It’s no secret that GPU manufacturers (have to) rely on board partners in addition to the so-called reference design and the production of their own graphics cards, in order to be able to exist on the market with a broad impact, and you can see it very clearly with the motherboards that you need the variety of custom designs as a manufacturer in the end, in order to be able to pick up as many customers as possible. Not only the actual features, but also the appearance and brand affinity play a very important role.
I already wrote over a year ago that Intel had contacted various manufacturers regarding the GPUs (a visit to MSI also played a role), whereby the possible cooperation with many well-known names in the graphics card industry could prove to be rather difficult. The biggest obstacle here should be the benefits of exclusive partnerships with the AIC (NVIDIA) and AIB (AMD), as many of these manufacturers, who have already been on the market for years, have entered into tough exclusivity agreements with one of the two major GPU manufacturers, which they obviously don’t want to risk or put at risk.
Such an exclusive partnership has advantages and of course disadvantages, because times are constantly changing and it is of course also immensely important which of the two previous GPU providers was ahead in the customer’s favor. The biggest disadvantage is when you can only play second fiddle for years because your GPU partner is behind in performance, of all things. On the other hand, the benefits also include preferential supply of GPU RAM kits, as well as better technical support from the local FAE and the R&D departments of the respective GPU manufacturers. I’m just reminding here of NVIDIA’s Greenlight program and the thumbscrews present daily to keep partners exactly on track. Something like this leaves no room for blue experimentation.
However, the big conglomerates like Asus, MSI and Gigabyte (as well as ASRock in parts for CPUs) are an exception, e.g. they support several manufacturers so far, because they can afford to do without exclusive contracts due to their size and market positioning. With ASRock, for example, I never understood why they got involved with only one of the two GPU manufacturers, but since ASRock jumped on the graphics card business so late, gaining experience and a certain focus on what was feasible probably also played a role here. This means that it is perhaps not impossible that the trio of potential Intel partners reported by the Chinese media could quickly become a quartet.
Raja Koduri, Senior Vice President and General Manager of the Accelerated Computing Systems and Graphics (AXG) Group at Intel Corporation (what a long title) had recently expressed himself rather unclear in an interview with Chinese media, who exactly (and in which form) would be a potential partner for custom designs, but it will probably come down to the above-mentioned three (or four) major manufacturers, while the classic graphics card manufacturers will certainly not be on board.
That companies like Asus and also MSI want to secure themselves a piece of Intel’s new graphics cake can almost be seen as confirmed, for Gigabyte it will probably also play a role to what extent the slightly stricken company can reposition itself and also contribute financially at all. If you ask in the classic strongholds of NVIDIA and AMD, you will only get a shake of the head and a twitch of the switch, especially since you may not really trust the whole thing yet. In these companies, you have way too much to risk in switching providers for it to really be worth the risk. And the first works reported don’t look that turgid either, so that one would throw one’s own philosophy out of the window because of the euphoria.
Interestingly, one hears behind the hand again and again the argument that one cannot estimate the time and perseverance with Intel’s foray into the world of end user graphics cards, which can of course decide the existence of one’s own company with such specialized manufacturers. If Intel suddenly loses interest after one or more attempts, you would be left with nothing.
Only three GPU variants at launch?
Intel’s ARC Alchemist graphics card rumors currently point to just three GPUs, which are said to target the high-end sector, the upper mid-range, and the and entry-level sector of the gaming market. Also here I had already reported a long time ago that the complete expansion with 512 EU is set (SKU1), just as the salvage variant of the large chip with then 384 EU (SKU2). More interesting should be the SKU5, which with 128 EU could round off the somewhat thin portfolio at least downwards. As my article from May 2021 certainly offers more interesting content and whoever is interested is recommended to read it:
Intel Xe-HPG based DG2 with 5 different chips – new slides and confirmations
If we remember back, I had already suspected a TDP of about 275 watts for the top model back then, but this could not be confirmed so far. But rumors are growing that this is exactly what could happen. This would at least put it on par with the GeForce RTX 3070 Ti, which is judged to be rather inefficient, and which would probably be attacked by the biggest Intel model in terms of performance. Such a constellation is certainly not really optimal, but in times where everything sells that somehow has a plug, that’s enough for a start. So let’s be surprised.