Benchmarks are not magic applications. They use predefined runs of certain sections of an application (in this case let's say 'games'). They spit out a cold uninteresting number that is suppose to 'define' the performance of the hardware being tested.

The problem with that is a benchmark doesn't test for everything there is to test. So the score you get doesn't really tells you how fast the hardware will be in the real world, it just gives you a place for that hardware on the results table.

3dMark doesn't tell me how fast my video card is, it just tells me how fast it runs 3dMark. There are a number of cards that will give you a very high 3dMark score and not perform that good in certain games (double performance in 3dMark and just 5-40% performance increase in games).

It it ONLY was an optimization for a certain benchmark (detected or not via the exe name) then yes, it was cheating. But since the same optimizations can be applied to games and other applications (detected or not via the exe name) it is not a cheat.

If the performance in real world applications/games is being improved via optimizations then the score in benchmarks must see an improvement too.

Benchmarks supposed to offer an 'image' of the real world performance. If the performance increases only in real world applications/games and not in benchmarks, then would not this be a bad thing for the company that produces the hardware (shouldn't they feel cheated for doing all this optimization work and not getting any recognition from the benchmark-savvy community)?

Time will tell how many other applications/games will have optimizations applied.

In theory, I agree, and it can be a difficult line to draw.

I suppose the key difference is a generalised optimisation vs an application specific one.

If it's a game, application specific optimisations are fine as long as it makes thing faster.

For benchmarks, though, it's testing generalised performance and subverting that is breaking some of point of benchmarking.

It is a little of a grey area with Intel's drivers, as it's a performance boost offered to certain applications rather than specific benchmark boosting code.

Given the crappy performance will quickly be revealed in real games, I can't say I'm bothered about it. Checking CPU utilisation on DirectX/OpenGL applications (not games) on an Intel IGP vs Nvidia/ATI might be interesting though..

So I guess that when the new Catalyst or whatever the nVidia driver is called claims 40% performance improvement for 1 game, 20% for another and 5% for 2 others and NOT for all the games in the world, it is a cheat, since it is targeting a specific application/game.

It's free perfromance improvement. You don't like free stuff if it doesn't come from you beloved company. Feeling some pressure? Fanboy-ism-1-0-1.

This isn't quite the same as the Nvidia and ATI cheating where they either included specific code for 3DMark or deliberately reduced output quality.

I'd expand on what Olle P said - this is not a cheat for 3DMark, it's an optimisation for a certain set of programs.

Basically some of Intel's recent graphics chipsets are severely limited in terms of shaders and suchlike.

If the program being used fits entirely within the hardware resources provided by the chipset, performance is quite good. Once it extends beyond the resources provided (many modern games do), performance is dire - hence there's a software shader emulation (possibly other enhancements too) for a limited selection of applications.

These applications are all listed in the registry and can be modified and added to if you wish, at least on the X3100.

Of course, it's important to remember that the 3100 is definitely not a gaming chip, and even the later 4500 chipsets aren't that hot either.

What I've read is that Intel straight out admit that the driver will divert some tasks to the CPU when running 3DMark Vantage. Their defence is that it's not *only* that program that get this treatment, but also a range of different games as well. Therefore they don't consider it to be a cheat.

I see nothing wrong with pushing some graphics tasks to be solved by a low priority CPU process, but it should be implemented as a general rule instead of limited to a few pre-determined applications.

"We reckon that between Intel and AMD, whichever one can figure out how to detect in a general way that the GPU is a bottleneck at the same time that the CPU is underworked, and vice-versa, and divert some of either workload to the currently underutilised resource will steal an important march on its competitor."

Yas, I believe that's called AMD Fusion.

Typical Intel smoke, lies and mirrors. Intel jiggered drivers improved nothing but the benchmark scores.
Read: From Tech Report -

"Here's the tricky part: the very same [AMD] 785G system managed 30 frames per second in Crysis: Warhead, which is twice the frame rate of the [Intel]G41 with all its vertex offloading mojo [read: Intel rigged] in action. The G41's new-found dominance in 3DMark doesn't translate to superior gaming performance, even in this game targeted by the same optimization."

Typical Intel buggery.

Only a fool would trust Intel or Microsucks. Naive, Gullible Fools have made these two criminal companies rich.

This is the another whine from them.

So sad, mow you can label Intel as untrustworthy, hypocrite and out right liar but that nothing new.

That Intel is lagging 8 years in driver development.

Hahahaaha laughabuzz... bzzt pop.

Ah Ha. So that's probably why the Core 2 & i7 benchmarks have been so much better than AMD's too. Intel simply targets the benchmark software to get high scores !

think again, larrabee team is composed of many losers from the original driver team that couldn't hack it in a product world for various reasons....so, dont hold your breath for product worthy larrabee software... in fact any production worthy software at intel is a non starter

That'll be why AMD is playing catch up for the past couple of years, will it?

Oh, I get it, Intel has such skill at obfuscation that the majority of the market cannot tell that core2 and i7 are shyte?

Of course. And there was a second (and third!) gunman behind the grassy knoll.

SOS, DD from InHell. The scum of the earth.

Could someone with an intel IGP whip up a program with the same name, or rename another program that doesn't use a lot of system resources, and run it just to get the graphical 'boost'?

When life (or your IT department) gives you lemons...

Just saying you have an Intel IGP is enough for anyone to guess how horrible it is.

Thankfully Larrabee software team isn't the same as the Intel IGP drivers team.

this behaviour is consistent with everything intel does - DISHONEST

I expected intel to cheat better than this, as they have in the past.

Unfortuantely it did not cheat well enough as to not expose its true colors here.

Disappointing.

Seems intel didn't learn from the mistakes from ATI and nvidia.
So they will probably make all the same mistakes.

Blimey , not seen this going on for a number of years since old_ati and nvidia kept blaming each other about 'cheating' this way..

Intel cheating by modifying its driver behaviour to do better at a benchmark so that the score improves from abyssmal to just pathetic ?

Why do they bother ? Surely their resources would be better spent on actually producing graphics chips and chipsets that have performance even vaguely in the same league as their competition.