Is Intel's entrance into the discreet GPU market a bad thing?

[GodisanAtheist]

The word on Intel's to-be-released GPUs is that it's likely they will compete in (at most) the mid-market and lower, so roughly the x60 level of performance (as an aside I appreciate that NV's naming scheme has become a sort of standard unit of performance for videocards). However, it appears AMD's best cards land at roughly this performance level as well.

I worry that Intel will really only end up competing with AMD at this performance level (and while it is a large part of the market, it can only be split so many times while remaining profitable to compete in). NV will likely maintain it's majority share and many of Intel's buyers wouod likely come from the same pool of people that currently consider AMD cards as an option.

As a result, we might end up with Intel losing interest in the small slice of the pie their investment returns, while AMD exits the discreet market to focus on custom designs due to the further reduction of market share. Nvidia is the only one that ends up winning, with their strong brand reinforced by top market products.

Instead of more competition, we end up with less.

Alternatively, Intel bumps off AMD's crippled RTG from the discreet market and thanks to deep coffers continues to push on, only to have one mediocre competitor replaced by another (with potentially terrible drivers, if history is a guide).

How do you think Intel's entrance into the market will disrupt the current order?

 

[crisium]

No.

So you think Intel will initially compete in sub $300 market which is where AMD competes, and just end up killing RTG? But what's the alternative, just AMD competing in sub $300? Might as well have 3. At least this way sub $300 customers have a good time.

They will scale up their cards eventually if they want to compete in the workstation.

I don't get how more competition is a bad thing, and even if RTG eventually folds then it's not less competition than before.

 

[Kenmitch]

Wait and see is what I'd suggest. It's too early to try and predict the outcome.

 

[ub4ty]

The word on Intel's to-be-released GPUs is that it's likely they will compete in (at most) the mid-market and lower, so roughly the x60 level of performance (as an aside I appreciate that NV's naming scheme has become a sort of standard unit of performance for videocards). However, it appears AMD's best cards land at roughly this performance level as well.

I worry that Intel will really only end up competing with AMD at this performance level (and while it is a large part of the market, it can only be split so many times while remaining profitable to compete in). NV will likely maintain it's majority share and many of Intel's buyers wouod likely come from the same pool of people that currently consider AMD cards as an option.

As a result, we might end up with Intel losing interest in the small slice of the pie their investment returns, while AMD exits the discreet market to focus on custom designs due to the further reduction of market share. Nvidia is the only one that ends up winning, with their strong brand reinforced by top market products.

Instead of more competition, we end up with less.

Alternatively, Intel bumps off AMD's crippled RTG from the discreet market and thanks to deep coffers continues to push on, only to have one mediocre competitor replaced by another (with potentially terrible drivers, if history is a guide).

How do you think Intel's entrance into the market will disrupt the current order?

No .. competition is always better and welcome.
I look forward to it. I'm tired of the duopoly

 

[VirtualLarry]

Instead of more competition, we end up with less.

I see what you're saying, longer-term. I know AMD is in a tough spot right now with GPUs, and their previous "head honcho" moved over to Bluer Pastures (at Intel), but I have faith that they will soldier on.

Don't forget, Sony and Microsoft are secretly priming AMD's R&D pipeline(s), if some accounts of history are to be believed. (That Polaris and Vega are really just dGPUs derived from console GPUs that Sony and Microsoft have paid for their designs.) So, don't count out AMD yet, as long as they have the Sony and Microsoft contracts for the consoles.

 

[zliqdedo]

You're asking a very complicated question. What you've described is one possible outcome, how likely I couldn't say, but it sounds logical. Another speculation could be that Intel actually delivers a very compelling product in all respects - hardware, software, and pricing - advertises it aggressively (possible given Intel's financial might), and ends up competing with NVIDIA as well, leaving enough of AMD's slice of the pie untouched to allow RTG to remain in business. Of course, you could also make the argument that a high-stakes 'war' between Intel and NVIDIA would result in great-value products for consumers, which would leave no room for a low-ball player like AMD. In all honesty, if RTG is to become a division unable of producing compelling, competitive products, it would exit the market sooner or later anyway, Intel or no Intel.

Which would be bad, but then again, if you think about it, NVIDIA is pretty much acting as if there isn't any competition even now. The last time I remember AMD making a difference is in the R9 290 era, which forced NVIDIA to immediately drop the price of the GTX 780 and consequently introduce the GTX 780 Ti - a card with a fully-enabled big GPU for $700, a class of card they might have been hoping to be able to sell for $1k+ from then on. This more or less carried over to the next generation - GTX 970 was relatively good value, and GTX 980 Ti - a card with a 601 mm² GPU - launched for $650. Of course, AMD had the Fury X back then. When NVIDIA got decisively back on top, things changed - the 10-series launch brought a $700 "flagship" card based on a mid-sized 314 mm² GPU. Sure, they later produced the GTX 1080 Ti for $700, much better, but at 471 mm² its GPU is not quite 'phat' or expensive to produce, in theory at least. As for the 20-series, well... We saw the arrogant SRPs, and the even worse realities of the market. Yes, the actual flagship will be available at launch for the first time since the GTX 580 - 8 years to be precise - and at 754 mm² it truly is big and juicy, but at $1000-1200, yeah...

Going by that logic, one way to shake up the market is to release a GPU powerful enough to compete with NVIDIA's best, which as you've said Intel is not planning on doing, at least not initially. If Intel executes well and with aggressive pricing in the lower segmet, however, the upper one might improve nonetheless - it'd be hard for NVIDIA to charge triple for a 50% improvement. And if this ends up driving AMD away, and we might be underestimated the RTG here, then we'd end up with a better competitor to NVIDIA since Intel would've had to fix their driver issues to become successful in the first place.

It's hard not to be excited about Intel's entry in the dGPU market. I don't think Intel and AMD would kill each other off since there's clearly space for at least two players in the market so it stands to reason that whatever happens at least one of them will survive, and who's to say there's no room for three players with absolute certainty? Even if Intel and AMD make each other's lives in the dGPU market hell, the minute one drops, the fortunes of the other improve exponentially. Not to mention, I think we might be underestimating both companies - why should AMD be a company that barely manages to compete, despite their current situation, and why should Intel be unable to produce good drivers? I don't think it's completely justified to compare Intel's GPU drivers to NVIDIA's and AMD's, since their GPUs aren't necessarily meant for playing games. If NVIDIA was making the same type of IGP for the same type of purpose as Intel, would their drivers be any better?

I also don't think NVIDIA's user base is set in stone, especially in the face of Intel. My understanding is that many of the current users buy NVIDIA video cards, not because they're thoroughly informed, but because they subjectively perceive NVIDIA as established, popular, better, faster, etc. Of course, many don't buy the cards that are actually faster than AMD's. So why wouldn't this crowd switch over to Intel - the other megabrand, potentially even more Rolex-y, from which they already have a microprocessor to begin with - Core i9 + Iris g9. Sick, bruh. Let's play some Minecraft on that.

 

[PeterScott]

I also don't think NVIDIA's user base is set in stone, especially in the face of Intel. My understanding is that many of the current users buy NVIDIA video cards, not because they're thoroughly informed, but because they subjectively perceive NVIDIA as established, popular, better, faster, etc. Of course, many don't buy the cards that are actually faster than AMD's. So why wouldn't this crowd switch over to Intel - the other megabrand, potentially even more Rolex-y, from which they already have a microprocessor to begin with - Core i9 + Iris g9. Sick, bruh. Let's play some Minecraft on that.

Not set in stone, but kind of set in thick packed mud. There is significant inertia to get people to switch. Even when AMD has comparable cards, people overwhelmingly buy NVidia.

I don't see why Intel would pull NVidia customers any more than AMD does, given how people deride Intel IGPs, and laud AMD APUs, and Intels reputation has taken bit of overall beating given the ongoing 10nm woes, and resurgent AMD CPUs.

Also, Intel seems to love margins even more than NVidia does, so it is hard to imagine them pushing perf/$ much.

If you asked me today how likely it is that I would buy each GPU brand in 2020, I would say 80% NVidia/18% AMD/2% Intel.

Luckily both Intel and AMD have good CPU businesses to subsidize the GPU business for a while, to create more options, and with two of them in the market, it is more chances of some alternative hitting the sweet spot to compete with NVidia.

For tech watchers, the GPU market is looking very interesting in the next few years. Intel entering the market, along with a big Ray Tracing push, and Tensor cores in consumer GPUs. Wild stuff.

I am actually most interested in getting Tensor cores in the consumer space. They could open many new capabilities, if they aren't driver locked to only first party access, which I can definitely see happening.

 

[GodisanAtheist]

No.

So you think Intel will initially compete in sub $300 market which is where AMD competes, and just end up killing RTG? But what's the alternative, just AMD competing in sub $300? Might as well have 3. At least this way sub $300 customers have a good time.

They will scale up their cards eventually if they want to compete in the workstation.

I don't get how more competition is a bad thing, and even if RTG eventually folds then it's not less competition than before.

- The 3d discrete GPU space has not supported 3 co-equal competitors through it's entire history. (I know it's a bit more complicated than the following, but I'm on mobile so bear with me) 3d FX was king, then Nvidia showed up and killed them, then ATI showed up and put up a good fight through the years, eventually as AMD, but now they're falling behind.

In a market that barely appears to support two players, what happens when there are three?

The scenario I'm suggesting, Intel hangs on just long enough to get RTG to exit the market, but subsequently exits the market itself. Intel is a company that is not comfortable being second, especially not by a large margin, and has time and again quickly pulled the plug on product lines that are not meeting expectations.

So instead of more completion, which I agree on a philosophical basis is always a good thing, the net result is we end up with less competition and a market that is so mature that the odds of a new competitor making their way in is slim to none. So no competition, with a monopoly for Nvidia, without even the token jabs provided by RTG at the moment.

Intel has a buttload of cash and has assembled a dream team of engineers and industry verterans to enter the market strong, but I'm certain that they will enter the market with more of a sloppy wet plop than a splash. There is just too steep a hill to climb to reach the summit NV has built.

Intel is likely planning to play rope-a-dope with NV by using the same borderline practices of bundling their cards with their CPUs to keep them relevant, buying time to build up the tech.

 

[VirtualLarry]

Intel is likely planning to play rope-a-dope with NV by using the same borderline practices of bundling their cards with their CPUs to keep them relevant, buying time to build up the tech.

Yes, I definitely see bundling in the future.

 

[Kenmitch]

Yes, I definitely see bundling in the future.

....And I see evil things!

AMD: Yo Intel what you wanna do about nvidia
Intel: JHH and that jacket gots to go
AMD: Yep....Lisa laughs every time she sees him in it. Calls him the squirminator!
Intel: You guys know gpus better than us so what ya guys think?
AMD: How about we come up with the idea and some of the IP while you guys foot the R&D bill?
Intel: You guys are putting the squeeze on us already with Zen products. I've heard the term pulling a rabbit out of your booty but thought it was a myth.
AMD: So what you think. We've got multiple solution already. Just don't have the manpower to get it done quickly.
Intel: Hmm....How long do you think it would take to develop and implement?
AMD: You doing the heavy hitting....Probably a couple months to develop and a generation to implement.
Intel: Sounds interesting....What's the projected outcome?
AMD: End game for nvidia on the PC. We could also watch them squirm and die a slow death if you prefer.
Intel: How? Is it legal? You know we really like to watch the other guy squirm.
AMD: I can't type that here silly guy. Of course it's legal....We don't play your games. We know all too well.
Intel: Give me a call at 1-800-END-GAME. You too funny dude!
AMD: Dang....You guys work fast when you got the flame under your booty!

Just saying....The future only knows what the future will bring is all.

 

[Elfear]

If you asked me today how likely it is that I would buy each GPU brand in 2020, I would say 80% NVidia/18% AMD/2% Intel.

Dang, you only give AMD an 18% chance of being in the GPU market 16 months from now? They definitely have some big hurdles to cross to remain competitive but I don't think they're that bad off.

 

[alcoholbob]

If Intel is willing to dump in 2 billion+ a year in R&D they can certainly compete with Nvidia, it's all a matter of how much they wanna invest in the AI, machine learning, and CGI ray tracing race.

 

[GodisanAtheist]

Dang, you only give AMD an 18% chance of being in the GPU market 16 months from now? They definitely have some big hurdles to cross to remain competitive but I don't think they're that bad off.

I think he's just saying that given what he knows now, there is an 80% chance he will buy an NV card in 2020. Similarly, there is an 18% chance he goes for AMD (assuming they offer an proper upgrade in performance for the right price) and a very slim 2% chance Intel will show up with anything compelling enough for him to purchase.

I'd wager something similar, but given I'm on a 980Ti now and have never had a problem switching between brands, if AMD shows up with even 1080ti performance for a moderate sum, they'll get a pity purchase out of me.

 

[Elfear]

I think he's just saying that given what he knows now, there is an 80% chance he will buy an NV card in 2020. Similarly, there is an 18% chance he goes for AMD (assuming they offer an proper upgrade in performance for the right price) and a very slim 2% chance Intel will show up with anything compelling enough for him to purchase.

I'd wager something similar, but given I'm on a 980Ti now and have never had a problem switching between brands, if AMD shows up with even 1080ti performance for a moderate sum, they'll get a pity purchase out of me.

Got it. Makes more sense now. I'd say my chances are about the same unfortunately.

 

[sandorski]

Don't be too quick to dismiss AMD. For the first time in a decade they finally have the Cash to spend on R&D and they have a lot of talent in regards to the GPU Market.

 

[GodisanAtheist]

Don't be too quick to dismiss AMD. For the first time in a decade they finally have the Cash to spend on R&D and they have a lot of talent in regards to the GPU Market.

- After getting absolutely ROASTED by the Polaris hype train, then getting smart and being only rug burned by the Vega hype train, I think I'll believe it when I see it

I know AMD has it in them, and their CEO is the first in a long time thats had their **** together, but they have a long hard road in front of them in both CPUs and GPU's. You can't just release one good product and all is well, you have to demonstrate that you can do it again and again year after year to rebuild that trust in the brand.

AMD's got their work cut out for them on both fronts, but their competition isn't sitting still (lucky for them Intel seems to be going backwards at the moment with their Specter mitigation). NV is still playing the game like a boss though...

 

[Qwertilot]

They have the cash for a decent R&D spend now but that won't show any effect as soon as 2020. Maybe 2023 or so?

 

[zliqdedo]

The scenario I'm suggesting, Intel hangs on just long enough to get RTG to exit the market, but subsequently exits the market itself. Intel is a company that is not comfortable being second, especially not by a large margin, and has time and again quickly pulled the plug on product lines that are not meeting expectations.

Okay but why would Intel exit in that case? Beating one of its competitors should open up enough market space for Intel to continue. If Intel is looking at the consumer dGPU market and thinking it can gain a 50%+ share overnight, beating Nvidia, then the people behind this project must be delusional, and I can't believe that. Even if first place is the end goal, forcing AMD out of the market would be an important step towards that, serving to embolden Intel. Yes, Intel has a tendency to canceling big projects but that must also be part of a learning process, it's hard to believe they would just throw massive cash at dGPUs just to see what happens, they must know that if they don't execute well, they would not be successful.

So instead of more completion, which I agree on a philosophical basis is always a good thing, the net result is we end up with less competition and a market that is so mature that the odds of a new competitor making their way in is slim to none. So no competition, with a monopoly for Nvidia, without even the token jabs provided by RTG at the moment.

Let's say that happens, and Nvidia stands alone. Judging by your other comments, you don't see AMD as a proper competitor to Nvidia - you said you were burned by both Polaris and Vega, and you don't seem to have much faith in AMD's future. Going by that logic Nvidia already stands alone, so how can Intel's entry into the market, or any sort of potential shake-up for that matter, be a bad thing? If AMD's jabs create sufficient competition I certainly don't see its effect now - it feels like the RTX cards as priced at the absolute maximum of consumer tolerance, forcing hands, an act of a complete monopolist.

Another thing that I noticed is that we're all discussing a scenario where Intel actually wants to compete in the general consumer and/or gaming market. Perhaps Intel wants to enter the non-general dGPU market which is a whole different story.

 

[TheELF]

Okay but why would Intel exit in that case? Beating one of its competitors should open up enough market space for Intel to continue. If Intel is looking at the consumer dGPU market and thinking it can gain a 50%+ share overnight, beating Nvidia, then the people behind this project must be delusional, and I can't believe that. Even if first place is the end goal, forcing AMD out of the market would be an important step towards that, serving to embolden Intel. Yes, Intel has a tendency to canceling big projects but that must also be part of a learning process, it's hard to believe they would just throw massive cash at dGPUs just to see what happens, they must know that if they don't execute well, they would not be successful.

AMDs GPU division wouldn't shut down,ever.
Somebody would buy it up and keep it running ,probably intel (not nvidia because monopoly) but also phone/mobile companies might love to get their hands on it.

 

[zliqdedo]

AMDs GPU division wouldn't shut down,ever.
Somebody would buy it up and keep it running ,probably intel (not nvidia because monopoly) but also phone/mobile companies might love to get their hands on it.

Quite possibly. As I've reasoned in previous comments, I think we might be underestimating AMD in the whole discussion, but it's still good to entertain possible outcomes.

 

[PeterScott]

AMDs GPU division wouldn't shut down,ever.
Somebody would buy it up and keep it running ,probably intel (not nvidia because monopoly) but also phone/mobile companies might love to get their hands on it.

Agreed. But even more than that, AMD wouldn't sell the GPU division either, unless AMD as a company was collapsing, and it was part of the liquidation sale. There was a story a while back that the reorganization to give more internal independence RTG, was a sign they were getting ready to split it off. That was total nonsense.

The reasons AMD bought ATI still exist today, and really the GPU has only has grown in importance, which is why Intel wants in.

The vast majority of devices are powered by chips that have both CPU and GPU elements (APUs/IGP/SoC). Having any kind of mainstream CPU business requires integrating GPUs. AMD absolutely needs RTG for that.

And even when you look at Gamers on Steam, the top 5 cards are all lower-mid NVidia 50 and 60 series cards. So to sell to most gamers buying discrete cards, you don't need to be competitive even in the space Vega sold. You just need good low-mid cards.

Bottom line AMD is never leaving the GPU business unless they collapse as a whole.

Intel OTOH, might show up with laughable GPUs and bail out again. Hard to say what happens here, but like AMD what they need most are good integrated solutions and strong 2050/2060 competitors. They are a big wild card.

 

[Qwertilot]

Tend to assume Intel are also - mainly? - after the gpu compute market. Better fit for them than gaming and huge potential profits.

 

[TheELF]

The reasons AMD bought ATI still exist today, and really the GPU has only has grown in importance, which is why Intel wants in.

The vast majority of devices are powered by chips that have both CPU and GPU elements (APUs/IGP/SoC). Having any kind of mainstream CPU business requires integrating GPUs. AMD absolutely needs RTG for that.

But after all these years they still can only add an iGPU to a max quad core because otherwise it would become a ridiculous monstrosity that motherborads would just be incapable of supporting.

 

[TheELF]

Tend to assume Intel are also - mainly? - after the gpu compute market. Better fit for them than gaming and huge potential profits.

https://en.wikipedia.org/wiki/Xeon_Phi
They already have that covered,they took a GPU design and used x86 cores ,making another new entry to rival their own older design... might be possible but I don't think so.

Both amd and nvidia have become so good at what they do, after so many years that the lowest GPU that nvidia makes is the 1030 and right now costs around $90 same with amd and the rx550 and they are pretty powerful,I think intel is going to target the market between iGPU and the 1030/550,providing cheap(ish) GPUs that are packed full of features like qsv Vulkan OpenGL HLSL OpenCL and so on and enough gaming power to justify buying them at a small cost,that's a realistic goal for intel to start with, compete against ancient amd/nvidia GPUs in price and beat those old ones at current gaming.

 

[PeterScott]

But after all these years they still can only add an iGPU to a max quad core because otherwise it would become a ridiculous monstrosity that motherborads would just be incapable of supporting.

A MB could support an IGP and 8core, maybe not max bandwidth to both simultaneously, but often you don't need to do that. AMD was almost certainly just looking to keep die size (and cost) down.

IIRC Raven Ridge has a bigger die than 8 Core Ryzens, while selling for less. If your lowest ASP die, already has the biggest area, the last thing you want is to make it even bigger.

AMD may be considering more CPU cores for 7nm APUs, but that may depend on whether they are putting more cores on 7nm CPU only dies.