Intel linking graphics speculation

Fasauceome · April 26, 2019

I saw this thread just now

And thanks to @Himommies the cool tech it was about was dug up, which I thought was pretty cool but it was also fairly obvious that this never came to market. A multi GPU chip along the lines of the DirectX 12 multi GPU magic trickery? Pretty awesome on paper. I looked into the product myself and found an interesting topic about it that sort of cleared up what I was looking for.

https://forums.anandtech.com/threads/what-happened-to-lucids-hydra-chip.297774/

Most people in the thread were saying it was just overhyped, cool in theory but the claimed 100% scaling was clearly not realistic. However there was a post that really struck my interest:

Quote

last i heard they were bought by intel, so my expectation is they are planning something for release around the same time as larrabee

What we know now is that Larrabee resulted in Xeon PHI and someone who went dumpster diving at Intel HQ sold Linus an Intel test GPU as part of their long lasting graphics project. First of all, if it was bought by Intel it explains why it was so delayed getting to market (ba dum tss) but it also got me thinking,

What if this chip is the design they're using for their new Xe graphics cards? Engineers previously hinted that the strategy being used for an Intel graphics card has to do with combining HD graphics chips already in CPUs on their own PCBs, and the revealed Xe name uses the "e" to denote how many compute units the card has. Do you guys think this is what happened to this chip?

Firewrath9 · April 26, 2019

the combing chips thing was april fools... it mentioned DX15 or something.

Fasauceome · April 26, 2019

4 minutes ago, Firewrath9 said:

the combing chips thing was april fools... it mentioned DX15 or something.

I didn't see anything about that, that Anandtech forum discussion was a full year after the product was announced.

LTT had a video about it

It's a super run into the ground April fools joke if it was one.

Results45 · May 16, 2019

On 4/26/2019 at 4:26 PM, fasauceome said:

I saw this thread just now

Spoiler

And thanks to @Himommies the cool tech it was about was dug up, which I thought was pretty cool but it was also fairly obvious that this never came to market. A multi GPU chip along the lines of the DirectX 12 multi GPU magic trickery? Pretty awesome on paper. I looked into the product myself and found an interesting topic about it that sort of cleared up what I was looking for.

https://forums.anandtech.com/threads/what-happened-to-lucids-hydra-chip.297774/

Most people in the thread were saying it was just overhyped, cool in theory but the claimed 100% scaling was clearly not realistic. However there was a post that really struck my interest:

What we know now is that Larrabee resulted in Xeon PHI and someone who went dumpster diving at Intel HQ sold Linus an Intel test GPU as part of their long lasting graphics project.

First of all, if it was bought by Intel it explains why it was so delayed getting to market (ba dum tss) but it also got me thinking,

What if this chip is the design they're using for their new Xe graphics cards?

Spoiler

Engineers previously hinted that the strategy being used for an Intel graphics card has to do with combining HD graphics chips already in CPUs on their own PCBs, and the revealed Xe name uses the "e" to denote how many compute units the card has. Do you guys think this is what happened to this chip?

Digging deeper, here's the story of what happened to LucidLogix:

After releasing their multi-GPU chip and Hydra Engine firmware in 2008/2009, reviewers found that the performance scaling only worked on a few games and even poor power management in the next version (the Lucid Virtu).

Lucid then after dropped development of multi-GPU products altogether and in 2014 began focusing on their "PowerXtend" firmware for smartphones.

The company and all of its IP was acquired by Google for over $44 Million this past July.

Might all that IP have contributed to Google's Stadia Platform? I bet it did -- directly or indirectly.

Sources:

Spoiler

.

Despite Lucid's novel approaches (from multi-GPU to VirtuMVP to PowerXtend) seems like it's demise was because of better execution by larger tech companies.

A good example of this was their "Virtual VSync" technology back in 2012. Sounds familiar? Well both Nvidia and AMD released similar solutions within a year or two after.

It even kinda shows in their ads. They were hilariously bad:

https://www.youtube.com/user/lucidlogixdotcom/videos

And then there's Intel: