AMD vs. Intel and Nvidia - The Next-Gen GPU War is ON!

[Wayde Robson]

The first salvo in the next-gen GPU war has been fired by Nvidia and one thing is for sure, AMD will be hard pressed to outperform Amepere when Big Navi finally arrives. But the big item that's come out of Nvidia's big launch is the RTX 3070, priced at under half of a GTX 2080Ti. I feel bad for computer shops stuck with too much inventory of last-gen Nvidia cards. A third-tier new card outperforming the previous generation's flagship really shows how significantly the current generation of GPUs have jumped.

But regardless of how Big Navi performs against Ampere, AMD's fingerprints can be found all over Nvidia's pricing for this gen's launch. Launching a $500 next-gen card that renders its previous gen's flagship obsolete is sets the stage for the "gaming war" ahead, not just with AMD's Big Navi, but also the next-gen console launch. I think Nvidia is trying to poach some console gamers over to the PC-side.

Here's a rundown of how AMD managed to make a huge comeback and become a gravitational force in technology, again. I am in no way a Team Red fanboy, but the company has my respect and maybe even some gratitude for injecting competitive pricing back into PC-building, again.

AMD Fuels Nightmares of Intel & Nvidia

 

[BoredSysAdmin]

AMD Strategy with GPU since 2013 has always been - have console makers pay for its development, then use the same R&D to bring to the discrete GPU market. This way AMD gets a free GPU R&D - hard to beat that.
That said Nvidia's Ampere's early/marketing numbers seem like a very strong entry offering big performance increase numbers and price cuts. I cheer for AMD (as in I own a bit of their stock), but Big Navi better bring it(tm).
Intel's Xe is, for now, a paper release, but since they (embracingly) require TSMC to make at least some parts of it, I am not sure how soon we'll see it since TSMC does not see Intel as a long-time customer.
Edit: Above may be not 100% accurate. I couldn't find a bulletproof source, but there are claims that Intel network chips made using 28nm tech are made by TSMC, still not the same thing as reserving a massive amount of time on highly desirable 7nm production lines for time-limited fashion.

 

[Irvrobinson]

Intel's Xe is, for now, a paper release, but since they (embracingly) require TSMC to make at least some parts of it, I am not sure how soon we'll see it since TSMC does not see Intel as a long-time customer.

A little advice, take everything you read in the semiconductor press with a grain of salt. Most of the writers have never worked in the semiconductor industry and have no expertise. Intel has used TSMC as a fab for all of its server networking chips for about 15 years. Intel is a long-term customer of TSMC. Intel reserves its fab capacity for CPUs and select other high-volume chips that usually associated with CPUs, like client and server chipsets.

Edit: I can think of one exception of lower volume chips being made in Intel fabs, and that would be the Altera FPGAs. Years ago Altera cut a deal for Intel to fab their chips, back when Intel fab process was state of the art, and later Intel acquired Altera. I'm not positive if Intel still fabs their own FPGAs, but I suspect they do.

 

[BoredSysAdmin]

@Irvrobinson So are you saying that semiconductor press is semiaccurate at best? /rimshot

 

[Irvrobinson]

@Irvrobinson So are you saying that semiconductor press is semiaccurate at best? /rimshot

Often full of crap. I remember one so-called "expert" who kept calling ARM a chip manufacturer. ARM hasn't and still doesn't sell chips, they sell what's called Intellectual Property in the chip industry, meaning they design blocks like CPUs and other types for inclusion in other company's chips. I understand this semi-accurate guy now works in marketing at Intel. Perhaps he'll learn something there.

 

[BoredSysAdmin]

I understand this semi-accurate guy now works in marketing at Intel. Perhaps he'll learn something there

On a similar note = Take everything you read on Internet public forums with a grain of salt, Charlie Demerjian working for Intel, that would the day hell freezes over.

https://www.linkedin.com/in/charliedemerjian/

 

[AcuDefTechGuy]

Nvidia RTX-3080?

Sounds kind of like Yamaha RX-A3080.

Dammit, I just bought my RTX-2080S a few months ago.

 

[Wayde Robson]

Often full of crap. I remember one so-called "expert" who kept calling ARM a chip manufacturer. ARM hasn't and still doesn't sell chips, they sell what's called Intellectual Property in the chip industry, meaning they design blocks like CPUs and other types for inclusion in other company's chips. I understand this semi-accurate guy now works in marketing at Intel. Perhaps he'll learn something there.

That's interesting, I've only read online that ARM was a chip manufacturer, but I can see how that might be confused for merely selling its IP for others to make. A lot of what gets called "manufacturers" don't actually have one factory to manufacture anything, just order it from bidders in China... as far as I know. I know a couple of guys in Waterloo that were deep in chip technology at the University there and they ended up owning patents of their own and make a pretty good living from it, but they also work.

 

[Wayde Robson]

Often full of crap. I remember one so-called "expert" who kept calling ARM a chip manufacturer. ARM hasn't and still doesn't sell chips, they sell what's called Intellectual Property in the chip industry, meaning they design blocks like CPUs and other types for inclusion in other company's chips. I understand this semi-accurate guy now works in marketing at Intel. Perhaps he'll learn something there.

Oh, what do you think of the prospect of Nvidia buying ARM?

I honestly think Intel getting into GPU with Xe Graphics is a way it can bridge into the spinoff industries that like to use GPUs.

 

[BoredSysAdmin]

@Wayde Robson
Nvidia won't buy ARM itself, I guarantee it, but they more likely to buy of one of the current ARM licensees chip developers, like Ampere Computing, or buy Calvimum from Marvell (much less likely)

 

[Irvrobinson]

On a similar note = Take everything you read on Internet public forums with a grain of salt, Charlie Demerjian working for Intel, that would the day hell freezes over.

https://www.linkedin.com/in/charliedemerjian/

I wasn't referring to him.

 

[Irvrobinson]

@Wayde Robson
Nvidia won't buy ARM itself, I guarantee it, but they more likely to buy of one of the current ARM licensees chip developers, like Ampere Computing, or buy Calvimum from Marvell (much less likely)

Well, the scuttlebutt is that NVDA is interested in ARM, but I agree with you, it doesn't strike me as a good idea. I also agree that one of the ARM-based CPU designers *is* a good idea. I doubt Marvell is interested in selling Cavium, but I like Ampere and think that would be NVDA's best target. Ampere is rightly aiming at the big cloud companies, and their design is very good, IMO. Ampere is also full of a lot of former Intel engineers who know a lot about designing successful CPUs.

 

[BoredSysAdmin]

I wasn't referring to him.

I'm lost, which one did you mean?

 

[Irvrobinson]

Oh, what do you think of the prospect of Nvidia buying ARM?

I honestly think Intel getting into GPU with Xe Graphics is a way it can bridge into the spinoff industries that like to use GPUs.

As soon as I read about the NVDA interest in ARM I didn't see the synergy. I'm with @BoredSysAdmin, I think buying Ampere is a much better choice.

As for Intel, IMO the most important technology they're working on is oneAPI, which unifies the programming model for heterogeneous computing. If they build a decent GPU and they succeed with oneAPI they could successfully challenge NVDA and AMD. Cuda is not exactly programmer friendly unless you're a SIMD programming expert. Software enabling is an important factor in how Intel has succeeded beyond just being chosen for the IBM PC.

oneAPI: A New Era of Heterogeneous Computing

Remove proprietary code barriers with a single, open, standards-based programming model for heterogeneous computing—CPUs, GPUs, FPGAs, and other accelerators.

software.intel.com

 

[Irvrobinson]

I'm lost, which one did you mean?

In that particular case, Ashraf Essa. But there is a long list of semiconductor and computer system industry writers who don't seem to know what they're talking about.

 

[BoredSysAdmin]

In that particular case, Ashraf Essa. But there is a long list of semiconductor and computer system industry writers who don't seem to know what they're talking about.

Ok, so Ashraf Eassa of TheMotleyFool and SeekingAlpha (both less than highly reputable sites) is indeed joined Intel's marketing team. How does this relate to rumors of TSMC not giving Intel much love?
Are you referring to these articles?

Intel Corp. Set to Fall Further Behind TSMC

Things are going to get ugly for Intel's manufacturing group.

finance.yahoo.com

TSMC Is Set to Beat Intel at Its Own Game | The Motley Fool

Chipzilla's about to get lapped by its fiercest competitor.

www.fool.com

 

[Irvrobinson]

Ok, so Ashraf Eassa of TheMotleyFool and SeekingAlpha (both less than highly reputable sites) is indeed joined Intel's marketing team. How does this relate to rumors of TSMC not giving Intel much love?
Are you referring to these articles?

Intel Corp. Set to Fall Further Behind TSMC

Things are going to get ugly for Intel's manufacturing group.

finance.yahoo.com

TSMC Is Set to Beat Intel at Its Own Game | The Motley Fool

Chipzilla's about to get lapped by its fiercest competitor.

www.fool.com

No, I haven't read them. Since Intel isn't in the foundry business, TSMC and Intel are not competitors, and Intel has indeed been a TSMC customer for a long time. Intel has also been rumored to have done business with Samsung's foundry business in Austin, but I haven't been interested enough to keep track.

I admit to reading several semiconductor press web sites, but mostly I use them to identify what might be news, and I search around until I find a good source for accurate information. Sometimes there isn't a good source, so the BS is all that's available.

 

[jandrewyang]

That Motley Fool article referenced cherry-picks the timeframe to show AMD outperforming NVDA. Given the timeframe under discussion in the bulk of the article, NVDA and AMD are nearly neck-and-neck. The 5yr comparison just happens to include AMD at its nadir where it was down nearly 90% from its previous high. I've attached YTD, 1yr, 5 yr, and lifetime charts of the comparison. AMD has definitely made a great resurgence and are fantastically positioned in the x86 and GPU space but the 5yr compare is literally the only window within which AMD has dramatically outperformed NVDA stock.

Disclosure: I hold positions in both AMD and NVDA

 

[Irvrobinson]

Edit: I can think of one exception of lower volume chips being made in Intel fabs, and that would be the Altera FPGAs. Years ago Altera cut a deal for Intel to fab their chips, back when Intel fab process was state of the art, and later Intel acquired Altera. I'm not positive if Intel still fabs their own FPGAs, but I suspect they do.

I was correct. Intel's Architecture Day 2020 presentation shows the next generation Agilex FPGA (sampling now) is based on Intel's 10nm fab process.

 

[shadyJ]

Nvidia has posted such strong performance numbers for their Ampere platform that AMD has its work cut out for it trying to match or even getting close to that level of performance. I don't think Nvidia is sweating AMD's moves at the moment. I am rooting for AMD because strong competition makes for better product choices for us consumers.