Futuremark Preps DirectX 11 Exclusive Benchmark

[Unknown Soldier]

Futuremark’s Shift in Approach Towards 3DMark Benchmark Points to Popularity of DirectX 11

3DMark benchmarks from Futuremark have become very important tools for measuring performance of graphics processors and gaming PCs. For many, 3DMark score is the main factor when choosing hardware. Unfortunately, historically 3DMark did not use the latest technologies extensively. But the things are going to change with the next version of the benchmarking suite.

“Our 3DMark team is working on some seriously awesome DX11 stuff. It looks great, and that focus and investment in the latest graphics tech has benefits for the game studio too,” said Jukka Mäkinen, chief executive officer of Futuremark, in an interview with ComputerAndVideoGames web-site. Meanwhile, according to the BrightSideOfNews web-site, Futuremark has decided to make its next-generation 3DMark exclusive for DirectX 11 application programming interface (API), a move that clearly points to potentially very high popularity of the new API.

Traditionally, Futuremark did not implement too many new technologies into its 3DMark benchmarks. For example, 3DMark03 merely used DirectX 9.0, whereas 3DMark Vantage did not use a lot of DirectX 10 technologies. However, it looks like in case of DirectX 11 the approach will be changed and the new benchmark will only be aimed at DirectX 11 hardware.

“Over the last decade, graphics processors supporting each new version of DirectX have achieved high volume shipments earlier in their life cycles than their predecessors. Our forecast is that DirectX 11 capable GPUs will continue this trend with the fastest transition between technologies ever, resulting in shipments surpassing DirectX 10 GPUs in 2010,” said Dean McCarron, the principal of Mercury Research, late last year in a report.

By present ATI, graphics business unit of Advanced Micro Devices, has already transitioned its lineup of graphics processing units (GPUs) to DirectX 11 chips. However, Nvidia Corp. yet has to start shipping its own DX11 graphics processors. Unless both companies ship DirectX 11 GPUs, the transition will only get slower.

News Source: http://www.xbitlabs.com/news/video/...ark_Preps_DirectX_11_Exclusive_Benchmark.html

 

[Lonbjerg]

So another useless synthetic benchmark?

 

[Jawed]

Actually Futuremark's "benchmarks" are worse than useless, since they're misleading, with awful documentation that makes no attempt to educate and elucidate.

Jawed

 

[AlexV]

Actually Futuremark's "benchmarks" are worse than useless, since they're misleading, with awful documentation that makes no attempt to educate and elucidate.

Jawed

Their timing is intriguing.

 

[fellix]

I would rather expect more feature tests diversification in this round, along with the usual "game" benches for the e-peen contests. Things like fill-rate on wide variety of surface formats (from INT8 to FP32, Z-only), setup rate evaluation, GS throughput, general compute performance and so on. All the things that make a synthetic application actually useful (at least here, in B3D ).
And a new more friendly interface is welcome -- the Vantage messed GUI was just horrible.

 

[AlexV]

Things like fill-rate on wide variety of surface formats (from INT8 to FP32, Z-only), setup rate evaluation

I think we'll have these covered internally with our own toys sometime in the next 2 weeks

 

[trinibwoy]

Their timing is intriguing.

Why? Isn't it coming out sometime near the end of the year? Not everything has to be an Nvidia conspiracy you know

 

[AlexV]

Why? Isn't it coming out sometime near the end of the year? Not everything has to be an Nvidia conspiracy you know

Indeed...so maybe you shouldn't be seeing conspiracy suggestions everywhere? Had I wanted to say anything involving an IHV or another, i would have.

 

[Gunhead]

It's funny how it's always been useful, for a useless synthetic benchmark. I've found it a useful ballpark figure of a given product's performance, for 10 years. Silly score contests aside. There must be something in the wrong that they accidentally did right.

 

[Lonbjerg]

It's funny how it's always been useful, for a useless synthetic benchmark. I've found it a useful ballpark figure of a given product's performance, for 10 years. Silly score contests aside. There must be something in the wrong that they accidentally did right.

They appeal to e-peen teens, that think their score actully means anything related to gaming...not something I will call a "good" thing

That is also why more and more review sites have dropped 3Dmark...as it is a useless "number".

 

[Jawed]

I admit, Gunhead has a point, if you happen to have it lying around already and you think a driver update has borked your graphics performance, it is a sanity check.

Jawed

 

[PSU-failure]

Actually Futuremark's "benchmarks" are worse than useless, since they're misleading, with awful documentation that makes no attempt to educate and elucidate.

Well, I quite disagree with the awful documentation...

It's sufficiently clear on how the score is calculated, the only real issue is the tests themselve and perhaps the harmonic mean weights.

Overall, the idea of benchmarking the CPU is good but only if graphics tests are not CPU bound already, the same goes for physics which not only shouldn't be favoring one GPU brand but should be integrated in the graphics tests too, perhaps like what they did back in 2000-2001 with 2 "details levels" for each scene, but with 1 being physics free and the other with it.

 

[FrameBuffer]

They appeal to e-peen teens,..

I know sort of like people who brag about their 15/15 Mbit Gbond SHDSL lines and their i7(at 3.5Ghz), 12GB RAM, 2 x 21" CRT's, 7.1 soundsetups /wink

 

[AlexV]

It's sufficiently clear on how the score is calculated

That's nice, but it does little to nothing to detail their tests inner workings. And that's what's important, not how an arbitrary number was generated by applying arbitrary weights and computing one mean or another.

 

[PSU-failure]

That's nice, but it does little to nothing to detail their tests inner workings. And that's what's important, not how an arbitrary number was generated by applying arbitrary weights and computing one mean or another.

For the audience they target, it's enough.

Sure some synthetic, multiple rendering engines, heavily documented tests would be way more interesting, but the goal is to compare hardware solutions at a somewhat high level.

The 2000-2001 benchmarks had a better documentation and I hope they'll restart from there for the next iteration, even if it's not enough for in-depth performance analysis, it would be better than the last ones, especially if PhysX is left out.

 

[Lonbjerg]

I know sort of like people who brag about their 15/15 Mbit Gbond SHDSL lines and their i7(at 3.5Ghz), 12GB RAM, 2 x 21" CRT's, 7.1 soundsetups /wink

Well...it's taking it of context now isn't it?
The PC specs where in regards to powerconsumpition in the High-End PC gaming..and the linespeed in regard to that 56K really, really, really, really, really, really, really, really, really, really needs to die
(And that SHDSL with ~2ms to the gateway is a perfect perk form my work...compared to ADSL lantency)

Besides, you don't see that in my sig do you?

 

[trinibwoy]

Indeed...so maybe you shouldn't be seeing conspiracy suggestions everywhere? Had I wanted to say anything involving an IHV or another, i would have.

So what's so intriguing about the timing?

 

[Rangers]

I think Futuremark messed up when they went to 3 scores (Performance/Extreme etc). Part of the draw was the simplicity, one number to give a broad representation graphics performance. They've lost that this "generation".

I suppose they didn't want people with integrated laptops and such to feel bad about a low score...but when I want to play with 3Dmark on a integrated graphics laptop, I just get a old version.

 

[jimmyjames123]

Their timing is intriguing.

The success of and interest in Futuremark's DX11 benchmark is to some degree contingent upon some actual competition in the DX11 space. If FM came out with the DX11 benchmark six months ago, their benchmark would probably have been largely ignored by the community, since there would be no basis for comparison due to having only one IHV with DX11 GPU's

 

[itsmydamnation]

I know sort of like people who brag about their 15/15 Mbit Gbond SHDSL lines and their i7(at 3.5Ghz), 12GB RAM, 2 x 21" CRT's, 7.1 soundsetups /wink

3.5 is weak.... dont even need a voltage bump to get that,