News Intel Arc Alchemist: Release Date, Specs, Everything We Know

[InvalidError]

I can't bet on any of these scenarios with confidence... and that feels very odd.

My bet is on: Intel only has 6-7nm crumbs from TSMC to make those DG2 chips on and won't have enough silicon to make a meaningful dent in the current market. DG2/Alchemist will be little more than an open-beta to help smooth things out for DG3/Mage.

 

[TerryLaze]

And again 100% of all Intel initial GPU output could easily land there with short-term benefits to the shareholders.

That could seriously hurt its adoption in the gaming (& custom game console) eco system.

I can't bet on any of these scenarios with confidence... and that feels very odd.

A big chunk of Xe is going to go to OEM systems, there is no two ways about that, intel and OEMs go hand in hand. I basically doubt that intel would even attempt this without having sold most of the cards already beforehand.
So miners would have to buy complete systems or a lot of them are going to end up with gamers.

Also gaming adoption will depend on how close the hardware is to nvidia/amd , if games don't need special optimizations to run ok on Xe because the hardware is similar enough or the drivers do a good enough job of translating everything then it will be automatically adopted.
Even if they end up being slower.
If on the other hand every game needs special optimizations we can forget about it already, it's going to take forever.

 

[InvalidError]

Also gaming adoption will depend on how close the hardware is to nvidia/amd , if games don't need special optimizations to run ok on Xe because the hardware is similar enough or the drivers do a good enough job of translating everything then it will be automatically adopted.

Don't forget price. Few DIY gamers will give Intel the time of day if their GPUs cost about the same for similar or worse performance with a mediocre to passable prior reputation for gaming driver support.

Miners don't care about gaming performance and gaming driver robustness, so they'll make sure to bring the street prices in line with hash rate as long as there is still money to be made.

 

[abufrejoval]

My bet is on: Intel only has 6-7nm crumbs from TSMC to make those DG2 chips on and won't have enough silicon to make a meaningful dent in the current market. DG2/Alchemist will be little more than an open-beta to help smooth things out for DG3/Mage.

Intel only having crumbs would be the reasonable estimate: Intel would follow AMD's approach to build credibility for their GPUs over several generations. That's really tough, when neither Nvidia nor AMD have really been slacking in GPUs, as Intel did in the CPU domain. What's the USP for Intel GPUs? It's mostly around gravity/integration not technology.

But now enter politics. TMSC might want to play the Intel chip designer faction against the Intel foundry and offer the designers significant allocation to diminish and hurt the foundry, which is still ailing from the 10nm desaster. TMSC has little interest in Intel's foundry to recover and they would also be competing on ASML EUV equipment (and all the other stuff you need to make chips).

TMSC has Intel desperate enough to run their leading design tech on their foundry and they surely don't want that to ever change again. If Intel were to be succssful as a 3rd party foundry, that could really hurt TMSC, so they know what to do. Again, I'd say they have little reason to fear Intel's foundry on technology terms, it's the politics that could turn the tide for Intel.

The Intel designers again may see a chance to use their Chipzilla weight and put money on the table, pretty much like Apple did for 5nm.

You can't really afford to go meek and you can't really afford to go big (and err) that often. Middle ground muddle won't get you anywhere either.

 

[C00kieMonster1015]

i just hope that INTEL will have monthly driver support like what NVIDIA and AMD is doing right now..... Driver Support takes time, money, and Manpower to sustain... and intel might abandon it when the next big thing after alchemist.

 

[eye4bear]

When I go in ICC's GPU tab, it shows my GTX1050 as 'Unknown' and no 11400/UHD730 IGP. The IGP does show up in Device Manager with drivers loaded and I can activate display outputs when I plug monitors into the motherboard, so the IGP and drivers are definitely loaded and working, ICC just isn't picking it up.

Another issue with Intel's drivers is that 9805 (still newest at time of writing) fails to register as a newer version than 9078 in Windows Update, so Windows keeps overwriting manually installed newest drivers with antique drivers that precede Rocket Lake's launch by four months every time WU does its thing. Either Intel isn't bothering to update Windows Update or something is screwed up there.

Edit: WU just overwrote my updated drivers with ancient drivers again, Intel's update tool is failing to update drivers even after two attempts at letting it auto-install after reboot. Turned off automatic driver updates in GPEdit and force-updated the drivers by manually picking the INF file from device manager. My Intel IGP driver experience so far is hot garbage.

 

[eye4bear]

Just one reason some of us gave up on Windoze and use Linux where we can totally control the update process. My GPU drivers are only updated when I say they can be.

 

[Rekikz]

It will be interesting to see just how far intel are willing to go on pricing. They are in a position where they desperatley need market share to get developers thinking about how they can optimise/support features for their GPUs, it wouldnt be too surprising if they would sell at cost with a long term strategy in mind.

 

[InvalidError]

It will be interesting to see just how far intel are willing to go on pricing.

Well, pricing got indirectly leaked by Intel revealing how much their "scavenger hunt" prizes are worth. With the top-end prizes listed as worth $700 including some promo materials and Xbox pass, we can expect the top-end SKUs to cost $600-650.

Looks like Intel is aiming for the competition's MSRPs instead of undercutting the market to drive adoption as is normally the case for the under-dog.

Not like they have any reasons to do otherwise while crypto is still on the up-and-up as supply shortages are expected to persist through most of the next year.

 

[TerryLaze]

Looks like Intel is aiming for the competition's MSRPs instead of undercutting the market to drive adoption as is normally the case for the under-dog.

Remains to be seen, depending on the performance they might be overpricing which would be more their style.
At least there is quite some buzz about intel working with game devs making things happen so it should at least work well enough.

 

[JarredWaltonGPU]

Well, pricing got indirectly leaked by Intel revealing how much their "scavenger hunt" prizes are worth. With the top-end prizes listed as worth $700 including some promo materials and Xbox pass, we can expect the top-end SKUs to cost $600-650.

Looks like Intel is aiming for the competition's MSRPs instead of undercutting the market to drive adoption as is normally the case for the under-dog.

Not like they have any reasons to do otherwise while crypto is still on the up-and-up as supply shortages are expected to persist through most of the next year.

FYI, the top prize had a $900 "approximate price" while the first place prize was $700 — plus extras. Best guess: DG2-512 at the top, in more limited quantities, and DG2-448 as the more readily available model.

 

[JarredWaltonGPU]

For anyone wondering, I just did a major refresh of all the information in this article to bring it up to date. (It was last updated in November 2021, so a lot has changed.) Performance data, from Intel's own benchmarks, is now included.

 

[passivecool]

@ intel: Everybody needs a hobby.

 

[egda23]

In the end, all will depend of the price war in the low - mid-level performance categories. Which are the categories most used by peoples with a discrete GPU.
At least according to the reports from Steam, which certainly has many flaws, but at least can analyse a very large number of users.