Futuremark Releases 3DMark Stress Tests

[btarunr]

Futuremark released a new stress-test mode for the popular 3DMark suite, which lets you test the stability of your machine, after all the overclock/volt wizardry is done. What this mode does is runs a continuous 3D scene from the bench of your choice ("Fire Strike" (standard, Extreme, and Ultra,) and "Skydiver"), in a continuous rendering, without stopping for any loading screens or other breaks. This is different from the "loop" custom test setting, which runs each test over and over again, with pauses to reload the scene.

In the free edition, you can stress for up to 10 minutes. The Professional Edition key lets you scorch your hardware for up to 40 hours. Stressing your hardware is not all that this mode does, there's also a stability-test component, which checks for frame-rate stability through the defined cycles of your stress. The app would flag your hardware as "passed" only if it can achieve a frame-rate stability of at least 97% (it should tell you if your overclock is not sustainable as it runs into a thermal throttle). Simply update your 3DMark installation to the latest version, to use this feature.



View at TechPowerUp Main Site

 

[bug]

Probably not the best place to ask this, but does anyone know what is 3DMark Firestrike adding to Windows startup during installation?

 

[RejZoR]

Funny, I talked about this a week ago. And now here it is. Though, not exactly what I had in mind. I'd prefer the usual Firestrike benchmark, that simply loops forever, like Heaven and Valley benchmarks. That would be a real world load, not some synthetic over the top demanding mode...

 

[Air]

Thats interesting. Any plans to add this test to reviews?

 

[bug]

Funny, I talked about this a week ago. And now here it is. Though, not exactly what I had in mind. I'd prefer the usual Firestrike benchmark, that simply loops forever, like Heaven and Valley benchmarks. That would be a real world load, not some synthetic over the top demanding mode...

Stressing using "real world load" is not that useful. What if the "real world load" changes to affect parts of the GPU that were not stressed before? It's better to stress as many area of the GPU as possible, imho.

 

[RejZoR]

Firestrike stresses graphic card enough, I'd just need the option to loop the existing benchmark till I end the test. Again, like Heaven and Valley benchmarks.

 

[dj-electric]

I deleted a bad joke but when come to think of that...

This is the RX480 up there in the picture. I mean, who can tell me otherwise

 

[jigar2speed]

I have a feeling GTX 1080 FE owners wouldn't find this app interesting.

 

[the54thvoid]

I have a feeling GTX 1080 FE owners wouldn't find this app interesting.

Yeah it must suck for them. Even when it's throttling it's still about 10% faster than my 980ti at 1500Mhz.

 

[erixx]

I already had this since a week or so... ?!

 

[natr0n]

As if 3Dmark isn't already stressful test enough they have to go one further...

 

[jigar2speed]

Yeah it must suck for them. Even when it's throttling it's still about 10% faster than my 980ti at 1500Mhz.

It should actually, if I were spending $1000, which is the average price currently, i would be pissed because reviews are telling me completely different story about the performance.

 

[bug]

It should actually, if were spending $1000, which is the average price currently, i would be pissed because reviews are telling me completely different story.

You do realize we're talking a few tens of MHz on cards clocked at over 1GHz, don't you?

 

[jigar2speed]

You do realize we're talking a few tens of MHz on cards clocked at over 1GHz, don't you?

No we are not talking about few tens of MHZ on cards over 1 GHZ, we are talking about throttling issues of GTX 1080FE cards which are selling for $1000 on average currently and they cant maintain their boost speed of 1733 MHz on default fan speed i.e. after certain period of time.
When gamers OC their GTX FE cards, they are forced to keep the fan speed at around 80% to 100% in order to maintain the boost/oc clock speed, that is really very annoying to ears.

 

[PP Mguire]

No we are not talking about few tens of MHZ on cards over 1 GHZ, we are talking about throttling issues of GTX 1080FE cards which are selling for $1000 on average currently and they cant maintain their boost speed of 1733 MHz on default fan speed i.e. after certain period of time.
When gamers OC their GTX FE cards, they are forced to keep the fan speed at around 80% to 100% in order to maintain the boost/oc clock speed, that is really very annoying to ears.

As long as the card maintains its base clock or higher then it's doing its job. Boost isn't guaranteed, 1733 is just the base boost for reference cards but on default fan speed if it stays at or above 1607 base clock then there's nothing to argue over. Anybody who pays $1000 on a midrange card (whether it's the top end now or not) is IMO pretty daft.

As to the fan comment, that's the norm with reference blower coolers. If you buy a reference card with the intention of OCing then IMO you have nothing to complain about. I kept my Titans at 85% with an OC 24/7 and didn't complain. Sure it was noisy, but I already expected it. That's why I went water 6 months later.

Sheesh, I couldn't imagine spending 1000 on a 1080.

 

[bug]

No we are not talking about few tens of MHZ on cards over 1 GHZ, we are talking about throttling issues of GTX 1080FE cards which are selling for $1000 on average currently and they cant maintain their boost speed of 1733 MHz on default fan speed i.e. after certain period of time.
When gamers OC their GTX FE cards, they are forced to keep the fan speed at around 80% to 100% in order to maintain the boost/oc clock speed, that is really very annoying to ears.

SO in your head, boost clock is something that should be sustainable 24/7. I see...

 

[jigar2speed]

As long as the card maintains its base clock or higher then it's doing its job. Boost isn't guaranteed, 1733 is just the base boost for reference cards but on default fan speed if it stays at or above 1607 base clock then there's nothing to argue over. Anybody who pays $1000 on a midrange card (whether it's the top end now or not) is IMO pretty daft.

As to the fan comment, that's the norm with reference blower coolers. If you buy a reference card with the intention of OCing then IMO you have nothing to complain about. I kept my Titans at 85% with an OC 24/7 and didn't complain. Sure it was noisy, but I already expected it. That's why I went water 6 months later.

Sheesh, I couldn't imagine spending 1000 on a 1080.

Agreed with you, but a lot gamers are purchasing FE cards because they were shown FE cards reaching 2.1GHZ OC via Nvidia's CEO himself and in reality we both know things are quiet different.

 

[PP Mguire]

Agreed with you, but a lot gamers are purchasing FE cards because they were shown FE cards reaching 2.1GHZ OC via Nvidia's CEO himself and in reality we both know things are quiet different.

That also showed a temp of IIRC 67c. It was probably running a hybrid cooler. Some reviews I read were successful in gaining a 2GHz boost; Guru3D achieved this with a 60% fan setting. Then again, kinda dumb to buy a card (especially reference) in the context of trying to attain what is more than likely a tweaked OC. I mean, in stock form a 1080 is 24% faster than a Titan X (4k) according to TPU. Anything past that I'd say is a 100% success regardless of how far the card can be pushed. As the owner of 3 Titan X cards that rustles my jimmies a bit but I look at it as a precursor to how great P100 will be when I get 2 of those. (Or shit, looking at Polaris maybe how good Vega will be in my machine idk at this point)

 

[Caring1]

Agreed with you, but a lot gamers are purchasing FE cards because they were shown FE cards reaching 2.1GHZ OC via Nvidia's CEO himself and in reality we both know things are quiet different.

In reality I have seen plenty running at 2.1 on Firestrike, they can't all be special review cards.

 

[Kursah]

Interesting to see them add, I think it should be unlimited time for the free verison or the "40 hour" limit rather. I'll stick to OCCT, which recently updated to 4.4.2, and is still free for unlmited testing as-long-as you're not joined to a domain network. Been using it for years now as my primary system stability test, OC or not. Haven't been let down yet... 3DMarks I could really take or leave anymore. Regardless, cool to see them add that feature to this release I suppose...just a shame they limit the free one to 10 minutes. An hour or two would've sufficed as an appropriate limitation IMHO.