[ad_1]
The mother of all leaks from Intel just happened. Intel recently hosted a highly publicized "Xe Unleashed" event, in which the GPU leadership presented its Xe methodology to Bob Swan and some other key people (some AIB representatives, such as ASUS, were also present). . from among them thought we should know about it too. We were able to get hold of presentation slides and even images of the teaser! You know the little "e" annoying in Intel Xe? Well, that's the number of GPUs.
Intel Xe Unleashed: e stands for GPUs, the flagship GPU of Xe 2 will be a dual "transparent" GPU, which will arrive on 6/31 next year
Intel Xe's philosophy is that innovation must be on three main fronts: process, microarchitecture and "e". We already know the first two, but 'e' is something that has not yet been implemented successfully. Yes, there were two GPUs, but they all had to swap part of the functionality and never be scaled linearly. Intel's graphics team thinks they have solved this problem. With a completely new architectural approach (Xe) and a software layer (OneAPI) that can seamlessly adapt to an unlimited number of GPUs, he is ready to address the years of neglect suffered by "e" in the industry.
We managed to get hold of 3 slides of the presentation given by Raja Koduri:
This slide is the cornerstone of the Xe philosophy and the great revealer of what e actually means. This also reveals the existence of the X4 graphics processor class, which, as you will see, is only one step in Intel's plan to dominate the GPU market.
They designed the API One to act as an intermediary between the Direct3D layer and the GPU (I am told they also have a Linux solution) and allow the user to evolve seamlessly between multiple GPU. Transparency is the key word here, as many GPUs can work consistently since a single GPU has never been created. According to the presentation presented at the Xe Unleashed event, the GPU will essentially be recording in the form of a large GPU. This will allow it to integrate with applications that may not be able to handle multiple GPUs and retain almost all backward compatibility.
Developers will not have to worry about optimizing their multi-GPU code, the OneAPI takes care of all that. This will also allow the company to exceed the usual lithographic limit of foundry dies, currently between ~ 800 mm and approximately. Why have a matrix of 800 mm when you can have two dies of 600 mm (the smaller the size of the die, the higher the yield) or four of 400 mm? Armed with an API and XE macroarchitecture, Intel plans to upgrade to Octa GPUs by 2024. This roadmap seems to indicate that the first GPU Xe class will be X2.
The provisional timeline of the first class of GPU X2 was also revealed: June 31, 2020. The X4 class will follow at one point in 2021. It seems that Intel plans to add two cores each year. So we should have class X8. by 2024. Assuming that Intel has the downscaling solution, it should be very easy to increase it. The only concern here would be the performance of the package – that Intel should be more than capable of handling, and binning should deal fairly easily with any wastage problem. Neither NVIDIA nor AMD have yet gone the way of MCM and if Intel can really offer this design, the sky is the limit.
Without further delay, here is the sequence of the teaser that our spy managed to take:
This is your first look at the official Intel Xe GPU, or more precisely the Intel X2 GPU. This short but rather spicy trailer gives a lot. The design cradles a carbon fiber aesthetic with blue accents (after what we've said, blue stripes glow in the dark!) And the first reference design will be realized in partnership with ASUS. You can also see very clearly two intake hoses for what appears to be an internal water loop.
My source has said that the map will actually have two modes. A standard mode, which will allow the dual graphics processor to run at a moderate clock speed for most users, and a turbo boost mode, which when connected to the upgrade of the 39; AIO, will allow the user to achieve clock speeds above 2.7 GHz exact) on both GPUs! This is an absolutely incredible feat that allows Intel to reduce the initial cost of its GPU. You can either buy the card with the AIO as a package, or pay less and upgrade later.
I am told that Intel plans to be very price competitive and when asked to indicate that their flagship product would be more affordable than any other partner in the market. This means we are looking at a maximum MSRP of $ 699 for the flagship X2 product. The X2 GPU will be based on the new 4D XPoint memory and will feature the Direct3D 14_2 feature level for hardware. Here are the complete specifications that were discussed during the event:
| Wccftech | Intel Xe GPU 2 |
|---|---|
| process10 | 7 nm (Intel) |
| Architecture | Xe |
| Flow processors | 12288 (6144 x2) |
| Basic clock (Boost) | 1600 MHz (2718 MHz) |
| Memory | 32GB 4D XPoint |
| Bandwidth | 8 TB / s |
| Peak performance | 66.8 TFLOPs |
| Size of the matrix | 600mm2 (x2) |
| GPU type | MCM |
| interconnection | Unknown (based on TSV) |
| Level of functionality | Direct3D 14_2 |
| TDP | 350W |
| Power connectors | 3x 8 pin |
| MSRP expected | $ 699 (to be confirmed) |
| launching | 6/31 2020 |
Update: clean images of Intel Xe 2 graphics processor teaser
[ad_2]
Source link