Intel Unveils Discrete GPU Prototype Development

[Vayra86]

Play fair dude, this is likely Raja's very first step towards something, ie intels present best igpu with hybrid shaders and an fpga added to make up for all thats missing if they don't fit cpu cores and more importantly the rest of its supporting circuitry related to specific purposes ie power control but, the fpga mainly acts as an interface between intels proprietary inter chip interface type and pciex since it's clear the igpu was designed without pciex in mind it makes sense to test what they can expect before scaling up the design, they won't actually make many of these even for themselves, it's a stepping stone chip clear as day.

The shop bought ones four or five years out imho and certainly won't be a dual chip solution maybe MCM though.
Because that FPGA could possibly be a game changer if used to pump the GPUs core performance intelligently and or to directly affect a new Api for mainstream use case acceleration, intel already do fpga in server so they know it's got legs.

You present some solid points there and I feel four to five years out is realistic. However - consider what other companies can and will do in five years. If anything, this relies mostly on Intel's ability to improve the GPU in a fundamental sense combined with Raja's expertise on the rest. That is quite a gamble, one I wouldn't put money on.

 

[medi01]

I need the power even if it costs me my soul!!!!

"Challenge accepted!" (c) Huang

 

[Apocalypsee]

Intel should start to develop it's own drivers for their IGPs before doing a dGPU, Windows OpenGL and Direct3D drivers are a joke.

I second this. Intel IGP driver is bad even though their hardware is capable, and compatibility with older games are terrible

 

[Casecutter]

Play fair dude, this is likely Raja's very first step towards something,

I've been reading... regrettable I didn't mean to denigrate. I just see Raja's name and folks thinking he's going to be a Jim Keller to Intel discrete Graphic accelerator for lowly mainstream consumer's. Reading here you and others are probably steering the thread in a proper direction, that this is more to deliver a package that geared toward AI, Deep Learning, Compute, Automotive and yes is probably looking to use different packaging, and interconnects (EMIB).

On the other side reading here I can see a logic and argument of this work is to perhaps also provide a offshoot for iGPU perpetuation for the consumer CPU against AMD APU's. The more I read the more I don't see this about a discrete Graphic gaming product for either desktops or laptops, as I don't think Intel wants in on maintaining gaming drivers.

Raja is tasked to lead and maintain clear focus and keep engineers on task. He's less about actual designing but more using his preeminence on building partnerships and the development of trends in such moving markets.

 

[mastrdrver]

That is so wrong, I don't even know from witch angle i could explain why. I guess none would be the best.

Technically I should have said will never make a good GPU as that should have been easily understood from the example i gave.

 

[Fluffmeister]

It's funny to see the shitty CPU companies try to play catchup.

 

[CKBRYANT]

There already are a lot more GPU makers than Nvidia and AMD, but none of them are for PC's.

ARM, Imagination Technologies, Qualcomm and Vivante all make GPU's, as well as technically S3/VIA...
I most likely missed some companies as well, but the problem is, none of them can keep up with Nvidia due to the amount of money they're throwing at their R&D of new graphics architectures.

Even Intel isn't going to be able to catch up any time soon. At best I'd expect something mid-level for the first 2-3 years, as it's expensive, it's resource intensive and time consuming to make GPU's.

I honestly expect Intel to go after the Edge Computing cases of the GPU market, like future of AI and future of data center. I don't foresee anything they showed off via their proof of concept or protype being anywhere as close to Larabee (consumer ready graphics focused on gaming and etc) I just don't foresee Intel going that route, it's too far entrenched with the Duopoly that exists between AMD and Nvidia for them to join the game unless they bought one or the other. Which won't happen. Intel would take >3-5 years to come up with a top tier solution and at a minimum 3 years to have a mid-tier GPU.

 

[TheGuruStud]

Larabee? is that you? back from the dead?

The wasted billions are weighing that coffin down pretty heavily...just like Itanic.

 

[las]

With Raja Koduri in the lead, this is doomed to fail.

 

[alchemist83]

There already are a lot more GPU makers than Nvidia and AMD, but none of them are for PC's.

ARM, Imagination Technologies, Qualcomm and Vivante all make GPU's, as well as technically S3/VIA...
I most likely missed some companies as well, but the problem is, none of them can keep up with Nvidia due to the amount of money they're throwing at their R&D of new graphics architectures.

Even Intel isn't going to be able to catch up any time soon. At best I'd expect something mid-level for the first 2-3 years, as it's expensive, it's resource intensive and time consuming to make GPU's.

Wow! Thanks for the 'varied' information about companies who aren't the big two (NVIDIA or AMD). The other ''GPU'' makers you state really aren't GPU contenders / makers. Their efforts are focused on SOC devices. Cos they make mainly SOC for mobile phones, tablets and the like. They used to make lowly onboard GPU's for laptops back in the day (S3/VIA and ARM still do to an extent). Back in the 00's S3/VIA had a few g-cards still (more in the 90's).

From what I can personally make out from 1 the Hype and 2 the pics above (notably the 1st one), is that Intel do indeed intend to make a g-card to compete with NVIDIA and AMD. The 1st pic shows a PCI-E interface via/ with the FPGA. Thats how yours and my g-cards work BTW. Also the title says ''Discrete GPU''. Our cards (seperate not intergrated ones, e.g GTX 1070 etc...) are Discrete GPU's.

AND! You reckon Intel cant compete with NVIDIA? Intel money wise have 9x times more Billion $ dollars than NVIDIA, have been around for twice as long - 50 years compared to 25 and have chip making research and capabilities coming out of every orifice that FAR surpasses NVIDIA or AMD. True the big 2 are further along R&D wise in this particular speciality. But Intel wanna make more dollar and have the dollar and previous experience in order to at least give it a shot.

So maybe Intel may will go up against the big 2 and maybe, finally we will have more choice! It might take a long while though and as CKBRYANT above has pointed out, Intel will most likely start with Datacenters and AI processing. You never know, perhaps their first GPU will be a Mining one! Lols, I bloody hope not...

 

[rtwjunkie]

Intel is a CPU company first.

i740 was a terrible GPU even for the day. It would be like calling a gtx 1030 a good midrange gpu.

Intel can not make a GPU and never will because it's a CPU company first.

This is akin to someone saying GE was an appliance company, “they’ll never be able to make jet engines.”

There is no reason any company should be pigeonholed. Proper application of resources and recruitment of Subject Matter Experts allows expansion to other fields, especially similar ones like GPU’s.

 

[mastrdrver]

This is akin to someone saying GE was an appliance company, “they’ll never be able to make jet engines.”

There is no reason any company should be pigeonholed. Proper application of resources and recruitment of Subject Matter Experts allows expansion to other fields, especially similar ones like GPU’s.

You're analogy is flawed. If Intel was a silicone company, then you'd be right.

But they're a CPU company and anything else will be a secondary business and be treated as one.

Otherwise they'd still be making discrete GPUs to this day.

 

[Prima.Vera]

Isn't the main reason we don't have a 3rd competitor the amount of proprietary stuff currently found in modern GPUs that requires a lot of callous licensing??

 

[bug]

You're analogy is flawed. If Intel was a silicone company, then you'd be right.

But they're a CPU company and anything else will be a secondary business and be treated as one.

Otherwise they'd still be making discrete GPUs to this day.

As evidenced here: https://www.statista.com/statistics/495928/net-revenue-of-intel-by-segment/
they don't have CPUs as a standalone chapter in their revenue stream. Whether it's "PC client group", "client computing group" and "data center group" or whatever, they're looking at these as more or less complete platforms.
Whether they've build cheap, weak, embeddable solutions by themselves or licensed them from a third party, Intel has been dealing with graphics chips for decades. Arguing now that because they've traditionally used those as IGPs they cannot possibly build a decent discrete graphic card is a bit far fetched, imho.

 

[mastrdrver]

...........Arguing now that because they've traditionally used those as IGPs they cannot possibly build a decent discrete graphic card is a bit far fetched, imho.

Yet history says otherwise.

 

[bug]

Yet history says otherwise.

What?