Intel Xe Rumors: Development reportedly stagnating and efficiency worse than its competitors
Intel received a lot of coverage on Xe, their enterprise and gaming GPU that should hit the market in 2020. The coverage was mostly due to the fact that several key staff from AMD moved to Intel to take control of the product. However, something is going on.
Last week several key staff members that last year joined the GPU team, left Intel. And now new rumors have surfaced, not particularly good ones. I will say this though, these are rumors, as vague as can be and nothing else.
Intel's first dedicated graphics card Xe should be released June 2020 but was already pushed back to 'late' 2020. Rumors, however, seem to get significant delays and technical problems. Under the direction of chief architect Raja M. Koduri (ex-AMD), the first models based on the in-house 10nm design for the desktop, mobile applications, workstations and data centers should be available by 2020. More recently we have seen announcements forecasted for exascale servers codenamed Ponte Vecchio. Believing current rumors, it seems Intel's advances in the development of Xe GPUs have stalled. And the efficiency of the chips should be well below that of the competition. The information comes from the forum member »wjm47196« of the Chinese website Chiphell.com (via Wccftech). It is not clear where he obtained that information, but some of them are very vague and superficial, others are more detailed. We summarize the key statements in bullet points:
- Work on the Xe GPUs does not seem to be going well
- Ponte Vecchio is unlikely to arrive within the next two years
- Currently, the efficiency of the Xe architecture is relatively low compared to the competition
- For the time being, there are no plans for custom models of Xe graphics cards
- Drivers would need a lot of time to improve
An official statement from Intel is not available at the time of writing the message.
Intel has used Supercomputing 2019 (SC19) to announce a new category of GPUs based on the Xe architecture and was promoting the Xe GPUs as purpose-built for data centre artificial intelligence and high performance computing (HPC) workloads, rather than gaming. The chip giant has named the compute-focused graphics processing unit after a bridge in Florence -- the Ponte Vecchio.
Intel Xe Rumors: Development stagnating and efficiency worse than its competitors - 12/04/2019 09:14 AM
Intel received a lot fo coverage on Xe, their enterprise and gaming GPU that should hit the market in 2020. The coverage was mostly due to the fact that several key staff froM AMD moved to Intel to ta...
Intel Xe Discrete Graphics based on 7nm - has HBM and carries codenamed Ponte Vecchio - 11/14/2019 09:09 AM
Intel has released a bit of info on their Xe line of GPUs, and that info is rather limited. First off, the product will be fabbed at a 7nm node. Secondly, and yeah think of this what you will, it's g...
Intel Xe graphics card spotted in benchmark database (no benchmarks though). - 11/08/2019 09:29 AM
By 2020 Intel's first graphics cards based on the Xe architecture will be released. The first test phase has already been completed, as Intels Robert "Bob" H. Swan mentioned in a statemen...
Intel Xeon E-2200 Processors Launch With Hardware Security Built-in - 11/01/2019 12:26 PM
Intel's Xeon E-2200 processors are available, bringing enhanced security usages with the additional layer of hardware-based security and manageability made possible by Intel Software Guard Extensions...
Raja Koduri hints at June 2020 Intel Xe graphics card launch - 10/07/2019 03:29 PM
For those that have been living under a rock and w onder who the man is, Raja Koduri was the guy behind many Radeon GPUs, and then moved to Intel where he is working on Xe amongst other things. Now, ...
Senior Member
Posts: 2068
Joined: 2017-03-10
I don't think that's how the GPU industry works. AMD gave up on the high-end with Polaris and look how that turned out - lost market share, lose prestige, and dominance by your competitor for half a decade. In this market, if you don't offer a high-end product which can rival your competitor's products, you don't get taken seriously. If Intel is serious about competing in the consumer GPU space, they absolutely must offer a high-end product from the get-go. This is especially pertinent since it's Intel, a company with deep pockets who can take the risks and swallow the costs (this isn't some small, start-up company that absolutely needs the revenue to survive).
Senior Member
Posts: 22075
Joined: 2008-07-14
AMD, through their acquisition of ATI, was already well established in the discrete graphics market. AMD chose to pull themselves out of the enthusiast market. That effectively told people that AMD themselves didn't believe they were capable of competing at that level.
Intel is in a very different position. They need to prove they can actually deliver a product, and support that product properly.
Senior Member
Posts: 2913
Joined: 2010-05-06
Wow I'm shocked at how so many people are underestimating Intel. Despite their manufacturing woes over the years, Intel still manages to post record profits. Intel has also been in the GPU business for two decades, so it's not like they are some n00b. Actually, I believe they are still the largest GPU manufacturer in the World in terms of market share. But the main reason why they are expanding their GPU business has nothing to do with gaming. It's because they see the tremendous success that NVidia has had within the HPC, accelerated comping and data center markets with their GPUs.
Why do you think Intel has been so hellbent on designing wider vectors for their CPUs? We are up to 512 bit wide vectors now, and it will probably be extended to 1024 bit in a few years. It's to compete with NVidia in those crucial and VERY profitable markets. So definitely don't expect Intel's first round of discrete GPUs to compete with NVidia and AMD's best when it comes to gaming. Gaming will be important, but definitely not on the same priority as those other markets.
More than anything else, Intel wants to be a one stop shop for businesses, governments, universities etcetera that need gobs of computational power. Without a strong and capable in house discrete GPU, their customers have to buy GPUs from NVidia or AMD rather than from Intel.
Intel should have done this years ago, because the writing was on the wall when NVidia released Fermi, the first fully programmable GPU. Well I guess there was Larrabee but that obviously didn't turn out well.

Senior Member
Posts: 13253
Joined: 2018-03-21
the mistake you're making is correlating profits with good products.
Senior Member
Posts: 5771
Joined: 2003-09-15
I expect this exact scenario to play out in RL.