Intel rep says people "probably won't" need discrete graphics in the future

Intel's already made some fairly bold promises at its Intel Developer Forum in Shanghai this week, and it now looks like it's getting into the prediction game as well, with one representative from the company telling TG Daily that people "probably won't" need discrete graphics cards in the future. That word comes from Intel Graphics and Gaming Technologist Ron Fosner, who was showing off a graphics demo running on a multi-core Nehelam system that, as you can see in the video at the link below, likely won't have NVIDIA or AMD rethinking their strategy just yet. Fosner also curiously looked to the past to back up his argument, saying that "if you look back into the mid 80's, there were no discreet graphics cards." Of course, all of this is all the more puzzling given that Intel is itself dabbling in discrete graphics with its Larrabee project, albeit under the guise of a CPU / GPU hybrid.

[Adam @ Apr 3rd 2008 10:59AM]

Well, if you look back to the mid-80s, there was no cell processing in video games. Therefore, I predict that we'll return to 8-bit gaming. Sega Master Systems for all!

[Zeus.:God @ Apr 3rd 2008 11:02AM]

Cell processing is mostly just a stop-gap solution for multiple cores.

[rv @ Apr 3rd 2008 11:51AM]

Way of. 4,000 years ago, we had no computers. Therefore, there will be no more computers by 2010.

[Jeebus @ Apr 3rd 2008 2:58PM]

Here's a clue to what Intel is talking about. They hired the guy who wrote the article.

http://www.pcper.com/article.php?aid=506&type=expert

While it won't run nicely even on Nehalem, in a couple of years, ray traced games are entirely doable. They'll look much nicer and won't require any GPU.

[Wes @ Apr 3rd 2008 11:02AM]

I could see it.... 8 core processor, these 4 do cpu, those 4 do gpu...

[Abuzar @ Apr 3rd 2008 1:19PM]

Run a 3Dmark06 CPU test. My Quad core only manages 3FPS. A SINGLE core GPU will easily manage 50 FPS on those settings.

[Wes @ Apr 3rd 2008 2:16PM]

O rly?

I'm not saying it works right now, what I'm saying is, in the year 3000, a single multi core chip may be able adequetly dedicate cores, or clusters of cores, to tasks that currently work best with a seperate chip.

[rektide @ Apr 3rd 2008 2:51PM]

@Abuzar: what "single core" GPU are you talking about? The TNT1? GPU's are all comprised of lots of cores. The geforce 3 series had ~24-64 specialized cores, the Geforce 8 series have 96-128 general purpose cores.

[konceptz @ Apr 3rd 2008 3:15PM]

Kind of like what we have now.

4 Core processor. 4 Core SLI.

[Greg @ Apr 3rd 2008 11:04AM]

Intel rep translator:
"People probably won't need discrete graphics"
TO
"No we didn't screw up by not purchasing a GPU maker"

[tom @ Apr 3rd 2008 11:33AM]

Have the rep define - his definition of "future"

[Greg @ Apr 3rd 2008 11:36AM]

"Future"
TO
"When this crazy vidya game fad finally passes"

[nh @ Apr 3rd 2008 3:07PM]

I dunno, I reckon it's more:

'people probably will need loud, proud and totally extravagant graphics in the future'

[Esat @ Apr 3rd 2008 11:04AM]

My laptop has a graphics card powered by my self esteem.

I can sometimes manage to see a pixel, but most of the time it turns out to be a speck of dust.

:(

[Ellianth @ Apr 3rd 2008 11:59AM]

Well God put out an Imagination Engine update like two weeks ago. Have you applied the patch? I had the same problem as you but Imagination Engine 3.6.1 is awesome. It only turns out to be dust 1/4th of the time. That's an improvement by anyone's standards!

[retro77 @ Apr 3rd 2008 2:30PM]

rofl...too funny

[kal326 @ Apr 3rd 2008 11:08AM]

Yeah I believe that one. Didn't somebody also say that nobody would ever need more then 640k of memory?

[Rex @ Apr 3rd 2008 11:37AM]

well, they guy did end up making more money till his retirement, than my entire clan will, so i guess you can forgive him

[Kurian @ Apr 3rd 2008 11:40AM]

I believe it was 1MB.

[Kurian @ Apr 3rd 2008 11:41AM]

And he was referring to the size of EXECUTABLE INSTRUCTIONS, not other data.
Even by todays standards, the executable instructions in a program are usually well below 1MB.

[shawnmos @ Apr 3rd 2008 11:49AM]

Somebody did, but it wasn't Bill Gates.

[Jay @ Apr 3rd 2008 12:03PM]

Well actually, I told my wife back in 1981 that the $149 16K Ram Pack for my ZX-81 would be all the memory I'd ever need...

[Alimas @ Apr 7th 2008 5:54AM]

Yes, it was a reference to conventional memory. Back when conventional memory (the first 640kb) was the only memory in a desktop.

[ssuk @ Apr 3rd 2008 12:59PM]

What your wife meant was; "You're not wasting any more money on that bloody computer."

[retro77 @ Apr 3rd 2008 11:09AM]

Lol @ 80's with no discrete graphics....pong ≠ crysis

[Ty @ Apr 3rd 2008 2:00PM]

Maybe homeboy only plays bejewelled.

[Rollins @ Apr 3rd 2008 11:15AM]

A CPU that handles GPU duties will probably enough for most people.

Most people not including gamers. Most people deal with integrated graphics solutions already.

[Chris @ Apr 3rd 2008 11:16AM]

I'm so glad Intel is out there, deciding what's best for me.

[raygundan @ Apr 3rd 2008 11:20AM]

This stuff goes in cycles, and Intel may have something here. GPUs have been big, parallel, highly specialized processors for a while. CPUs are suddenly getting very parallel with lots of extra cores that frequently are underused, and GPUs are suddenly becoming so general-purpose and can run plain old code. (See the GPU port of Seti@Home for a quick example.)

If GPUs start to lose the specialization that makes it worthwhile to dedicate a separate chip to graphics, and CPUs continue to gain extra cores that aren't doing anything... it only makes sense to start doing the graphics work on the CPU again.

Specialized hardware only stays specialized until either the CPU can do it with trivial effort, or it becomes a CPU on its own.

[rektide @ Apr 3rd 2008 1:11PM]

Your comments are the closest I've seen in this thread to anything even resembling an informed truth.

The main issue is that GPU instructions simply work on far larger blobs of data. Gimme the dot product on these two vectors takes 8 different values for input. Run this transform takes a 4x4 matrix. These are SIMD instructions, "single instruction, multiple data" and yes Intel has been tacking on more and more from its start at MMX all the way through to SSE4, but broadly speaking the CPU is not designed to pull in such large chunks of data.

Not only is the GPU better at doing math, but it has a HUGE number of simple CPU's on it. When you throw out the fancy things that make CPU's fast like out of order instructions and superscalar execution, the basic "do math" part of the CPU is very small (its called the ALU, arithmetic logic unit), and you can put a helluva lot of them onto silicon. The Geforce 8 & 9 series have 96 - 128 processors on them! ATI has _320_ processors. The comments on here that "well with 8 cores just use a couple for graphics" are ignorant of the fact that the GPU already has a large array of 2Ghz cores.

The CPU has to deal with far more complex tasks, and thats what makes it a very poor way-to-big GPU. For now, the refined specialization of a GPU is unparalleled. Just count: 320 cores v. 4 cores. an 80:1 parallelization ratio.

[Andir3.0 @ Apr 3rd 2008 2:33PM]

Well, I don't know if you're talking about something specific, but GPU processors are very small specialized stream processors and they each handle small chunks of data as opposed to the general processing of the CPU. Unless you are talking about bit bandwidth, but those are split between the available specialized stream processors. Your post was kind of vague. If you invert everything you said about data size vs core, then you pretty much got it. Those 320 Cores on the GPU are nowhere near as capable as a CPU in instructions. If that were the case, we'd all have 320 core desktops now. It's the basic difference between RISC and CISC essentially.

[rektide @ Apr 3rd 2008 3:10PM]

Everything in the GPU is optimized for one thing; SIMD, because thats 93% of what the GPU sees. The "small chunks" of data the GPU cells handle are still bigger than the 4 or 8 byte words the CPU's are optimized for. There are some SSE extensions that operate on comparatively larger sets of data, but even in these cases the CPU is hobbled by its fetch unit.

As I stated the reason the CPU is so much bigger is because it had a lot of added complexity to optimize performance when dealing with simple non-data-intensive operations and doing intensive flow control, which is why there is out of order execution and superscalar design. The 320 cores are no where near as good at flow control, but they're probably at least as good at doing cross products; they are pure data crunching machines.

Yes most post was vague, I try to keep stay high level and keep my word count under an 8 bit value when I'm not on Aces' talking to computer engineers.

[sully @ Apr 3rd 2008 11:21AM]

My c64 investment is SURE to pay out HUGE!!!!

[Kiwi616 @ Apr 3rd 2008 11:24AM]

I think they have it backwards...."we won't need CPU's anymore, because all processes will be handle by a GPU!"

Who are they kidding, games are getting more and more detailed and complex. Not to mention the CPU processors aren't keeping up at ALL right now. My MB runs crysis fine...NOT!

[DeadPixel @ Apr 3rd 2008 11:26AM]

thats a horrible prediction. they are just mad they didn't buy into after market graphics!

[Chebwa @ Apr 3rd 2008 11:28AM]

I have been toying around with Intel's 82945G chipset lately (which is by no means bleeding edge), and its 3D capabilities have really blown me away. If Intel really pushed themselves (which they apparently intend on doing), I see no reason why this claim couldn't be 100% true in just a couple of years. Discrete graphics have become absurdly complex lately anyway, it'd be nice to see various integrated options.

[digiboi @ Apr 3rd 2008 11:40AM]

Haven't we been here before?

Sorry - but the reason Intel didn't succeed in eliminating the graphics card before is because the industry keeps pushing the boundaries. 3D graphics in the OS, streaming video, multi-touch - its not just games that push the envelope.

[Chebwa @ Apr 3rd 2008 1:01PM]

Yeah, but we are hitting a wall soon. The average home computer is hideously overpowered for most simple tasks. It's coming to a point where software is going to be key, not hardware. This is only proven by the extreme success of "underpowered" computers like the Eee. Most people don't need even a Core 2 Duo, much less the 4 or even 8 or 16 core systems on the horizon.

Once that wall is hit, Intel will be able to shrink and shrink and shrink until a very capable fully integrated solution is perfect acceptable among casual gamers/editors, and discreet cards are left for the extremely, EXTREMELY high end users.

[billmao @ Apr 3rd 2008 11:31AM]

They're probably wrong.

[cduran01 @ Apr 3rd 2008 11:37AM]

Intel has a point here. Back in i386 days there were math co-processors, the i486 killed that. Now with multi-core processors, general CPUs are getting powerful enough to handle graphic acceleration.

This will be the ideal solution for performance boost, you eliminate one of the bottlenecks of discrete GPU systems......the bus between the CPU and the GPU. No matter how fast that bus is made it could never be as fast as a direct bus on silicon. In this case the only part that needs to be separate is the video memory, you wouldn't want to have memory access conflicts between video and non-video functions.

[cduran01 @ Apr 3rd 2008 11:37AM]

Intel has a point here. Back in i386 days there were math co-processors, the i486 killed that. Now with multi-core processors, general CPUs are getting powerful enough to handle graphic acceleration.

This will be the ideal solution for performance boost, you eliminate one of the bottlenecks of discrete GPU systems......the bus between the CPU and the GPU. No matter how fast that bus is made it could never be as fast as a direct bus on silicon. In this case the only part that needs to be separate is the video memory, you wouldn't want to have memory access conflicts between video and non-video functions.

[farfisa @ Apr 3rd 2008 11:42AM]

Actually the full quote was:
"[people] probably won't [need discrete graphics cards] when they're dead."
which makes perfect sense.

[owl @ Apr 3rd 2008 11:42AM]

I think he is probably right. Just like we won't ever need more than 640KB of RAM.

[d0n0van @ Apr 3rd 2008 11:44AM]

I watched the video. I observed a system running under 20 FPS with all eight cores at 75% utilization. No, thank you, I don't want it. I'll keep my graphics card and my delicious 60+ FPS gaming experience.


But to my surprise, we ran out of that address base for applications within—oh five or six years people were complaining.


SOURCE: http://americanhistory.si.edu/collections/comphist/gates.htm

[Kurian @ Apr 3rd 2008 11:45AM]

The problem will be stuffing the 2GB of graphics memory onboard. The way the Intel rep portrays it, they intend to replace the equivalent of a 9800GTX or so. It will probably have to sit on the motherboard and be far away from the C/GPU and will cause a performance hit.

And don't say you can share the main RAM. You will be laughed at.
2GHz 0.8nanosecond ram will be required.

[why not the LS2LS7? @ Apr 3rd 2008 12:13PM]

Xbox 360 shares the main RAM for the graphics card. And it seems to do quite well.

[bcollinstex @ Apr 3rd 2008 12:25PM]

360 is a purpose built machine with well defined parameters.

[rektide @ Apr 3rd 2008 1:29PM]

AMD's had onboard memory controllers, Intel is adding them soon, so latency is not an issue (your "far away"). You talk about the need for 2Ghz ram, but PC's already have 1.6GHz ram, so thats not outlandish. The difference is GPU's have a much wider pipe, for every cycle they send far more data. This is called the bus width. DDR3 PC1333 is 1.3GHz and provides 12.8GB/s, whereas an entry level graphics card is nearly 60GB/s of data, even though they only run 1.4GHz. data transfered = speed * bus width, and GPU is king (256 or 512 bit v. 64 bit).

The xbox360 has a shared ram but its significantly faster than most PC's, 22GB/s. It also has 10MB of special purpose extremely fast eDRAM that runs 256GB/s, which gives it a fast "working area," but rest assured most of the graphical content in your game get stored on that shared ram. Source: http://en.wikipedia.org/wiki/Xbox_360_hardware#Memory_and_system_bandwidth

[Kurian @ Apr 3rd 2008 1:58PM]

Oh and seeing that the 360 doesnt even come close to an ancient 7800GTX in terms of image quality and ingame graphics, shared = FAIL.

[why not the LS2LS7? @ Apr 3rd 2008 4:01PM]

Kurian:
When the Xbox 360 was being designed the 7800GTX wasn't out yet.

And a 7800GTX cost more than an Xbox 360.

Not taking into account the relative cost of solutions: fail.