News Intel GPUs - Falcon Shores cancelled

[LightningZ71]

In a market that is essentially controlled by a duopoly (AMD and NVIDIA) to gain ANY traction, which includes programmer and developer mindshare as well as customers willing to try your new thing, you ABSOLUTELY HAVE TO sell at a loss to provide maximum value to customers to then have enough critical installation numbers to justify anyone giving a care about performance on your products.

Intel HAS to burn cash to gain a foothold in the market. They have zero choice. They can't chase the bottom end for numbers because their own iGPUs as well as AMD's are gutting that market at the same time. The $199-$259 area is where the battlefield is. 12GB with 4060 range performance is a solid target as you hit NVIDIA where they are most stingy (VRAM) as well as undercutting AMD on price (which is where AMD draws it's market).

Buyers don't care how much the silicon costs. They just know what performance level they can afford.

 

[gdansk]

In a market that is essentially controlled by a duopoly (AMD and NVIDIA)

That's very generous to AMD

 

[adroc_thurston]

With B580, Intel has a great chance against 4060, especially overseas

They do not?
It's like twice the die area lol.

Their main objective is creating a customer base who will buy their future cards

this meme again? Pat was literally just sacked for doing too much FCF torching.

Intel HAS to burn cash to gain a foothold in the market.

Then they don't enter the market. That simple.

Buyers don't care how much the silicon costs.

Yeah they do, better PPA means more oomf per dollar. always.

That's very generous to AMD

they literally own home consoles.

 

[LightningZ71]

That's very generous to AMD

Generous is maybe a bit harsh, but they are the clear #2 if you don't include game consoles and much more competitive if you do include consoles and their APU iGPUs in the total market.

 

[LightningZ71]

They do not?
It's like twice the die area lol.

this meme again? Pat was literally just sacked for doing too much FCF torching.

Then they don't enter the market. That simple.

Yeah they do, better PPA means more oomf per dollar. always.

they literally own home consoles.

Not entering the market is a decision for the past. They entered it. They can choose to leave it and write off a ton of development costs, or they can try to find a way to make it work.

PPA means more omph for COST. It matters note once it's on the shelf. And with both well within most power budgets, that doesn't matter either.

 

[GodisanAtheist]

If anything I am now more interested in seeing what AMD and NV do with respect to their 8600 and 5060.

If either of those cards provide the usual 15-20% performance boost for less than $300 then stick a fork in 'em, Intel Arc is done.

 

[adroc_thurston]

write off a ton of development costs

There isn't even that much.

It matters note once it's on the shelf. And with both well within most power budgets, that doesn't matter either.

Yeah it does.
Better PPA means you either sell more oomph iso sticker price (aka better market penetration) or sell iso perf for less (aka also better market penetration).
Again, you can't bludgeon the market as PPA-centric as GPU with this turd.

If either of those cards provide the usual 15-20% performance boost for less than $300 then stick a fork in 'em, Intel Arc is done.

The problem is they're both 128b parts so 8GB is a pretty hard limit.

 

[DAPUNISHER]

Why? Support has been there since Ryzen 3xxx and Intel 10th gen, and supposedly even many Intel 8/9th gen for some mobos. No reason to fix it.

Nonsense. Cast a wider net catch more fish. So many old office PCs and cheaper retail boards that don't have the setting exposed. Very few are going to go through the process of modding the bios to expose it. All those fish getting away will hurt, just as it has with A series.

 

[Hans Gruber]

Generous is maybe a bit harsh, but they are the clear #2 if you don't include game consoles and much more competitive if you do include consoles and their APU iGPUs in the total market.

A duopoly would be Boeing and Airbus. Considering Nvidia owns almost 90% of the GPU market. That is classified as a monopoly.

 

[adroc_thurston]

That is classified as a monopoly.

They even remotely aren't.
One real slip and they'll be bleeding MSS (especially DC MSS) like a gaping wound.

 

[Hans Gruber]

They even remotely aren't.
One real slip and they'll be bleeding MSS (especially DC MSS) like a gaping wound.

88% of the GPU market is not a monopoly?

 

[gdansk]

88% of the GPU market is not a monopoly?

Or 95% of the DC GPU market share?

 

[adroc_thurston]

88% of the GPU market is not a monopoly?

No.

Or 95% of the DC GPU market share?

Also no.
NV was a monopoly in DC like 5 years ago. Not now.

 

[poke01]

selling a product at a loss is more than fine if your flush with cash, Intel is not.

These have to sell WELL.

 

[adroc_thurston]

selling a product at a loss is more than if your flush with cash, Intel is not.

Loss is a relative thing, but this just isn't a viable PPA part.
If this was a 200mm^2 die, would've been alright.

 

[blckgrffn]

So just trying to put this into context of cards we know... this could be like a 12GB 3060Ti? About the performance of a 2080 Ti when you squint? 6750XT? Something short of the 4060 Ti 16GB and its laughable $450 price tag though.



I mean, for $250 seems alright. This could be a strong contender in prebuilt market where at least a 500W PSU is the norm and "moar memory" is attractive.

190W and all that die area is pretty nutso, I agree. That's crazy inefficient.

Now, about those drivers. Please mature faster and require less software magic to function.

 

[gaav87]

If this is 272mm2 than g31=420-450+ xD and bmg X4 must have been like 520mm2
Thats rdna2 perf/size on n5 xD

 

[igor_kavinski]

He must really be worried about Pat.

 

[Kepler_L2]

If this is 272mm2 than g31=420-450+ xD and bmg X4 must have been like 520mm2
Thats rdna2 perf/size on n5 xD

G31 is 390mm² IIRC.

 

[adroc_thurston]

G31 is 390mm² IIRC.

epic

 

[poke01]

G31 is 390mm² IIRC.

This makes Apple’s iGPUs look sane….

 

[ToTTenTranz]

When they say that, they're projecting hopium of Intel making NV GPUs cheaper.

How is this hopium if they're literally increasing performance per dollar to the end user by a substantial amount?

The customer side of me couldn't care less if this is using a 100mm^2 or 400mm^2 chip. And as long as I don't need to buy another power supply, I don't really care if it consumes 100 or 300W either.

People looking for a GPU that plays great at 1080p or 1440p can buy this, instead of paying an extra 50% or more for a Nvidia equivalent. So if this doesn't force Nvidia to lower their prices, their GPUs can rot on the shelves for all I care.

 

[adroc_thurston]

How is this hopium if they're literally increasing performance per dollar to the end user by a substantial amount?

Selling 270mm^2 N5 dies at a cost is not a viable business model and you know it.

The customer side of me couldn't care less if this is using a 100mm^2 or 400mm^2 chip.

Yeah, the customer side cares about buying nVidia. And good news, NV has way better PPA than this and can market squeeze as the result!

instead of paying an extra 50% or more for a Nvidia equivalent

NV cuts prices a wee bit and they win. Intel can't price it any cheaper due to BOM.
You need competitive or better PPA to pry into this market.

 

[511]

If this is 272mm2 than g31=420-450+ xD and bmg X4 must have been like 520mm2
Thats rdna2 perf/size on n5 xD

I have heard G31 is around ~400mm2

 

[branch_suggestion]

It doesn't look bad against a 2yo IP crippled with the lack of SLC and a tiny 2M L2. On a worse node.

Curse you Microsoft.