Although the date was ultimately not what many wanted, it looks like Intel has everything ready to launch this new generation of GPUs under its name. But it does not matter if it is not competitive and not if the price is not in line with said performance and the market. As expected, we are seeing more light than darkness, that’s all we know for now.
Intel is ready, DG2 will be presented at the most important event in the world
At least that is what was published on this occasion. And will Intel present in style what is perhaps the most advanced product of the decade in desktop graphics cards. Its new DG2 graphics card will therefore be presented at CES 2022, based on the Xe HPG design and architecture for gaming, it will rely on several variants to directly compete with AMD and NVIDIA.
At least five starter versions will arrive from this DG2. All of them differ in threads and although there are no names they have already been baptized by these numbers, these being: 96, 128, 256, 384 and 512, which we already knew . What we had no certainty beyond certain rumors, it is the version with 448 EU, which would arrive later and of which we have today interesting information.
Assuming it is rumored that it will include the same lithographic process as its sisters (TSMC at 6nm), the use of the 5nm node ringing behind the scenes is therefore ruled out, so Intel gets an improved process, but not cutting edge. . enough advantage to compete with RDNA 3 and Ada Lovelace (that doesn’t really mean either).
Amps vs RDNA 2 vs Xe, Intel will measure up in the mid and high end
The rumor is clear: this 448 EU DG2 will be a strong competitor for high-end gaming GPUs. With 3584 Shaders, it looks like its performance will be very close to that of the current RTX 3070, while the low-end 128 EU version is said to be somewhat ahead of the GTX 1650.
What can we expect from the more powerful 512 EU version? Okay, if everything is correct (take out the salt shaker and add salt as you like) the frequency of this DG2 512 would be around 2.2 GHz, it would have 16 GB of GDDR6 (no GDDR6X) coupled with 256 bits of tape busy.
In terms of performance and consumption, we are talking about between 220 and 250 watts as TDP and it would be between the RTX 3070 and the RTX 3070 Ti, so it could not compete at the top of the range with either of the two rival companies. It is said that this is not because of less advanced hardware than NVIDIA and AMD, a worse architecture as such, but because of the lack of optimization that current engines have to work with Tiles and that Intel would have problems and delays with their drivers.
In any case the date is approaching, it is time to refine and work with the video game companies, we will see if they finally manage to compete decently on their first try.
