i did put it in the other slots and i had the same issue, i put this card in my wife's rig and it gave artifacts in her Photoshop, so it's definatly the card.

Also, the Memory is alittle hotter than it should be in that photo because i was in the middle of video editing.

 

[Theocnoob]

Maybe you got multiple cards from a bad batch. That temperature is fine and artifacting in PS is definitely screwy.

 

[Kohta]

Maybe you got multiple cards from a bad batch. That temperature is fine and artifacting in PS is definitely screwy.

Hey OC, if you get a free chance run the FFXIV benchmark.

http://www.finalfantasyxiv.com/media/benchmark/na/

(The fastest download is probably fileplanet)

When it opens up, choose the "Miqo'te" reason being is because this benchmark forces soft-body Physics on the GPU instead of the CPU, which would be her tail, hair, movement and armor, the CPU is going to be tasked with some textures so your CPU speed will also effect your score. The Miqo'te is the most demanding one of the tests because of the physics. and run it on (High) 1920x1080, make sure you don't have any AA/AF force in the Nvidia CP, i'm eager to see what kind of results these cards are supposed to produce.

If your card isn't stable, you will see one of three things happen, the benchmark will crash, the shadows will flicker or not render at all, or you will see the FPS bar in the benchmark dip like a heart monitor, the reason it does this is because the way it's designed is if it has any errors, such as the shadows not being there or the physics get screwy, it will re-render the scene causing you to drop in fps to 0.

If you have multple GPU's i can walk you through how to enable them both for the benchmark, i ran this test on my first Matrix and scored 7000-ish, which seemed low considering my 5970 got alittle over 8200 overclocked, and the GTX 580 is supposed to be nearly neck and neck with this benchmark, according to other testers, this game and benchmark is optimized for Nvidia, that's why it should come real close to the 5970, and iv'e seen it do it before.

 

[Theocnoob]

Happy to help. Downloading the file now. I will run it and let you know in a few minutes.

This is my system btw

2600K
4.5Ghz
Asus Maximus IV 1809 bios
1600Mhz 8 8 8 24 2N 2x4GB DDR3
120GB SSD boot
GTX 580 DCUII

 

[Kohta]

Happy to help. Downloading the file now. I will run it and let you know in a few minutes.

This is my system btw

2600K
4.5Ghz
Asus Maximus IV 1809 bios
1600Mhz 8 8 8 24 2N 2x4GB DDR3
120GB SSD boot
GTX 580 DCUII

That's perfect, our systems are nearly identical, though i use my SSD as a main HDD, and my memory is 2133mhz, i had to lower my OC to 4.5ghz because i was at 5.0 @ 80C, this matrix pushed the heat directly into my heat sinks fins making it run 10-15c higher, if i'm doing prime95 and running a windowed benchmark my CPU hits 91c over the period of an hour, or atleast it did when the card was benching, 4.5ghz @ 1.365v puts me around 70C load with the card blowing heat into it now, so this'll be perfect. not to mention i feel better with 1.365v versus 1.52v

 

[Theocnoob]

I got 6700 points at 840Mhz... Hang on though had 'adaptive power' bs on...

K, 7158 @ 940Mhz. HTH.

 

[Deanzo]

Just FYI, the cores are not cherry picked.

You are paying for what they have done/added to the PCB.

Which imo may help if the core is up to it.

Aslo, that you can find other cards doing more with less v means very little.
The very few that have crazy clocking cards are going to post all about it, where everyone else doesn't.
I've had/used over 20 580's, not that I really do much air testing as it's pointless, but I've never got close to 900mhz
Though thats just a number, what should really matter to you is, how many more fps does this card give you over the last, not some silly % number for the overclock.

 

[Kohta]

I got 6700 points at 840Mhz... Hang on though had 'adaptive power' bs on...

K, 7158 @ 940Mhz. HTH.

That's what i got at 940 back on my old thread with my card that started artifacting on me.

http://www.overclockers.com/forums/attachment.php?attachmentid=98022&stc=1&d=1312442345

but see, i had 940 at 1.125v and eventually i kept having to raise the voltage because it would crash or artifact while i was playing, all while keeping temps in check, finally i decided to make up a ridiculous benching method, like running 3 GPU heavy simulations at once and i found out 890mhz required 1.15v, long story short, i RMA'd, got this one, and the same thing happened, only it happened quicker because i went ahead and used my little benchmark method on each clock and voltage, artifacts appeared and didn't go away, even at stock.

The good news is my artifacts are gone now and before i pack this one up in 3 or 4 hours to sip it back off i'm going to mess with it again, my new one should be here around 8am or 9am.

is your 940mhz at 1.15v?

 

[Kohta]

Just FYI, the cores are not cherry picked.

You are paying for what they have done/added to the PCB.

Which imo may help if the core is up to it.

Aslo, that you can find other cards doing more with less v means very little.
The very few that have crazy clocking cards are going to post all about it, where everyone else doesn't.
I've had/used over 20 580's, not that I really do much air testing as it's pointless, but I've never got close to 900mhz
Though thats just a number, what should really matter to you is, how many more fps does this card give you over the last, not some silly % number for the overclock.

When you're trying to find the max potential overclock, which is what every overclocker does at some point, you will push it as far as it will go, and then figure out what you're willing to settle with, 890mhz is not what i am willing to settle with, and this isn't a reference cooler, so i should be able to pass what a reference design can do on it's lowest end, i have expectations and standards, and the minute i spend more money on a card than a reference designed and accept a mediocre overclock, is the day mine and everyone else's standards have died.

890mhz @ 1.15v is not acceptable, i would tell anybody to send it back if they bought a matrix and couldn't pass that. Let's not get on the topic of my opinion and how it is my own, standards and expectations included, been through this already with another member and i won't settle for less-than, you can by all means but i buy premium to overclock, not accept until the next thing comes along, and that is my decision to make, if i deem this card an overhyped-overocker that's for me to say, and so far, that's how it is, maybe 9AM can change my mind.

Also, don't forget they both shot me artifacts on different occasions.

Additionally, based on what you said, i'm paying for what they did to the PCB if the core is up to it, absolutely, but why would i waste my money getting a core that isn't up to it, we don't buy Corvette's with a Geo Metro 3 cylinder and consider that acceptable, and what happens when you buy a corvette with it's correct 430hp engine, but it spits and sputters, we take it back and use our warranty. A $500 video card is the same as a $50,000 car in both worlds comparatively. aka. it's a lot of money for what it is.

 

[Kohta]

I got 6700 points at 840Mhz... Hang on though had 'adaptive power' bs on...

K, 7158 @ 940Mhz. HTH.

One more thing OC, if you have the Heaven Benchmark, give that a shot, 1920x1080, normal tessalation, 16xAF, 4xAA, this card won't finish that benchmark at 830mhz on stock voltage, i don't care what GPU core you have, that's crap, around the part of the benchmark where you see the grass start to blur overlooking the island is where it stalls, 830, 820 all stall, i can OC it to 818mhz and get through it on stock voltage, but that's not what i;d even call an OC.

 

[Theocnoob]

That's what i got at 940 back on my old thread with my card that started artifacting on me.

http://www.overclockers.com/forums/attachment.php?attachmentid=98022&stc=1&d=1312442345

but see, i had 940 at 1.125v and eventually i kept having to raise the voltage because it would crash or artifact while i was playing, all while keeping temps in check, finally i decided to make up a ridiculous benching method, like running 3 GPU heavy simulations at once and i found out 890mhz required 1.15v, long story short, i RMA'd, got this one, and the same thing happened, only it happened quicker because i went ahead and used my little benchmark method on each clock and voltage, artifacts appeared and didn't go away, even at stock.

The good news is my artifacts are gone now and before i pack this one up in 3 or 4 hours to sip it back off i'm going to mess with it again, my new one should be here around 8am or 9am.

is your 940mhz at 1.15v?

1.1

It'll do it on less. I run 1.1 for guaranteed stab.

 

[Tokae]

I ran that test a week ago. But with dual 6970s not 580s.

8566 on high!

http://www.overclockers.com/forums/showpost.php?p=6933729&postcount=134

Sorry for the threadjack but no one else ran this test.

 

[Kohta]

I ran that test a week ago. But with dual 6970s not 580s.

8566 on high!

http://www.overclockers.com/forums/showpost.php?p=6933729&postcount=134

Sorry for the threadjack but no one else ran this test.

I don't mind at all, in fact, here is how to run dual GPU's on the benchmark.
open up the FFXIV Benchmark folder, the open the data folder and rename "FFXivWinBenchmark" to 3DMark06

Next Download this:
http://translate.google.com/transla...seesaa.net/article/153685727.html&sl=ja&tl=en

Extract it and place it inside the data folder where you renamed the file. Then run the benchmark using what you just renamed and it should screen wrap and use both GPU's.

If you've done this and still got 8566 then that doesn't seem right, a couple of 6870's not overclocked can hit 8500 with no problem, the 6970's should be atleast 20% better.

 

[Kohta]

The new one just came in an hour or so ago, befor ei even installed it i ripped the heatsink off and replaced the TIM with IC Diamond, the TIM that was on it didn't even cover the whole GPU, speaking of, these GPU's are HUGE compared to ATi's, you could fit 3 or 4 ATi/AMD chips inside one of these monsters.

Put it all back together, tossed it in, fresh install of drivers and immediatly began OC'ing.

This time the results were 1000% better.

1.075v (stock)
818......√
820√
830√
...
880√
900X

1.088v
900√
910√
920√
930√
940X

1.1v
940√
950√
960X

1.113v
960X

1.125v
960√
970X

1.15v
970√
980√
990√
1000X

1.23v
1000√
1010√
1020X

Even though i can do 1000mhz, 100% fan speed could not hold onto that 1.2v even with the awesome TIM. it hit 78C max with 100% fan speed during any of the benchmarks and it climbed quick.

For the sake of heat/longevity i decided to go with 1.1v 950mhz, it runs cooler, staying around 60c at 50% fan speed and the benchmark results were not much different from 950 - 970, maybe 25-30 more depending on the benchmark, there isn't any difference between 950 and 960mhz on FFXIV benchmark, so i may have hit the CPU bottleneck, i doubt it, but it's possible or that benchmark atleast.

 

[Tokae]

I don't mind at all, in fact, here is how to run dual GPU's on the benchmark.
open up the FFXIV Benchmark folder, the open the data folder and rename "FFXivWinBenchmark" to 3DMark06

Next Download this:
http://translate.google.com/transla...seesaa.net/article/153685727.html&sl=ja&tl=en

Extract it and place it inside the data folder where you renamed the file. Then run the benchmark using what you just renamed and it should screen wrap and use both GPU's.

If you've done this and still got 8566 then that doesn't seem right, a couple of 6870's not overclocked can hit 8500 with no problem, the 6970's should be atleast 20% better.

Hmm no idea as to why I got that score then. I run dual monitors and it went to fullscreen mode on my main one and I could see it using the crossfire on the 2nd one.. Perhaps I will try it again!

*Edit*

Is there a screenshot somewhere of dual 6870s for the score? I had my cpu @ 4.6 ghz, both cards @ 950 core mhz and 1425 memory mhz for that run and that is what I ended up with.. I also ran 3dmark11 just before that FFIV benchmark and came up with P10113 marks