News Intel: Arc A770 GPUs Will Launch “Very Soon”

[Admin]

Intel details features and capabilities of Arc A770 8GB and 16GB boards, Arc 750 cards.

Intel: Arc A770 GPUs Will Launch “Very Soon” : Read more

 

[LolaGT]

Yup, any day now.

After so may of those statements it is kind of meaningless.
When I see a boatload of actual benchmarks up on YT then I will think they might be out.

 

[InvalidError]

Alchemist has been "coming soon" for almost a year now. This has reached stale meme status at this point.

 

[Ogotai]

bah, who really cares about Arc now ? no one i know cares about it any more. they did like 6+ months ago, now, not so much.

 

[ezst036]

I'm looking forward to it. Intel has to do it as they can. The second gen will probably also have release to market issues - likely a lot less as lessons are learned. The third gen will be the one that has a flawless rollout.

 

[jp7189]

I'm looking forward to it. Intel has to do it as they can. The second gen will probably also have release to market issues - likely a lot less as lessons are learned. The third gen will be the one that has a flawless rollout.

Let's hope there is a 2nd and 3rd gen. By all accounts 1st gen is a massive hemorrhage of financial loss. It's going to be hard to convince the bean counters to keep pouring money in to this project.

 

[InvalidError]

Let's hope there is a 2nd and 3rd gen. By all accounts 1st gen is a massive hemorrhage of financial loss. It's going to be hard to convince the bean counters to keep pouring money in to this project.

The first generation of almost anything new Intel gets into always makes spectacular losses. The first generation of almost anything of substantial complexity by any company rarely covers the costs to start the thing from scratch. Though Xe-based GPUs aren't exactly new since Xe-based IGPs have been around since Tiger Lake, which make the recurring hiccups that much more unexpected. Still, Intel will need a graphics division for IGP even if it scraps dGPUs, so the likelihood of Intel quitting after only one bad run is effectively zero.

As I have written in other threads predicting the end of ARC, as AMD and Nvidia drift further away from sub-$300 GPUs, the need for decent IGPs to pick up that slack will become much greater and Intel giving up on decent graphics performance would hand over most of that market to AMD by default. I doubt Intel can afford to hand over a substantial chunk of its 60% of all PC graphics market share (and 70% of IGPs) to AMD, so Intel effectively cannot afford to fail here even if it takes three generations to produce something worthwhile and on schedule.

 

[Neilbob]

I'm looking forward to it. Intel has to do it as they can. The second gen will probably also have release to market issues - likely a lot less as lessons are learned. The third gen will be the one that has a flawless rollout.

They're on around their 22nd generation of x86 CPU releases and half the time don't quite manage a flawless rollout ...

 

[velocityg4]

While performance for mid range gamers may be decent. I don't think I'd want to be a alpha tester for Intel's ability to make reliable performance gaming drivers. AMD and nVidia still have issues with many years of experience and constantly have to release driver patches. I'd hate to think what the Intel experience will be like. Especially on games more than a few years old which they have little incentive to make patches for.

 

[The Historical Fidelity]

I would have felt this was good for the GPU market a year ago, but ARC being released now (a couple months away from nvidia 4000 and amd 7000 release) means that Intel a750 and a770 won’t be competing with the rtx 3060 and rx 6600xt in price and performance, but instead have to be priced even lower towards the next gen rtx4050 and rx 7500xt. That is not good margins for graphics cards that were meant for the 2021 market.

 

[InvalidError]

That is not good margins for graphics cards that were meant for the 2021 market.

Intel was never going to get great margins from Alchemist simply due to being the new kid on the dGPU block with a 20+ years long history of quirky IGP drivers regarding gaming. Alchemist was always going to be a financial sacrifice for Intel to identify its worst weaknesses. Many of those realizations are likely too late for Battlemage, so Celestial is probably be the earliest we can expect real competition.

 

[The Historical Fidelity]

Intel was never going to get great margins from Alchemist simply due to being the new kid on the dGPU block with a 20+ years long history of quirky IGP drivers regarding gaming. Alchemist was always going to be a financial sacrifice for Intel to identify its worst weaknesses. Many of those realizations are likely too late for Battlemage, so Celestial is probably be the earliest we can expect real competition.

Except for their margins will be worse than what they figured was tolerable in 2020 for the 2021 season. In fact now, in order to price their cards for the new gen of AMD and nvidia, they will have to sell each gpu die at a loss. It would be better financially to simply drop alchemist as their shareholders are already pissed off at their returns and focus on battlemage

 

[InvalidError]

In fact now, in order to price their cards for the new gen of AMD and nvidia, they will have to sell each gpu die at a loss. It would be better financially to simply drop alchemist as their shareholders are already pissed off at their returns and focus on battlemage

For Intel to hit the $140 price point on the A380, Intel is likely already selling those at a loss. For Intel to be announcing the A580-770 "Soon", the chips must already be quite far along production if not already shipped to AIBs. At that point, Intel is better off selling the chips it already has for whatever it can get for them than having to buy back stock that has already been sold, compensate AIBs for recalling orders and wasted production prep, and eating all of those additional costs on top of getting $0 for the chips.

Then you have the additional setbacks of not putting the chips out there for developers to play with, not having customers to find and report more bugs to fix drivers before next-gen hardware launches, not having user bug reports filter through the driver developers and help identify hardware flaws in time to fix them before next-gen silicon is finalized instead of requiring more driver work-arounds in the future, etc.

From a development point of view, cancelling Alchemist now would mean Intel won't have the field data it could start gathering today to improve drivers and hardware until Battlemage launches, which means many of the hardware bugs that could have been avoided if found with Alchemist will still be present by the time Celestial's silicon is finalized. I can imagine it setting ARC's entire development back by one extra generation or two until competitive silicon.

 

[The Historical Fidelity]

For Intel to hit the $140 price point on the A380, Intel is likely already selling those at a loss. For Intel to be announcing the A580-770 "Soon", the chips must already be quite far along production if not already shipped to AIBs. At that point, Intel is better off selling the chips it already has for whatever it can get for them than having to buy back stock that has already been sold, compensate AIBs for recalling orders and wasted production prep, and eating all of those additional costs on top of getting $0 for the chips.

Then you have the additional setbacks of not putting the chips out there for developers to play with, not having customers to find and report more bugs to fix drivers before next-gen hardware launches, not having user bug reports filter through the driver developers and help identify hardware flaws in time to fix them before next-gen silicon is finalized instead of requiring more driver work-arounds in the future, etc.

From a development point of view, cancelling Alchemist now would mean Intel won't have the field data it could start gathering today to improve drivers and hardware until Battlemage launches, which means many of the hardware bugs that could have been avoided if found with Alchemist will still be present by the time Celestial's silicon is finalized. I can imagine it setting ARC's entire development back by one extra generation or two until competitive silicon.

I totally agree, they should not abandon it now with production underway, I was talking about when they missed their target date in 2021. Now they are in a rock and a hard place. Potentially having to give away ARC dies to AIB’s as their pcb, coolers, and firmware costs will stay the same. Any way you look at it though, Intel shareholders will not be happy and to cover the losses from their gpu division, the cost of Intel CPU’s may go up further to satiate shareholder expectations that Intel increase their margins.

 

[InvalidError]

the cost of Intel CPU’s may go up further to satiate shareholder expectations that Intel increase their margins.

They can try. Though I suspect tumbling PC sales and AMD's growing market share are going to force Intel's pricing hand. Post-COVID isn't going to be a good time for manufacturers.