Intel is bringing out their graphics card to compete against the graphics giants AMD and Nvidia. The long-awaited Intel Xe Graphics card has entered the testing phase. Intel first teased their GPU in CES 2020, where we saw a slight glimpse of the graphics card. The new Intel Xe GPU is the next step to the future.
Intel’s GPU History: Failure after Failure
Intel had several other prototypes on the works, but only one of them worked properly. Their first Project Auburn didn’t work as well as they thought it would.
The first GPU, the Intel740, or i740, was a decent card. The performance was abysmal, compared to what they advertised. The i740 performed decently in Quake II at a resolution of 800×600, anything above that didn’t work. Even pairing the i740 with a better CPU didn’t give a better gaming experience. Intel did further GPU development, but still, they didn’t perform well, and the final product never made it to the market.
Intel again picked up the GPU project from its ashes of failure and focused on their new GPU engineering known as Project “Larrabee.” The GPU worked well not as a GPU but as a General Compute Processor. The project then became Xeon PHI, powering many supercomputers. Again the promised GPU didn’t even make it to the market. So when Intel announced their new GPU project in 2017, people were naturally skeptical.
So what is new and good about Intel’s new GPU?
The Intel Xe GPU is based on the same Gen 12 graphics on Intel’s onboard GPU solution. The Gen 12 graphics aren’t that strong, but the GPU will be scaled to increase performance. Intel’s Gen 12 and Gen 11 graphics are available in 10th Gen Ice Lake and upcoming Tiger Lake CPU. The integrated GPU in from the 8th generation to 10th generation Core series CPU is ok but not suitable for gaming.
The Xe graphics cards are based on the Ponte Vecchio architecture and use a 10nm+ process node. The Xe-LP will be the entry-level graphics solution for home pc solutions. Whereas the Xe-HP GPU is for high-end systems and applications, and the Xe-HPC is strictly for data centers, etc.
The Entry-level GPU Xe-LP seems to have caught traction, and Intel showed their rendered images and prototype during CES 2020 in the form of DG1 GPU. The Xe GPU might implement ray-tracing in their GPU. AMD will be releasing Ray-Tracing features with their RDNA2 based Navi 21 GPU, whereas Nvidia is looking to release their newer Ampere Graphics card with more Ray-Tracing cores. As the Ray-Tracing feature hit the market and mainstream gaming community, it is more likely from Intel to implement Ray-Tracing in their GPUs.
Leaked Driver Information sheds some light on the GPUs
Intel is still under research, and the Ponte Vecchio is moving to a 7nm process node. A leaked graphics driver showcased some details about the graphics card: iDG2HP128 and other codes listed inside the leaked driver hinting to 128 EUs based graphics cards. EU’s are execution units similar to Nvidia’s Cuda Core or AMD’s Streamline Process.
Intel’s Gen 11 graphics is two times faster than its predecessor. The beefiest Gen 11 graphics have 64 EUs, but Intel claims the Gen 12 graphics will be quicker and more reliable. The upcoming Tiger Lake CPU will feature the new graphics design. The discrete form of the Gen 12 graphics card will be more powerful than the integrated ones.
Intel is going for HBM based memory rather than DDR6. The Head of Intel Graphics Unit and former Radeon Technology Group leader Raja Koduri said the cards might be using HBM2 memory. Using the HBM2 based memory will be expensive, but Intel seems to deny it the claim.
But in a press conference slide, Intel showcased using the HBM2e memory and can implement their 3D chip stacking Foveros technology. The Xe graphics card boasts a TDP of 75W to 500W. The Lower end cards will have a TDP ranging from 75W to 300W, whereas the 500W cards are engineered for data centers.
DG1 Intel’s First Prototype
The DG1 is a GPU prototype from Intel, showcased at CES 2020. The new GPU isn’t that great. Since this is modern architecture and a new GPU, there were some hiccups. The DG1 was initially said to ship with their new Tiger Lake CPUs, but they have produced some samples for their ISVs. The ISVs will then work on DG1, optimizing, and checking its architecture.
The DG1 was seen in works during the Intel Keynote and on Intel’s show floor. The DG1 performed decently in the game title “Warframe.” A professional Esports player or a veteran gamer could tell that the game was playing below 60FPS. But that is just the teaser as the graphics card wasn’t fully optimized. The DG1 was operating using a PCIe 3.0 interface.
Looking at the DG1 displayed on Intel’s show floor, the GPU draws power from the PCIe Lane, meaning it has a low TDP. The DG1 is their first lineup of their Xe architecture GPUs. The Xe architecture has been divided into Xe LP, Xe, and Xe HPC GPUs. Intel stated that the DG1 is not a 3D acceleration solution but a “media and display engines” offering accelerating performance while gaming and video editing work.
Pricing and Availability
The pricing isn’t as transparent as of now. The card’s specification and performance haven’t surfaced online. If Intel intends to compete with the giants of the GPU industry, the pricing will be around 200$ for the entry-level and higher as the GPU performs scales. The datacenter GPU will cost thousands of dollars.
We, as ordinary consumers, are interested in mainstream graphics cards. It all depends on Intel about the pricing of the graphics card as they intend to target more on the higher end market spectrum. As the release date comes closer and closer, we will keep you updated on the topic.
As for the release date, there were hints towards a 2020 release, but the Ponte Vecchio release got pushed to 2021. Intel has showcased its GPU in several events and conferences, pointing to a 2020 release. But no news update has surfaced online. We will keep you updated for further news about Intel GPU after it is announced, so stay tuned.