Furmark burnin not pushing GPU clock speeds.

[Redrocket90]

So I decided to OC my Asus Strix R9 285OC to boost my ego in Firestrike, and DL'd Furmark to test stability.

I overclocked the GPU from 954Mhz (stock) to 1.1Ghz without touching the voltage. I ran Furmark for roughly 20 minutes to make sure it was stable enough to get through a bench and temp leveled off at about 78C.

I then ran Firestrike, which crashed on the second graphics test. Erm, ok? So I bumped the voltage a bit and retried. Same thing. Ran in Furmark and crashed in Firestrike. Ok, time for some investigating.

I discovered that Furmark was running my GPU in the mid 800s to 900s instead of the 1100 it was supposed to. Then I ran just enough Firestrike to see what the peak clock was in that and it showed 1100Mhz as it should. I couldn't actively monitor during the Firestrike test, but I assume it was running the GPU wide open, which is why it crashed during Firestrike and not Furmark.

I did get a Firestrike pass @ 1050Mhz, but I can't get a good stress test because a stress test doesn't do me any good if it's running at 85% clock speeds.

Any idea what might be going on?

FWIW I just bought 3dmark's full version on steam, which now includes a stress test, but it would be nice if I could get Furmark working, too. I like to run both CPU and GPU burnin to test temps at full gaming load (which turned my CoolerMaster HAF XB into a space heater, btw)

I'll update this after I run the 3dmark stress test to verify that it works.

Athlon X4 860K @4.3Ghz
Asus A88X pro
8 GB (2x4)
Strix R9 285OC @1050Mhz

Using openhardwaremonitor for temps and clocks.

 

[SinxarKnights]

Furmark is throttled to prevent damage.

You can try to rename the executable to get around it but not sure that works anymore.

 

[Redrocket90]

Furmark is throttled to prevent damage.

You can try to rename the executable to get around it but not sure that works anymore.

Isn't the whole point of it to test stability? How can it do that if it throttles itself?

 

[SinxarKnights]

Isn't the whole point of it to test stability?

Not really, the point of it is to cause an overheating condition.

It is throttled because of the overcurrent protection built in. It is very possible the VRMs could burst into flames if it was not throttled, potentially causing a short when the graphics card goes up in flames taking the rest of your PC and/or your house with it (not kidding, this is why overcurrent protection is there).

 

[Redrocket90]

Isn't the whole point of it to test stability?

Not really, the point of it is to cause an overheating condition.

It is throttled because of the overcurrent protection built in. It is very possible the VRMs could burst into flames if it was not throttled, potentially causing a short when the graphics card goes up in flames taking the rest of your PC and/or your house with it (not kidding, this is why overcurrent protection is there).

So is it Furmark that is doing the throttling? If so, how does it know to throttle since nothing else seems to share that same "courtesy"?

Also, if I'm not adjusting the voltage, shouldn't I still be safe in the VRM department?

 

[SinxarKnights]

No it is the BIOS of the graphics card that does the throttling. It is the type of load. Furmark and OCCT are designed specifically to pull as much power as the PSU can deliver and overheat the card. It is known as a "power virus". Furmark used to be blocked at the driver level and you could apparently rename it to bypass the limiter and destroy the card. Now though it has proper protections built in so any power virus will be throttled properly.

There are ways around it if you want to destroy it that badly. You can mod the BIOS to override the protections. It really has nothing to do with the voltages used for overclocking. It is the VRMs that regulate power to the entire card, the type of load Furmark uses causes an excessive amount of current (aka amps) draw. The VRMs regulate the voltage at a stable voltage while adjusting for current draw, all dynamically of course, but with a lot of waste heat (voltage regulation circuitry is quite interesting).

Some Nivida cards do this as well but not sure about my 970, never tried it. My R9 280 had the same thing happen as with your 285. That is when I read up on the subject and found out that this has actually been the case for quite some time with a lot of AMD cards, but not all though. It appears to be architecture dependent and I believe all GCN AMD cards has this built into the BIOS.

 

[Redrocket90]

Ok, i really appreciate the replies. It makes sense that my card would throttle to protect itself, but im still unclear why it only seems to be doing it for Furmark, while 3dmark has no problem kicking it's teeth in. In the Furmark burnin, it didnt even TRY to get the clocks to top out. Heck, they barely ever even approached stock speeds.

btw, i ran the 3dmark stress test and passed with 99.9% stability at 1050Mhz.

 

[SinxarKnights]

Try Valley with these settings in the first image: http://www.bit-tech.net/hardware/graphics/2014/05/19/amd-radeon-r9-280-review-feat-xfx/8

Here is the link to the valley benchmark:https://unigine.com/products/benchmarks/valley/

Here is what I got with my R9 280:

Sadly I deleted my old 3dmark scores and only kept the highest with my current card: http://www.3dmark.com/fs/8093686 (this is for a 970 so not really helpful).

What is in your PC?

 

[Redrocket90]

I know everything is running well and performing as it should be harware wise. I hold the 3rd highest firestrike score of all the AMD X4 860K with R9 285 single card systems. I just want to properly stress test my card at the overclocked speed so I know if I can get any more out of it while staying safe.

I'll try the Unigine test to see if that will push the clocks harder than furmark.

 

[SinxarKnights]

Yeah Furmark isn't good. Just forget it and move on.

Got links to your 3dmark scores? Heaven I found to be good for finding artifacts, I was able to push my 280 higher but in the smoke in Heaven you could see it artifacting on that. I got the 280 before the 285 came out and I heard it doesn't overclock nearly as well.

 

[Redrocket90]

This is the best to date for mine. http://www.3dmark.com/fs/8933150

I got the 285 up to 1073Mhz stable, but I also bumped the Vram up to 1500, which I think may have pushed things in the opposite direction because it dropped my Firestrike score a fair amount. Is that a thing? can overclocking the Vram slow it down? It was still running fine, it just hurt the bench scores in valley and firestrike.

I might try to get my CPU to 4.5, but I think 1073 Mhz on the GPU with 1450 Mhz Vram is about all I'm gonna get out of the graphics. My graphics are leaps and bounds ahead of my CPU anyway, so it's pretty pointless to try to push the graphics card too much harder until I get a better CPU.

One other mistake I've made: rather than doing it right by maxing each individual component one at a time, I've been giving it little bumps here and there. A little Vclock here, a little Vram there, maybe a pinch of voltage for flavor.

I might start over and do it right, this time.

Do you recommend finding the limit of the card without touching voltage, or is it ok to overvolt the card like you do a CPU? If so, is there a good utility to use for that? I've been using Crimson for all my graphics overclocking so far, which doesn't show the actual GPU voltage. Only a %.

 

[SinxarKnights]

That isn't for voltage. It just allows the card to consume more power before being throttled by letting it exceed the TDP a bit. It is fine to set it to 110% or 115% for overclocking.

I took a bunch of pics of my old scores but for the life of me I can't find them anywhere. And yeah if the card slows down after overclocking something on it is likely at the limit and unstable. I got that too once it was unstable on the memory.

Over on bit-tech they had the Powercolor r9 285: http://www.bit-tech.net/hardware/graphics/2014/09/02/amd-radeon-r9-285-review/1
On the overclocking page they were fairly close to what you got, 1030 core/1550 memory.

So yeah that is pretty good so far, you can try increasing the power limit to see if the memory is stable at 110% or so.

 

[Redrocket90]

Yeah, I think you're right about furmark. Its fine for testing cooling, but as far as stability goes, it's useless.

I think I'm pretty happy with where I'm at on my GPU. My CPU already can't keep up. Now if I could just figure out to get past this wall I'm at on my CPU clock...

Thanks for the help.

 

[Redrocket90]

Well, I lowered the settings to 1050 core, 1425 mem, 105% power, and managed to get my CPU stable at 4.4Ghz. I got my best Valley, Firestrike, and passmark tests to date. The first time I ran firestrike was a 100 point improvement, but it had a time error so my test was invalid 🙁 That would have put me at #2...

I might have to try it at 4.5Ghz.

 

[Redrocket90]

Bruhaha! I got it! I ran it again and got #2. http://www.3dmark.com/fs/9024270

Crashed at 4.5 tho. Maybe later.