Rumor : Intel to show discrete GPU at CES 2019

[App4that]

I don't see much coming out of this. This is just Intel being angry at Nvidia for taking AI moneis away from them. 

[Notional]

2 hours ago, mqe said:

Ye and 10 years before ryzen, pls tell me you sent this to strengthen my argument and didn t want to undermine it with ten year old success

What did Intel do a little over a decade ago? As in the years leading up to AMD's massive fall in market share?

[killcomic]

1 hour ago, VegetableStu said:

come to think of it, an intel GPU most definitely can't do CUDA stuff, so if they were to go head to head against nvidia via OpenCL wouldn't they kill AMD instead? ._.

Depends on what market they are going for. If they concentrate on gaming as opposed to content creation, lack of CUDA might not be too detrimental. 

[RadiatingLight]

6 hours ago, The Viking said:

 

6 hours ago, AnonymousGuy said:

IIRC this was a % of new sales chart, which means that Intel had 65% of the sales and AMD had the remaining 35% meaning they were still losing overall market share just at a lower rate.

 

 

Well, just an outlier I think.

When looking at the same chart but updated, it doesn't look so good for AMD

NOTE: Not an intel shill, just presenting an updated version of the same source. I'm strongly rooting for AMD to succeed

 

[GoldenLag]

anyone wanna guess how much of a fortune the first Intel GPU will cost on it launch? (im counting this as the first in know there are other just ssshh)    im guessing 1000+$.

[killcomic]

3 minutes ago, GoldenLag said:

anyone wanna guess how much of a fortune the first Intel GPU will cost on it launch? (im counting this as the first in know there are other just ssshh)    im guessing 1000+$.

I doubt they'll be able to go against AMD's or Nvidia's top end. My guess is that it's going to be a mainstream low or mid range card priced with a price to match.

[GoldenLag]

5 minutes ago, killcomic said:

I doubt they'll be able to go against AMD's or Nvidia's top end. My guess is that it's going to be a mainstream low or mid range card priced with a price to match.

the issue with that notion is how intel is going to get optimization for low end cards when they start off with no market share at all. without a high end card or mid-range card say a 1060 in performance i will find it hard to believe they will have much optimization going for it. AMD is still rolling cgn and have maintained thei optimization to a degree. Intel will have to push a lot of money into this if they want to succeed with low end cards and even high end cards. 

[Quadriplegic]

9 hours ago, mqe said:

I don t believe it. Why would they want to do this now? If anytime they would have done it before ryzen when it was just intel not two competitors.

@ all amd fans pls be quiet amd was irrelevant before ryzen

What do dedicated GPUs have to do with Ryzen?

[GoldenLag]

Just now, Quadriplegic said:

What do dedicated GPUs have to do with Ryzen?

especially considering polaris has been quite compelling at their respected pricepoint. There is the graphics division of AMD and then there is the CPU division. Intel is quite similar in that regard and i doubt intel would have managed to put in a GPU earlier than this

[leadeater]

8 hours ago, VegetableStu said:

come to think of it, an intel GPU most definitely can't do CUDA stuff, so if they were to go head to head against nvidia via OpenCL wouldn't they kill AMD instead? ._.

Any improvement to OpenCL will help AMD, even if it's Intel pushing in to the space with their own product. CUDA is killing AMD, they need OpenCL to gain feature/performance/market parity.

[GoldenLag]

2 minutes ago, leadeater said:

Any improvement to OpenCL will help AMD, even if it's Intel pushing in to the space with their own product. CUDA is killing AMD, they need OpenCL to gain feature/performance/market parity.

if AMD opens their open-source tech to Intel when intel pushes into the market things like hairworks and other proprietary tech that kills AMD performance would die off as the open source is more viable for intel. in many ways this will help AMD and lower Nvidia*s grasp on the market. 

[Guest The Viking]

2 hours ago, RadiatingLight said:

 

 

 

Well, just an outlier I think.

When looking at the same chart but updated, it doesn't look so good for AMD

NOTE: Not an intel shill, just presenting an updated version of the same source. I'm strongly rooting for AMD to succeed

 

-graph-

the graph themselves aren't 100% reliable, it's made from people benchmarking systems. So you have situations like this

 

AMD may have lost even more marketshare, but I think it's more about Intel's new cpus are out so everybody's benchmarking them. AMD will probably go back up slightly once next quarter numbers are in thanks to ryzen refresh.

 

but, they do give an approximate look at marketshare, so yeah, I guess coffeelake is hurting those AMD sales

 

[Sniperfox47]

9 hours ago, VegetableStu said:

come to think of it, an intel GPU most definitely can't do CUDA stuff, so if they were to go head to head against nvidia via OpenCL wouldn't they kill AMD instead? ._.

That's the best part for Intel! They don't have to worry about any of that because AMD already put all the work into CUDA to OpenCL converters!

 

https://github.com/ROCm-Developer-Tools/HIP

 

*Edit: meant OpenCL, not OpenGL. I blame autocorrect.

[Sniperfox47]

2 minutes ago, VegetableStu said:

whoa what?! O_O TIL

Just to clarify meant OpenCL, not openGL. Typo.

[bcredeur97]

17 hours ago, AresKrieger said:

Well this isn't their first attempt, they could have been working on one for years and brought people on in an attempt to hasten progress and pound out the kinks. 

 

Granted this is a rumor so season to taste

I think we heard a couple of rumors over the past few years that never really came to light.... but does show that this could of been in the works for a long time now.

[bcredeur97]

on topic of AMD marketshare... while I really like the new ryzen stuff I see very little of it on store shelves which is what the majority of the market share is coming from....

 

they honestly need more laptop stuff. Get in the big name brands. nice stuff that makes people want to buy it.

Being best choice for a price/performance PC build only gets you so far...

[asus killer]

15 hours ago, bcredeur97 said:

on topic of AMD marketshare... while I really like the new ryzen stuff I see very little of it on store shelves which is what the majority of the market share is coming from....

 

they honestly need more laptop stuff. Get in the big name brands. nice stuff that makes people want to buy it.

Being best choice for a price/performance PC build only gets you so far...

"normal stores" (i mean like big stores that aren't dedicated to computers but that also sells them, and i would say they sell a lot more than dedicated ones) start to have amd laptops, but it's really almost a Intel show, i would say about 90%. And most people prefer a laptop than a desktop this days. And the advertising space is also almost all Intel.

[leadeater]

18 hours ago, bcredeur97 said:

on topic of AMD marketshare... while I really like the new ryzen stuff I see very little of it on store shelves which is what the majority of the market share is coming from....

 

they honestly need more laptop stuff. Get in the big name brands. nice stuff that makes people want to buy it.

Being best choice for a price/performance PC build only gets you so far...

Might see more laptops with them now that the APUs are a thing, couldn't really do mass market laptops with no integrated GPU. Wonder what the laptop areas in stores will look like in 6 months time.

[laminutederire]

On 08/05/2018 at 3:22 AM, killcomic said:

Depends on what market they are going for. If they concentrate on gaming as opposed to content creation, lack of CUDA might not be too detrimental. 

Lack of CUDA for ML may not be an issue. They'll just finally pump money into OpenCL to help making it progress fast enough to catch up on what's important.

It might be good for AMD as well to be honest. (And good for science)

[leadeater]

5 minutes ago, laminutederire said:

Lack of CUDA for ML may not be an issue. They'll just finally pump money into OpenCL to help making it progress fast enough to catch up on what's important.

It might be good for AMD as well to be honest. (And good for science)

I find it amusing how entrenched CUDA actually is in the scientific and education community considering how open source do it yourself they are.

[laminutederire]

4 minutes ago, leadeater said:

I find it amusing how entrenched CUDA actually is in the scientific and education community considering how open source do it yourself they are.

I think it may be because most research is done through start-ups, or company-university projects in that particular area.

That and... to be fair, computer science people are often lazy. So are mathematicians, but those are lazy the good way when computer scientists are lazy the bad way.

On top of that you add that top universities are mostly about money, and Nvidia does cater to them by giving them gifts and so on.

The cherry on top is that computer science degrees are often too specialized, so optimizing on GPU is something that most programmers/theroricists don't really know about, and ML theory is something electrical engineers don't do as much. So you get two crowds that should work together to make OpenCL great, which don't. So one crowd use the API from the other crowd of Nvidia who somewhat did it for them.

[Trixanity]

Intel becoming a heavy weight could be a double-edged sword. It could spell the end for a lot of Nvidia's proprietary crap but Intel may yet just add their own proprietary crap and we could still end up with a lopsided market despite having a third player.

 

There is no way Intel's presence has any positive effect on Nvidia but it's still undetermined if it would help or harm AMD.

We'll see.

[Eroda]

just because they could be unveiling it doesnt mean they have a product. maybe this is the annoucement stating their intent ( because they already have been working on it )

[orbitalbuzzsaw]

27 minutes ago, Eroda said:

just because they could be unveiling it doesnt mean they have a product. maybe this is the annoucement stating their intent ( because they already have been working on it )

We know Intel's been working on a dGPU in some capacity for years - that's what became Xeon Phi. I wouldn't be surprised if we see an early prototype at next year's CES. Granted, it probably won't make it to market before about Q3 2020

[Eroda]

could quite iterally be a Wafer  of a "dGPU"