News Intel Battlemage, Celestial GPUs Are Booked in at TSMC: Report

[Admin]

Intel has signed up for large orders of Battlemage and Celestial GPUs at TSMC. Products are expected in H2 2024 and H2 2026, respectively.

Intel Battlemage, Celestial GPUs Are Booked in at TSMC: Report : Read more

 

[MBOO7]

"Most positively, the driver team has been beating expectations with the rate and quality of updates over recent months. "
Pleeeasse...
So we do now give kudos to a multi billion dollar company.. that released a highly dysfunctional and broken product and is now fixing their mutiple own fk ups? Something that should have beeen actually a given ???
Come on! I know what you mean, but the way you say it sounds just deliberately wrong.
I highly disagree with your statement on this one, but I can at least agree to the rest of what you said.

 

[gg83]

"Most positively, the driver team has been beating expectations with the rate and quality of updates over recent months. "
Pleeeasse...
So we do now give kudos to a multi billion dollar company.. that released a highly dysfunctional and broken product and is now fixing their mutiple own fk ups? Something that should have beeen actually a given ???
Come on! I know what you mean, but the way you say it sounds just deliberately wrong.
I highly disagree with your statement on this one, but I can at least agree to the rest of what you said.

People expected the driver team to fail and quit. So they are shattering expectations. The bar is lower for sure though.

 

[atomicWAR]

So we do now give kudos to a multi billion dollar company.. that released a highly dysfunctional and broken product and is now fixing their mutiple own fk ups?

We do it all the time but your not wrong...

People expected the driver team to fail and quit. So they are shattering expectations. The bar is lower for sure though.

The bar wasn't just lower, it was buried under six feet of fud. Its amazing Intel managed to dig the danged thing up and get their card functional at all. I honestly expected Intel to already have killed the project at this point. Much to my delight they haven't as we are in desperate need of more competition in the GPU space. The question is will Intel get competitive or fold first?

 

[Elusive Ruse]

All Intel needs to do is deliver a working midrange GPU on launch day and they'll sell a ton of them.

 

[InvalidError]

The A750/770 are still punching a long way below expectations in most games going by FP32 performance. I hope future drivers will consistently unlock more of that seemingly often under-tapped potential.

All Intel needs to do is deliver a working midrange GPU on launch day and they'll sell a ton of them.

The crypto-craze days are gone. AMD and Nvidia have working GPUs gathering dust on store shelves and warehouses right now with sales at 20+ years lows. Having products on the market isn't enough anymore, they also need to be priced low enough for normal gamers to actually want to buy them. Intel also bears the additional burden of new-kid-on-the-block status and that will take 2-3 technically and marketably successful generations to achieve. If it tries to sell almost exactly the same performance at almost exactly the same price as AMD or Nvidia, it will fail.

 

[Co BIY]

I think the the fact that the GPUs will be produced at TSMC reduces the impact of Intel entering the market for the consumer. It means that all three competitors are going to have the same base tech and the same base costs. TSMC will be taking a large cut out of all three.

I believed adding Intel's manufacturing weight to the GPU market supply would be very beneficial. It also made more sense for continuity with it's iGPUs.

 

[InvalidError]

I believed adding Intel's manufacturing weight to the GPU market supply would be very beneficial. It also made more sense for continuity with it's iGPUs.

Since Meteor Lake and beyond will use IGP tiles, the IGP tiles can be made at the same place as their monolithic dGPU counterparts. No continuity issues there.

 

[Howardohyea]

I'm pretty concerned about Intel's GPU division, Q3 2024 would put it just in front of RTX 5000 and RX 8000 series, and unless Intel could leapfrog Nvidia and AMD they're going after RTX 4000.

Furthermore if their previous leaked "target of <225 watts" holds true, and assuming an enormous jump in efficiency I think they're going to land somewhere around 3080/4070Ti levels of performance, and competing with a later generation at the same time.

 

[InvalidError]

I'm pretty concerned about Intel's GPU division, Q3 2024 would put it just in front of RTX 5000 and RX 8000 series, and unless Intel could leapfrog Nvidia and AMD they're going after RTX 4000.

I wouldn't be concerned at all about that as Intel is likely still two generations away from getting their drivers coming fully up to par and won't be considered a credible competitor until then no matter how good the hardware may hypothetically be on paper.

Right now, the A750 barely competes with AMD and Nvidia GPUs that have half as much raw compute performance with a few games faring drastically worse. Regardless of how much Intel's drivers may have improved since launch, they still have a very long way to go.

 

[RedBear87]

Still waiting for MLID to say that Battlemage is cancelled and if not Celestial is cancelled and if even that doesn't come to happen, then Celestial's successor is cancelled for sure. It's like a mantra.

 

[purple_dragon]

Intel's design tends to scale well with resolution, higher resolution = better performance. If they can sort out their driver problems and refine the silicon they have a chance.

 

[InvalidError]

Still waiting for MLID to say that Battlemage is cancelled and if not Celestial is cancelled and if even that doesn't come to happen, then Celestial's successor is cancelled for sure. It's like a mantra.

MLID already said BM and Celestial were most likely dead months ago and reiterated that statement a week or so ago.

With Meteor Lake using BM IGP tile and Arrow Lake using BM or Celestial IGP tile depending on SKU, Xe has at least two generations ahead of it at the absolute minimum.

As I have written in other "Xe doom and gloom" threads, I cannot imagine Intel giving up on having its own in-house IGPs and now that IGPs are becoming tiles, it would be pretty easy for Intel to spin off its IGPs as dGPUs by simply slapping an IO tile next to an IGP tile to handle video outputs, PCIe, VRAM (if not 3D-stacked on GPU by then) and whatever else.

 

[renz496]

Still waiting for MLID to say that Battlemage is cancelled and if not Celestial is cancelled and if even that doesn't come to happen, then Celestial's successor is cancelled for sure. It's like a mantra.

Doesn't matter how promising the tech is if intel still can't turn the discrete gpu into a profitable venture with Battlemage they probably will pull the plug

 

[renz496]

I'm pretty concerned about Intel's GPU division, Q3 2024 would put it just in front of RTX 5000 and RX 8000 series, and unless Intel could leapfrog Nvidia and AMD they're going after RTX 4000.

Furthermore if their previous leaked "target of <225 watts" holds true, and assuming an enormous jump in efficiency I think they're going to land somewhere around 3080/4070Ti levels of performance, and competing with a later generation at the same time.

Intel biggest concern is to put out decent product and start making profit rather than have to worried about competing head to head with the big two. Also competing on the same level as nvidia/amd is not even on their radar when they start this venture.

 

[TerryLaze]

Doesn't matter how promising the tech is if intel still can't turn the discrete gpu into a profitable venture with Battlemage they probably will pull the plug

Intel is doing all of this for the server GPU market, dGPU desktop just happens because they spend so much on the server GPUs that it just makes sense to also make desktop GPUs because the additional cost is small, comparatively, and it gives them some sales and some free advertisement due to articles like this one.
As long as the server GPU division makes money they will also be keeping the desktop GPUs.

Also their GPUs don't have to be a great seller to DIY, if they can secure some OEM deals to put GPUs into low/mid end prebuilds then that's all it takes for them to make it profitable.

 

[InvalidError]

Also their GPUs don't have to be a great seller to DIY, if they can secure some OEM deals to put GPUs into low/mid end prebuilds then that's all it takes for them to make it profitable.

If Intel's dGPUs aren't any good, practically nobody will want them in OEM systems either including the OEM themselves who would get stuck with the unsold stock and complaints about what sold through.

If Intel's graphics get pounded into the ground so bad that OEMs and IGPs are the only markets Intel can get into, it would make more sense to make bigger IGPs and axe dGPUs altogether. Roughly the same performance class without the worries of PCIe slots snapping off in transit.

 

[Ogotai]

Also their GPUs don't have to be a great seller to DIY, if they can secure some OEM deals to put GPUs into low/mid end prebuilds then that's all it takes for them to make it profitable.

yea ok sure. chose an arc gpu that is slower then the equivalent amd or nvidia gpu, for the same price ? no thanks. id rather get the nvidia or amd gpu. if given the option to choose, ill deselect the intel gpu and pick any thing else if the option exists. so this could not work at all for intel.

arc ranges in price from $330 for the 750 to $480 for the 770, for those those prices i can get at least a 6500 XT or a gtx 1650, and know those will work for everything.

they need to get their drivers sorted out, and working. plain and simple, seems, its not quite there yet.

 

[jkflipflop98]

The driver situation will be much better in the future. They're going with "how the other guys do it" approach. Consumer and Pro instead of the Alchemist Consumer/Pro/HPC/LowPower. That right there doubles the amount of attention each package will receive.

 

[-Fran-]

MLID already said BM and Celestial were most likely dead months ago and reiterated that statement a week or so ago.

With Meteor Lake using BM IGP tile and Arrow Lake using BM or Celestial IGP tile depending on SKU, Xe has at least two generations ahead of it at the absolute minimum.

As I have written in other "Xe doom and gloom" threads, I cannot imagine Intel giving up on having its own in-house IGPs and now that IGPs are becoming tiles, it would be pretty easy for Intel to spin off its IGPs as dGPUs by simply slapping an IO tile next to an IGP tile to handle video outputs, PCIe, VRAM (if not 3D-stacked on GPU by then) and whatever else.

He's never said the architectures were killed, only some specific products for the line up were going to be if things didn't change drastically, which seem to have not.

His predictions/leaks of AXG and all the shenanigans that have ensued ever since have been, while not 100% accurate, pretty spot on given the circumstances. He didn't predict Raja would be pushed out of Intel though, but for all the crap people has been giving him (MLiD), that same people is now showing clear signs of copium with ARC.

Like... The same people made fun of AMD and the "fine wine" and then when Ryzen was launched on AM4, but ARC somehow will be ok with the building on fire behind them. I can only shrug at that; laugh at times.

Regards.

 

[TerryLaze]

If Intel's dGPUs aren't any good, practically nobody will want them in OEM systems either including the OEM themselves who would get stuck with the unsold stock and complaints about what sold through.

If Intel's graphics get pounded into the ground so bad that OEMs and IGPs are the only markets Intel can get into, it would make more sense to make bigger IGPs and axe dGPUs altogether. Roughly the same performance class without the worries of PCIe slots snapping off in transit.

If they weren't good enough then intel wouldn't make large orders for them...unless you can come up with a good reason.
I'm just saying that gamer-market penetration isn't going to be a sign for if they are profitable or not.

yea ok sure. chose an arc gpu that is slower then the equivalent amd or nvidia gpu, for the same price ? no thanks. id rather get the nvidia or amd gpu. if given the option to choose, ill deselect the intel gpu and pick any thing else if the option exists. so this could not work at all for intel.

arc ranges in price from $330 for the 750 to $480 for the 770, for those those prices i can get at least a 6500 XT or a gtx 1650, and know those will work for everything.

they need to get their drivers sorted out, and working. plain and simple, seems, its not quite there yet.

I talked about pre-builds, like the ones you see on the shelfs in super-markets, or that big companies get for their offices.
Gamer-market isn't everything especially in the beginning when everybody else has decades of head start, although I'm sure they will also get better in that market as well.

 

[InvalidError]

If they weren't good enough then intel wouldn't make large orders for them...unless you can come up with a good reason.

Committing to order is no guarantee of actually being able to sell the stuff anywhere near the initially intended price. When Intel originally ordered Alchemist wafers, it was likely expecting to sell the A770 for $600-700 but had to lower expectations by almost half to actually shift units by the time it launched with more price cuts likely incoming once the software bundle offer expires.

If BMG starts at ~2X the A770 on performance as rumors suggest, there is a pretty high likelihood it will price itself out of consideration from most people willing to give it a shot again.

 

[gg83]

I'm pretty concerned about Intel's GPU division, Q3 2024 would put it just in front of RTX 5000 and RX 8000 series, and unless Intel could leapfrog Nvidia and AMD they're going after RTX 4000.

Furthermore if their previous leaked "target of <225 watts" holds true, and assuming an enormous jump in efficiency I think they're going to land somewhere around 3080/4070Ti levels of performance, and competing with a later generation at the same time.

But at the same time the 4070 is basically a 3080 +$100. So the newer generations might not be a big jump at the midrange? It seems like AMD and Nvidia only make the high-end cards significantly better than the previous generation. I read one chart showing the 3050 on par with the 1060 non-rtx.

 

[InvalidError]

It seems like AMD and Nvidia only make the high-end cards significantly better than the previous generation. I read one chart showing the 3050 on par with the 1060 non-rtx.

In benchmarks, the RTX3050 is much faster than the GTX1060, consistently a few steps ahead of the GTX1660 Super and ~10% behind the RTX2060.

 

[TerryLaze]

Committing to order is no guarantee of actually being able to sell the stuff anywhere near the initially intended price.

Did anybody say that it is?!
A large commitment still shows that at least intel things that they will be good enough to sell them, maybe not at the 50% + margin they would hope to, but still.
And it's not even confirmed that it is a large commit, I'm just saying that if it is a large one...