News Intel GPUs - Falcon Shores cancelled

[Hans Gruber]

Ok so my leaks from my plumber intel source were correct. xD
I can see this sit between 4060ti and 4070 after 6-12months. In some benches its close to 4070FE already... Overclocking is borked but when it will work easy 4070FE perf...
G31 is close to 4070ti in some scenarios/games. in some its between 4060ti and 4070fe. View attachment 113118View attachment 113119

The B580 is not the top Battlemage chip. The top chip is supposed to be 4070-4070ti performance if Intel releases it. Their roadmap has Celestial releasing Mid 2025 or Q3. One could argue Battlemage is their proof of concept chip to ready drivers for Celestial and to gain market share.

 

[Kepler_L2]

300 sq mm chip (plus VRAM, big board) sold for $250 takes capacity from AMD to sell 4x 75mm Zen 5s sold from $1200 (consumer) to $5000 (server) chiplets. Intel might take 100 bucks loss on those, but still be "quid in" in terms of saving their Xeon sales that otherwise would be lost.

Chances are Intel decided to use booked N4 for something more useful than this B-grade loss leader though, but either way strategy works in suppressing competitor's volumes, they should really get a Pat on the back for devising it...

A few years ago that might have worked as a strategy, but these days Intel absolutely does not have the money to try that.

 

[Hans Gruber]

A few years ago that might have worked as a strategy, but these days Intel absolutely does not have the money to try that.

Did you not get the memo about the Biden chips act?

 

[adroc_thurston]

300 sq mm chip (plus VRAM, big board) sold for $250 takes capacity from AMD to sell 4x 75mm Zen 5s sold from $1200 (consumer) to $5000 (server) chiplets. Intel might take 100 bucks loss on those, but still be "quid in" in terms of saving their Xeon sales that otherwise would be lost.

Chances are Intel decided to use booked N4 for something more useful than this B-grade loss leader though, but either way strategy works in suppressing competitor's volumes, they should really get a Pat on the back for devising it...

Intel does not have the FCF to do anything like that.
It ain't 2018 no more.

Did you not get the memo about the Biden chips act?

Pennies.

 

[Hans Gruber]

Intel does not have the FCF to do anything like that.
It ain't 2018 no more.

Pennies.

8,500,000,000,000 pennies which is $8.5 billion in Biden chips act funding.

 

[blckgrffn]

8,500,000,000,000 pennies which is $8.5 billion in Biden chips act funding.

Have you looked at the capital outlays into fabs that funding required to pull in?

And that comes into play, what, 4-6 years from now?

 

[Hulk]

At the end of the day most people aren't intested in the node their graphics adapter was built on or if it uses a few more watts than a similar one. They care about the up-front cost and fps. This isn't a home run for Intel, but it is a double. They got it out the door, it beats last gen on a smaller, more efficient die, and the drivers for the most part work. I am happy for Intel as they needed just something to go their way.

If they can get out 750/770 Battlemage parts in a timely manner that can compete with the next tier of cards up the stack they may have nudged their foot into the part open door of the graphics market. Funny thing here is that in the old paradigm, nVidia is Intel and nVidia could crush them out of the market by simply beating them on price as they have the pockets to "get rid of them" in that manner, as Intel used to frequenty do with their "up starty" competition.

Does Battlemage use a different driver from Alchemist?

 

[Hulk]

Forgot to mention it looks like what Raja had said a while back about hardware resource allocation was spot on. Some parts of Alchemist are over built and those resources should have gone to the bottlenecked area. Intel not only figured it out but they did what you are supposed to do with the next gen product, they correctly allocated hardward resources so everything is "evenly matched" in the GPU. This is why I believe Battlemage outperforms Alchemist with few cores and shaders. Yes, I know frequency was bumped up as well.

Except for frequency, on paper Alchemist is better than Battlemage in every way, yet Battlemage is more performant. To quote Jurrasic Park, "Well, there you have it."

 

[adroc_thurston]

8,500,000,000,000 pennies which is $8.5 billion in Biden chips act funding.

That's pennies given Intel yearly capex. Get real.

Except for frequency, on paper Alchemist is better than Battlemage in every way, yet Battlemage is more performant. To quote Jurrasic Park, "Well, there you have it."

Utterly irrelevant metric given the final product PPA is ~ACM.

 

[burninatortech4]

Impressive stuff. I was expecting to say "Why get this when I can just get a 7600XT 16GB?" and that's no longer the case.

Competition is good (3-way now). Intel really needed this.

 

[reaperrr3]

Impressive stuff. I was expecting to say "Why get this when I can just get a 7600XT 16GB?" and that's no longer the case.

Competition is good (3-way now). Intel really needed this.

Except
- Intel is probably losing money on this product, something they can't really afford anymore. They needed this kind of performance 2 years ago, and they needed this kind of perf/mm² and perf/W on N6, rather than N5/4.
- if they do gain some marketshare with this, it'll be mostly at AMD's expense, rather than Nvidia's, which only really increases the chances of Nvidia being the only one left at some point
- Soon we'll probably be able to say "why get this when I can just get a 8600XT 16GB", anyway.

 

[SolidQ]

why get this when I can just get a 8600XT 16GB

Unless AMD gonna surpise us and put 16gb, instead 8gb on N44

 

[GodisanAtheist]

Unless AMD gonna surpise us and put 16gb, instead 8gb on N44

- Figure it'll be the same play as this gen: 8600 8GB and 8600XT 16GB

 

[SolidQ]

Figure it'll be the same play as this gen: 8600 8GB and 8600XT 16GB

16GB would be fine, looking at DF B580 video, 7600 there pretty poor.


8gb is poor

 

[dr1337]

- Intel is probably losing money on this product, something they can't really afford anymore

And what if they're actually making a profit?

I have an extremely hard believing this is a loss leader product, not like it's even super complicated, just an iteration on existing tech. Intel of all companies needing to buy mindshare doesn't make sense when they already have had compelling products one generation before. I indeed staunchly believe that good products sell themselves.

 

[Kepler_L2]

And what if they're actually making a profit?

I have an extremely hard believing this is a loss leader product, not like it's even super complicated, just an iteration on existing tech. Intel of all companies needing to buy mindshare doesn't make sense when they already have had compelling products one generation before. I indeed staunchly believe that good products sell themselves.

B580 should be 10%-ish cheaper to manufacture than the 4070 Super, which has an MSRP of $600. So either this is sold at a loss, or Jensen is robbing gamers blind.

 

[Win2012R2]

A few years ago that might have worked as a strategy, but these days Intel absolutely does not have the money to try that.

This was planned when they had the money - in 2021 their annual revenue was $79 bln, Pat joined 15 Feb 2021.

B580 should be 10%-ish cheaper to manufacture than the 4070 Super, which has an MSRP of $600. So either this is sold at a loss, or Jensen is robbing gamers blind.

Before AI crazyness NVIDIA's gross margins were 60%-ish in 2021-23, but even that was certainly propped up by much higher margin DC products, so if we assume that NVIDIA gets 50% of the above price as gross profit then Intel must be selling at a loss, plus they won't have as good economics of scale for GDDR6 and other parts.

So IF Intel loses 100 bucks on each B-sale, then how long can they entertain with their current financial position?

 

[MrTeal]

B580 should be 10%-ish cheaper to manufacture than the 4070 Super, which has an MSRP of $600. So either this is sold at a loss, or Jensen is robbing gamers blind.

Even if the BOM is breakeven, it's hard to imagine they're covering all the engineering costs and making a profit on the product as a whole.

 

[dr1337]

Even if the BOM is breakeven, it's hard to imagine they're covering all the engineering costs and making a profit on the product as a whole.

If it were so expensive for them, I can't imagine intel even bothering. It would probably be their craziest investment. Giving a competitor fab extra money just to sell products at a loss would be the first thing I'd expect the board to cancel.

It could be true they sell at a loss sure, anything is possible, but it's more likely Nvidia and even AMD are selling cards at more than 200% markup. For intel to invest in desktop consumer facing GPUs on the level as a Playstation, Nintendo, or Xbox is suicide IMO and I'd be sad if that's what they're doing because that means they're gonna crash big time.

 

[CastleBravo]

B580 should be 10%-ish cheaper to manufacture than the 4070 Super, which has an MSRP of $600. So either this is sold at a loss, or Jensen is robbing gamers blind.

I think both are true.

 

[tajoh111]

8,500,000,000,000 pennies which is $8.5 billion in Biden chips act funding.

In terms of Fabs this is nothing. TSMC regularly spend 30-35 billion annually on R and D and Fabs with a better, cheaper and more efficient workforce.

Over a period of 4 years, I would say Intel needs 100 billion to keep up with TSMC on top of their own funding as a result of waste.

Intels numbers of 10% faster than a RTX 4060 are definitely on the optimistic side. $250 pricing is fair based on risk and performance on this chip. WIll be a loss on this product at this price. Too bad because aside from the financials seems like a decent product.

 

[poke01]

I want Intel to succeed but it seems some are against Intel joining the GPU market. Tell me what the **** did AMD do these last 6 years HUH? All they did was follow Nvidia.

In 2 generations, Intel has better upscaling, frame gen and better arch than AMD? Tom Peterson and his team know their stuff. Xe3 is coming next year, the point is Intel has yearly GPU arch updates.

The Intel team also fixes bugs faster and drivers updates are more common. I think by C770 or D770, Intel would gained great rep for their GPUs.


PS. Look into Intel's Framegen, its better than AMDs and Nvidias.

 

[poke01]

One more thing that people are missing out on, if Intel cancels or reduces funding for their GPU department, intel client cannot survive. SoC with big iGPUs on laptops are the future and GPUs form a ecosystem for your products.

 

[linkgoron]

Except
- Intel is probably losing money on this product, something they can't really afford anymore. They needed this kind of performance 2 years ago, and they needed this kind of perf/mm² and perf/W on N6, rather than N5/4.
- if they do gain some marketshare with this, it'll be mostly at AMD's expense, rather than Nvidia's, which only really increases the chances of Nvidia being the only one left at some point
- Soon we'll probably be able to say "why get this when I can just get a 8600XT 16GB", anyway.

It's quite clear that AMD is the one who has the most to lose from these cards. AMD failed with RDNA3, and RDNA4's might improve perf/watt or other metrics, but it won't really move the bar forward with regards to performance. I'm not sure where Navi 44 is supposed to land, but it looks like Intel is pretty competitive. Especially if Navi 44 lands with 8GB of memory.

In general, 7700XT/4060 performance at 1080p RT, and 6700xt/4060 performance in raster. That's not too bad for $250, this is probably what people wanted from Alchemist. If Intel can bring out a decent B770, AMD has a real challenger.

 

[Hulk]

B580 should be 10%-ish cheaper to manufacture than the 4070 Super, which has an MSRP of $600. So either this is sold at a loss, or Jensen is robbing gamers blind.

The second option. He's one of the richest men on the planet. Good for him.