You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
Do we know what's exactly happening on the Intel Alchemist rollout? Intel has been talking about Alchemist's potential a lot except the official release date.
The rumor is that Intel GPU have lots of compatibility problems. They could be fixed by installing patches. However, those patches varies depends on the SKU of CPU. End users will be very mad trying make them work.
Launching now will ruin Intel GPU brand in the long run.
I don’t think there are any 100% credible sources on what’s really going on at Intel with ARC right now. But there are a few things we do know:
- ARC has been delayed several times, it’s launching 3-6 months after early plans
- ARC has some good attributes: Strong Raytracing performance for the class of GPU, hits above it’s weight for hardware encoding, etc.
- DX12 performance is mostly acceptable (as long as you have a modern system with ReBAR support), outside of these conditions performance is fairly inconsistent
- Intel doesn’t have decades of hand-optimizing for DX9, DX10, DX11 games that AMD/Nvidia has. They had to have considered this in their roadmap. (The reason AMD/Nvidia driver packs are several Gigabytes is because…)
- Intel acknowledged working actively on at least 5 generations of ARC at one point. Even mentioning a to be named “E” generation product earlier this year.
- Raja (head of Intel Graphics) has publicly said Intel is committed to ARC after press rumors of cancellation came out
- Intel’s graphics rep (guy taken from Nvidia) said that they will price Alchemist based on the worst class of performance (i.e. non optimized DX9-11 games). This will put pressure on margins but will also establish market share.
- They’ve launched their first low end product in China; which smells like a giant beta test (i.e. 10th gen Cannonlake chips), though they’ve committed to launching more products in the US and Europe soon.
We are also hearing some rumors consistently:
- OEMs have reported confusion on launch dates and availability, and some reports seem to indicate Intel internally is in disagreement with the roadmap timing
- Alleged discussion of cancellation of ARC inside. (Either in entirety or first generation Alchemist)
Based on all of this, with no inside knowledge I suspect there are internal organizational communications “opportunities” for Raja and Pat to address.
I also suspect there’s some kind of friction inside between bean counters, strategists, and the technical staff for how much to invest in ARC going forward. The Q2 loss is certainly going to focus attention on “grow market share at a loss” products (“contra-revenue”). As a final kicker, they planned to launch with RTX 3070+ performance but it now looks like their top product will be around 3060Ti performance while the 4000 series is launching very soon.
Intel certainly should not be throwing in the towel. They've spent billions just getting the pleasure to be able to start building discrete GPU's.
Intel want's in on those thousand of dollars per card. No way intel is going to let that drop.
Of course it was expected to not be so great at the start. Getting drivers up and running takes a lot of effort and time too. Give it a few more years. 2026 maybe we will see something competitive.
Intel certainly should not be throwing in the towel. They've spent billions just getting the pleasure to be able to start building discrete GPU's.
Intel want's in on those thousand of dollars per card. No way intel is going to let that drop.
Of course it was expected to not be so great at the start. Getting drivers up and running takes a lot of effort and time too. Give it a few more years. 2026 maybe we will see something competitive.
The challenge is Intel's financial strength is not as good as before. Will Intel's shareholders and Board of Directors have the patience to wait for another two, three, or four years?
My take is that entering a market as a 3rd competitor is quite challenging if ecosystem is well-developed - it's not just "improved drivers". It's going to need to be a 5 year well-funded haul to achieve parity:
- DX12 performance is mostly acceptable (as long as you have a modern system with ReBAR support), outside of these conditions performance is fairly inconsistent
- Intel doesn’t have decades of hand-optimizing for DX9, DX10, DX11 games that AMD/Nvidia has. They had to have considered this in their roadmap. (The reason AMD/Nvidia driver packs are several Gigabytes is because…)
My take is that entering a market as a 3rd competitor is quite challenging if ecosystem is well-developed - it's not just "improved drivers". It's going to need to be a 5 year well-funded haul to achieve parity:
