Discussion Intel current and future Lakes & Rapids thread

[igor_kavinski]

I hope keller keeps him at arm's distance.

Dang! Now I have this image of Raja trying to lick Keller's face like a puppy while Keller, with arms stretched out and keeping Raja at bay, yelling "Down, boy! Down!"

 

[igor_kavinski]

In spite of delays, Raja still deserves a little bit of our appreciation I think.

I don't dispute that the A770 16GB is a good product, especially in raytracing. But Intel was already bleeding when they hired Raja and he dug his pointy nails all at once into Intel, making them bleed even more. I bet they wonder if the same job could have been done for a few billion $ less and that money could have been used for better CPU designs.

 

[SiliconFly]

I don't dispute that the A770 16GB is a good product, especially in raytracing. But Intel was already bleeding when they hired Raja and he dug his pointy nails all at once into Intel, making them bleed even more. I bet they wonder if the same job could have been done for a few billion $ less and that money could have been used for better CPU designs.

Well, creation & funding of AXG (and raja) was decided by the Intel board. He's just an employee hired for a very specific purpose. In spite of delays, I think he's delivered something thats gonna save Intel moving forward. Meteor tGPU is his creation as well. He was well into Battlemage when he left.

Soon we'll know whether Intel's AXG investment was a success or a failure depending of MTL's tGPU performance. If it performs well, it may very well take Intel to a whole new level! And if that happens, he'll definitely share some of the credits.

 

[SiliconFly]

I don't dispute that the A770 16GB is a good product, especially in raytracing. But Intel was already bleeding when they hired Raja and he dug his pointy nails all at once into Intel, making them bleed even more. I bet they wonder if the same job could have been done for a few billion $ less and that money could have been used for better CPU designs.

Yeah. Arc ray tracing performance was a bit of a surprise actually. Esp. when XeSS was better than AMD's FSR.

Guy is definitely talented. I think he just over-promised and delivered late.

 

[Exist50]

Yep. I said this before, but it looks like Intel is going "all in" on their foundry play and counting on not only regaining process leadership for their own products, but also gaining lots of foundry customers at the same time. If it works, Gelsinger will look like a genius but lots of things have to go nearly perfect for it to work at this point. There is not much room for error anymore at Intel.

The fabs are clearly what he's prioritizing, but I'm skeptical of the wisdom of sacrificing the design side, which is realistically both paying the bills today, as well as the generally more profitable industry. Like, would succeeding in foundry be worth scuttling their attempts to get into the AI chip market?

I dunno about cutting employees tbh. amd had and still has far less than intel does and they're kicking intel's ass six ways from sunday.

There is an argument to be made there, but AMD's been rapidly growing as they attempt to cover some of the existing markets Intel does. The acquisition of Xilinx, a broader client and datacenter SoC portfolio, etc. Also, Intel's behind. Catching up is usually harder than staying ahead. AMD had to take some pretty drastic cuts themselves to be able to invest in Zen, which continue to hurt them today, and they also got very lucky by intercepting the peak of Intel's failures with both design and fabs.

And perhaps more importantly, we don't have any real insight into how many people Intel has in design to begin with. Their fabs represent the lion's share of their total employee count, and they don't even publish that explicitly.

After the release of ChatGPT and competitors? Nope. Intel needs GPUs to stay competitive in the AI space.

A similar thing was said about phone SoCs, and we see how badly they bungled that. Wouldn't be the first time they lost out on a huge emerging market because of shortsightedness.

Yeah they got rid of Raja after all! He was like 25% of their GPU division all by himself!

Everything I've heard about Raja's time at Intel is that he was great at making pretty powerpoints and finding dumb ways to waste incredible amounts of money*, but terrible at actually getting anything done.

*Sometimes at the same time. Heard the claim that he outsourced his powerpoints to some fancy design firm.

 

[Thunder 57]

The fabs are clearly what he's prioritizing, but I'm skeptical of the wisdom of sacrificing the design side, which is realistically both paying the bills today, as well as the generally more profitable industry. Like, would succeeding in foundry be worth scuttling their attempts to get into the AI chip market?

There is an argument to be made there, but AMD's been rapidly growing as they attempt to cover some of the existing markets Intel does. The acquisition of Xilinx, a broader client and datacenter SoC portfolio, etc. Also, Intel's behind. Catching up is usually harder than staying ahead. AMD had to take some pretty drastic cuts themselves to be able to invest in Zen, which continue to hurt them today, and they also got very lucky by intercepting the peak of Intel's failures with both design and fabs.

And perhaps more importantly, we don't have any real insight into how many people Intel has in design to begin with. Their fabs represent the lion's share of their total employee count, and they don't even publish that explicitly.

A similar thing was said about phone SoCs, and we see how badly they bungled that. Wouldn't be the first time they lost out on a huge emerging market because of shortsightedness.

Everything I've heard about Raja's time at Intel is that he was great at making pretty powerpoints and finding dumb ways to waste incredible amounts of money*, but terrible at actually getting anything done.

*Sometimes at the same time. Heard the claim that he outsourced his powerpoints to some fancy design firm.

Raja has the uncanny abailty to fail upwards.

 

[A///]

There is an argument to be made there, but AMD's been rapidly growing as they attempt to cover some of the existing markets Intel does. The acquisition of Xilinx, a broader client and datacenter SoC portfolio, etc. Also, Intel's behind. Catching up is usually harder than staying ahead. AMD had to take some pretty drastic cuts themselves to be able to invest in Zen, which continue to hurt them today, and they also got very lucky by intercepting the peak of Intel's failures with both design and fabs.

main issue I have with amd is their polishing isn't there like there is with intel even though I wish xtu wasn't so ass ugly.

 

[A///]

Raja has the uncanny abailty to fail upwards.

I'd like to think when he was coked out of his mind years ago at that launch even drink in hand smoking a big cigar and dancing with some random woman he had some recognition of the good life apart from likely sleeping with that poor woman. guy must be one damn smooth talker to keep getting his foot int the door at better and better places.

 

[Timorous]

AMD had to take some pretty drastic cuts themselves to be able to invest in Zen, which continue to hurt them today, and they also got very lucky by intercepting the peak of Intel's failures with both design and fabs.

Agreed with this 100%. It took the combination of AMD delivering something and executing well for 6 years combined with Intel failing to execute for nearing on a decade for AMD to be in the position they are in. Had AMD failed or Intel managed to actually execute sometime in the last 5 years even the current picture would be very different. Even Sapphire Rapids launching around the same time as Milan would have made a massive difference.

 

[BorisTheBlade82]

Even Sapphire Rapids launching around the same time as Milan would have made a massive difference.

Yep, but not by as much as I would have expected in 2020. As it looks now, SPR even loses against Milan by most (relevant) measures.

 

[eek2121]

Sounds exciting! But any pointers and/or references to 5.3 - 5.3 ghz numbers? Cos' most of the articles I've come across; tend to get their numbers from rumors or guesstimates. 🙁

Yes. 😉

 

[igor_kavinski]

Everything I've heard about Raja's time at Intel is that he was great at making pretty powerpoints and finding dumb ways to waste incredible amounts of money*, but terrible at actually getting anything done.

*Sometimes at the same time. Heard the claim that he outsourced his powerpoints to some fancy design firm.

Who is the real hero behind Intel ARC? That person deserves recognition.

 

[igor_kavinski]

guy must be one damn smooth talker to keep getting his foot int the door at better and better places.

He's a number 4 (born on 31 Aug).

3+1=4

These people often get fame and fortune. Jack Nicholson is 4 too (April 22. 2+2=4)

 

[Exist50]

Who is the real hero behind Intel ARC? That person deserves recognition.

They've had hundreds if not thousands of people working on it, just as with any similar project. I don't think it's fair to look for one hero or villain in what's really a team effort.

Though I'm not very impressed with DG2. Sure, they got something out the door, but looking at the amount of silicon, it should really be competing with the 3070 instead of 3060 (if not higher), and of course it was very late to market. They need to close the gap a lot if they're to have any chance.

 

[mikk]

DG2 is very important for Battlemage because they can polish the drivers+software from this and the learning curve from DG2 should benefit Battlemage, game devs can add XeSS and so on. I'm curious how MTL tGPU will perform. They don't have to use the GDDR6 memory controller which is a bottleneck according to Intel and who knows what they have changed.

 

[A///]

He's a number 4 (born on 31 Aug).

3+1=4

These people often get fame and fortune. Jack Nicholson is 4 too (April 22. 2+2=4)

The last time someone worked numerology with me was in 96 at a gas station outside Vegas and she was wearing very little in the way of clothing.

 

[igor_kavinski]

The last time someone worked numerology with me was in 96 at a gas station outside Vegas and she was wearing very little in the way of clothing.

96 is the year of the Rat. It's for new beginnings. I'm guessing you ignored her but if you had accepted her offer, your life could have turned out different 😛

 

[LightningZ71]

Actually I think raja doesn't deserve all the bashing that I've seen on the web. He's done some good too!

If we turn the clock back, Intel had to completely scrap the Larabee project after pouring in billions because it under performed. Well, in reality, it was a total disaster!

Their iGPUs have been stagnant for more than a decade with tiny incremental updates that aren't even worth a mention.

Raja started with a clean slate, cut out all the old crap, delivered decent hardware (not top notch but still good) and most importantly intel now has super stable and clean graphics drivers because of his past efforts. That itself is a humongous task. He did take his own sweet time, but without his knowledge and experience, Intel would be having another Larabee in their hands.

Instead, now they're in a position to compete with AMD/Nvidia/Apple in the lower to mid-range graphics which by itself is an massive achievement. In spite of delays, Raja still deserves a little bit of our appreciation I think.

On Intel's iGPUs, Ice Lake is a good example of where Intel was headed in the pre-Raja days. With minor tweaks, it is essentially just the GT 630 graphics product incrrased in EU count and clocked higher than previously to take advantage of the additional bandwidth that DDR4-3200 and higher LPDDRX products made available. While it certainly was a notable improvement across the whole stack over coffee lake, it didn't really advance the ball a whole lot. In fact, it was AMD's decision to scale back the Raven Ridge iGPU design from 11 CUs to 8CUs in Raphael/Lucienne that kept things as close as they were.

 

[Khato]

On Intel's iGPUs, Ice Lake is a good example of where Intel was headed in the pre-Raja days. With minor tweaks, it is essentially just the GT 630 graphics product incrrased in EU count and clocked higher than previously to take advantage of the additional bandwidth that DDR4-3200 and higher LPDDRX products made available. While it certainly was a notable improvement across the whole stack over coffee lake, it didn't really advance the ball a whole lot. In fact, it was AMD's decision to scale back the Raven Ridge iGPU design from 11 CUs to 8CUs in Raphael/Lucienne that kept things as close as they were.

Whereas I'd say that Tiger Lake is a good example of where Intel was headed in the pre-Raja days. Ya know, the IGP currently used by all of Intel's consumer processors. Just using 3Dmark time spy numbers from notebookcheck for a rough gauge, Intel went from 389 on Skylake/Kabylake/Cometlake to 744 on Icelake to 1479 on Tigerlake. That's a pretty good improvement thanks to the poor starting point.

Raja definitely gets to take credit for the failed product of ATS and corresponding delays of DG2.

 

[LightningZ71]

I was under the impression that Xe/tiger lake's iGPU was heavily influenced by both Raja AND as a response to Raven Ridge. The design is decidedly more modern than 630, the size is the reaction.

 

[Exist50]

Envisioning a Simplified Intel® Architecture

Intel investigated architectural enhancements and modifications for a 64-bit mode-only architecture.

www.intel.com

Intel just published a proposal for a version of X86 (X86-S) with some of the legacy instructions and other outdated functionality removed. In particular, no 32b OS support, nor native 16b support. This proposal does still support 32b apps, however, so probably not a big deal for most people.

 

[Thunder 57]

Envisioning a Simplified Intel® Architecture

Intel investigated architectural enhancements and modifications for a 64-bit mode-only architecture.

www.intel.com

Intel just published a proposal for a version of X86 (X86-S) with some of the legacy instructions and other outdated functionality removed. In particular, no 32b OS support, nor native 16b support. This proposal does still support 32b apps, however, so probably not a big deal for most people.

While I get what they are doing that page is stupid. Intel 64? Or EMT64? Or is Intel 64 Itanium? None. It is AMD64 or x86-64.

I get it. They are not going to say "Oops, we got caught with our pants down!". But to make it sound like "Intel 64" was their creation? Ugh.

 

[aigomorla]

While I get what they are doing that page is stupid. Intel 64? Or EMT64? Or is Intel 64 Itanium? None. It is AMD64 or x86-64.

I get it. They are not going to say "Oops, we got caught with our pants down!". But to make it sound like "Intel 64" was their creation? Ugh.

Raja is hoping you ignore all that, and ask him again why his name is Raja.
OH wait... he's ditched Intel even.
I guess no one cared why his name is Raja.

Raja Koduri, a leader in Intel's graphics division, is leaving the company to work on AI

Koduri joined Intel to help the company move into graphics, among other projects, and now he's leaving to create a company working on generative AI for gaming and more.

www.pcgamer.com

No but seriously, im really wonder who the hell is at upper management in intel.
They never make deadlines, they have aweful target timing, and they seem to think what is best for apple users apply to intel.

Someone tell them apple left the ship, and most intel users do not like MAC's.

 

[A///]

Raja is hoping you ignore all that, and ask him again why his name is Raja.
OH wait... he's ditched Intel even.
I guess no one cared why his name is Raja.

Raja Koduri, a leader in Intel's graphics division, is leaving the company to work on AI

Koduri joined Intel to help the company move into graphics, among other projects, and now he's leaving to create a company working on generative AI for gaming and more.

www.pcgamer.com

No but seriously, im really wonder who the hell is at upper management in intel.
They never make deadlines, they have aweful target timing, and they seem to think what is best for apple users apply to intel.

Someone tell them apple left the ship, and most intel users do not like MAC's.

oujia boards and magic 8 balls.

 

[coercitiv]

They are not going to say "Oops, we got caught with our pants down!". But to make it sound like "Intel 64" was their creation? Ugh.

To be fair, it sounds like everyone names this instruction set the way they want:

Various names are used for the instruction set. Prior to the launch, x86-64 and x86_64 were used, while upon the release AMD named it AMD64.[1] Intel initially used the names IA-32e and EM64T before finally settling on "Intel 64" for its implementation. Some in the industry, including Apple,[2][3][4] use x86-64 and x86_64, while others, notably Sun Microsystems[5] (now Oracle Corporation) and Microsoft,[6] use x64. The BSD family of OSs and several Linux distributions[7][8] use AMD64, as does Microsoft Windows internally.[9][10]