The New 3DMark Benchmark - Testing Smartphones to Multi-GPU Gaming PCs

[5150Joker]

http://www.3dmark.com/3dm/22925

Something seems off. Oh well, I'm sure Nvidia and AMD will be devoting far too much development effort to get 3DMark working smoothly with their products.

That's lower than what I got with my laptop. It's not the benchmark, its your system, something is dragging your combined score down. Your GPU was otherwise 3k faster than mine. It's pretty amazing that a notebook is so close to a desktop 690 in performance.

 

[RussianSensation]

Christian Ney @ Ocaholics is all over this with an in depth review.

New 3DMark Tested - 25 GPU, CPU and Memory Scaling

Fire Strike

"The first graphics test focuses on geometry and illumination effects. Particles are painted with half resolution and dynamic particles are activated. In total there are 100 spot lights, which cast shadows and 140 more light points, which don't cast shadows. On average there are 3.9 million vertices and 500'000 input patches have to be invoked. Per frame 5.1 million triangles need to be rasterized. Furthermore 1.5 million compute shaders are invoked per frame to enable particle simulations and post processing effects. On average 80 million pixels have to processed.

The second Fire Strike test is mainly about particle and GPU simulations. Particles are rendered with full resolution and dynamic lighting effects are enabled. The GPU has to render two smoke fields, six spot lights are casting shadows and 65 more don't cast shadows. On average there are 2.6 million vertices and 240'000 input patches which need processing per frame. In other words there are 5.8 million triangles per frame and 8.1 million compute shaders need to be invoked per frame. In total 170 million pixel are processed."

Graphics scores

Performance
HD7970 GE CF = 15,389 (+21%)
GTX680 SLI = 12,743

Extreme
HD7970 GE CF = 6,892 (+17%)
GTX680 SLI = 5,885

Looks legit. This is essentially what we saw with games that use DirectCompute for DX11 effects such as Sleeping Dogs, Hitman Absolution, Dirt Showdown, Sniper Elite V2. If future games start to utilize compute shaders for more DX11 effects, GK110 and even more compute-focused Maxwell would make a lot of sense for NV. This also explains why leaning towards GCN parts as opposed to Kepler for PS4/Xbox 720 could have made more sense given their long-term life expectancy.

 

[tigersty1e]

nvidia will be able to increase their scores after they release new drivers that lower the image quality... i mean "optimize"

 

[moonbogg]

http://www.3dmark.com/3dm/32095

Not bad I guess.

 

[amenx]

Let the dick waving begin!

 

[amenx]

nvidia will be able to increase their scores after they release new drivers that lower the image quality... i mean "optimize"

Last instance of anyone doing that, whether ATI or Nvidia, was in 2003. They cannot do it again. The ever-vigilant tech press will uncover any attempt of that nature if it is made.

 

[AdamK47]

That's lower than what I got with my laptop. It's not the benchmark, its your system, something is dragging your combined score down. Your GPU was otherwise 3k faster than mine. It's pretty amazing that a notebook is so close to a desktop 690 in performance.

The problem is Nvidia's Quad SLI profile. If I go to 2-way I get this:

http://www.3dmark.com/3dm/41128

 

[ShadowOfMyself]

Meh only 1 real next gen test? 3dmark keeps getting worse...

I actually liked Cloud Gate the most out of those tests... Id like to see a "high quality" version of it, without all the blurring

At least this seems to reflect real world performance though... But I wish they would focus on making something that is visually impressive rather than just taxing computers for the sake of it

Oh here, lets make you render 45345 trillion polygons with 500 light sources etc, whats the point? Anyone can do that, it would be far more impressive to make a highly optimized benchmark that actually justifies the performance hit

 

[moonbogg]

The problem is Nvidia's Quad SLI profile. If I go to 2-way I get this:

http://www.3dmark.com/3dm/41128

Yours might be a little bigger, buy mine is straighter I bet!

 

[alexruiz]

Is anyone with catalyst 13.1 encountering the issue of "driver not FM approved"?

 

[aaksheytalwar]

Is anyone with catalyst 13.1 encountering the issue of "driver not FM approved"?

Mine crashes without a score

 

[psolord]

Um I am trying to buy the darn thing by using the "Loyal benchmarker bonus" option, but I don't see the required field to use my pro 3d mark 11 key anyway.

I tried putting it in the promo code, but nothing happened. I took all the steps prior to logging into my paypal account, but I didn't see any relevant field.

Any ideas?

 

[Termie]

Um I am trying to buy the darn thing by using the "Loyal benchmarker bonus" option, but I don't see the required field to use my pro 3d mark 11 key anyway.

I tried putting it in the promo code, but nothing happened. I took all the steps prior to logging into my paypal account, but I didn't see any relevant field.

Any ideas?

FYI, I believe you can get the same discount (25%?) buying it from Steam: http://store.steampowered.com/app/223850/

 

[bgt]

 

[psolord]

Yes I saw that in Futuremark's website, but it would come cheaper if the charge was in USD and paid in euros.

Oh well, small difference, not worth it. I guess it's time for my first Steam app, amongst a gazillion games. Steam it is. Thanks.

 

[psolord]

Lol it has achievements! xD

 

[bgt]

 

[sontin]

Fire Strike

"The first graphics test focuses on geometry and illumination effects. Particles are painted with half resolution and dynamic particles are activated. In total there are 100 spot lights, which cast shadows and 140 more light points, which don't cast shadows. On average there are 3.9 million vertices and 500'000 input patches have to be invoked. Per frame 5.1 million triangles need to be rasterized. Furthermore 1.5 million compute shaders are invoked per frame to enable particle simulations and post processing effects. On average 80 million pixels have to processed.

The second Fire Strike test is mainly about particle and GPU simulations. Particles are rendered with full resolution and dynamic lighting effects are enabled. The GPU has to render two smoke fields, six spot lights are casting shadows and 65 more don't cast shadows. On average there are 2.6 million vertices and 240'000 input patches which need processing per frame. In other words there are 5.8 million triangles per frame and 8.1 million compute shaders need to be invoked per frame. In total 170 million pixel are processed."

Graphics scores

Performance
HD7970 GE CF = 15,389 (+21%)
GTX680 SLI = 12,743

Extreme
HD7970 GE CF = 6,892 (+17%)
GTX680 SLI = 5,885

Looks legit. This is essentially what we saw with games that use DirectCompute for DX11 effects such as Sleeping Dogs, Hitman Absolution, Dirt Showdown, Sniper Elite V2. If future games start to utilize compute shaders for more DX11 effects, GK110 and even more compute-focused Maxwell would make a lot of sense for NV. This also explains why leaning towards GCN parts as opposed to Kepler for PS4/Xbox 720 could have made more sense given their long-term life expectancy.

Wow - you found out that a card with 32% more compute and 50% more bandwidth can be 17% faster.
I think that is a huge step for the mankind. :hmm:

 

[guskline]

Here is my FireStrike score for my 2500k Asrock rig @4.5 Ghza with a EVGA GTX670 FTW:
3DMark Score 5985
3DMarks Graphics Score 6756
Physics Score 8158
Combined Score 2654
Graphics Test 1 31.68988800048828 FPS
Graphics Test 2 27.38019561767578 FPS
Physics Test 25.900400161743164 FPS
Combined Test 12.344982147216797 FPS

Here is the Firestrike score for my FX 8350 @ 4.6Ghz with the exact same Graphics card:
3DMark Score 6004
3DMarks Graphics Score 6710
Physics Score 8978
Combined Score 2628
Graphics Test 1 31.381799697875977 FPS
Graphics Test 2 27.260066986083984 FPS
Physics Test 28.50474739074707 FPS
Combined Test 12.22535228729248 FPS

Here is the Firestrike score for my FX8150 @ 4.3Ghz with a Gigabyte 7870OC graphics card:
3DMark Score 5059
3DMarks Graphics Score 5705
Physics Score 7432
Combined Score 2174
Graphics Test 1 27.97727394104004 FPS
Graphics Test 2 22.28139877319336 FPS
Physics Test 23.594879150390625 FPS
Combined Test 10.115769386291504 FPS

Here is the Firestrike score for my Asus 2500k @ 4.5Ghz With a PNY GTX680:
3DMark Score 6111
3DMarksGraphics Score 6946
Physics Score 8258
Combined Score 2667
Graphics Test 1 32.742347717285156 FPS
Graphics Test 2 28.0255069732666 FPS
Physics Test 26.2176456451416 FPS
Combined Test 12.40527629852295 FPS

 

[james1701]

Is anyone with catalyst 13.1 encountering the issue of "driver not FM approved"?

It's in the FAQ. You need to install the 13.2 beta drivers.

 

[Mir96TA]

Where you all are D/L from ?
Cause my 3D mark is a regular 3D mark
I can't seems to get those fire strike etc

 

[Termie]

Where you all are D/L from ?
Cause my 3D mark is a regular 3D mark
I can't seems to get those fire strike etc

http://www.futuremark.com/support/downloads

 

[Termie]

Here is my FireStrike score for my 2500k Asrock rig @4.5 Ghza with a EVGA GTX670 FTW:
3DMark Score 5985
3DMarks Graphics Score 6756
Physics Score 8158
Combined Score 2654
Graphics Test 1 31.68988800048828 FPS
Graphics Test 2 27.38019561767578 FPS
Physics Test 25.900400161743164 FPS
Combined Test 12.344982147216797 FPS

Here is the Firestrike score for my FX 8350 @ 4.6Ghz with the exact same Graphics card:
3DMark Score 6004
3DMarks Graphics Score 6710
Physics Score 8978
Combined Score 2628
Graphics Test 1 31.381799697875977 FPS
Graphics Test 2 27.260066986083984 FPS
Physics Test 28.50474739074707 FPS
Combined Test 12.22535228729248 FPS

Here is the Firestrike score for my FX8150 @ 4.3Ghz with a Gigabyte 7870OC graphics card:
3DMark Score 5059
3DMarks Graphics Score 5705
Physics Score 7432
Combined Score 2174
Graphics Test 1 27.97727394104004 FPS
Graphics Test 2 22.28139877319336 FPS
Physics Test 23.594879150390625 FPS
Combined Test 10.115769386291504 FPS

Here is the Firestrike score for my Asus 2500k @ 4.5Ghz With a PNY GTX680:
3DMark Score 6111
3DMarksGraphics Score 6946
Physics Score 8258
Combined Score 2667
Graphics Test 1 32.742347717285156 FPS
Graphics Test 2 28.0255069732666 FPS
Physics Test 26.2176456451416 FPS
Combined Test 12.40527629852295 FPS

Wow, incredibly consistent results - you have three systems that score nearly the exact same across the board! Interesting to see the 8350 in fact beating the 2500k.

By the way, what's your HD7870 overclock? Notice that yours beat mine (below) by just 15 points in the graphics score - could we have the same overclock? Actually we both take one of the graphics tests - perhaps one responds more to core and the other to memory clocks. Not surprisingly all three GTX670FTW scores are pretty much identical, and very close to a stock 680. My guess is mine is slightly ahead due to PCIe 3.0, which adds 1-2%.

Here's my first run, before I test with a serious overclock. This is with a GTX670@1189 and a 3770k@4.0.

And here's my run with an i7-860@3.25 and an HD7870@1100/1350:

Just as an aside, the HD7870 never ceases to impress me. It cost half as much as the GTX670 and provides 80% of the performance. While lots of people ooh and ahh about the 7950/7970, I think the gamer sweet spot is still around the $200 mark, as it's been for a long time.

 

[guskline]

Thanks for the comments Termie. The 7870 is the Std Gigabyte 7870OC with a factory oc of 1100. That's it no other mods but it's the one that has the 3 fans.The EVGA GTX670 FTWs are identical cards. One in the FX 8350 @ 4.6Ghz and the other in the 2500k @ 4.5Ghz. I had some Christmas Newegg gift cards and snagged the 7870OC when it was $229. An incredible card. I feel the same way about the EVGA GTX670 FTW. No need to OC the GPUs.

BTW amazing how close our 7870s are despite different CPUs

 

[Termie]

Thanks for the comments Termie. The 7870 is the Std Gigabyte 7870OC with a factory oc of 1100. That's it no other mods but it's the one that has the 3 fans.The EVGA GTX670 FTWs are identical cards. One in the FX 8350 @ 4.6Ghz and the other in the 2500k @ 4.5Ghz. I had some Christmas Newegg gift cards and snagged the 7870OC when it was $229. An incredible card. I feel the same way about the EVGA GTX670 FTW. No need to OC the GPUs.

BTW amazing how close our 7870s are despite different CPUs

Cool, thanks for the info.

I find 3dMark really helpful, and have used it through the years to compare various systems. You can see from our incredibly close results that it really is a great test of graphics cards - makes it easy to diagnose GPU issues. Due to the standard resolution and settings, a given graphics card at a given clock will pretty much get the same score every time, within a small margin of error. It's really an excellent tool.