There are numerous rumours about Intel roaming the news today, from CPUs being discontinued to brand new GPUs. [H]ard|OCP have heard that Kaby Lake X does not have long to live, the end of this year may see the end of that chip and a modified Skylake-X family which have moved to solder under the heatspreader to allow higher frequencies to be reached. The TIM change is not the only interesting thing about the Skylake-X refresh, its TDP could be as high as 300W, based on information from Supermicro's PR.
As far as new products go, the little birds suggest Cascade Lake is unlikely to appear this year, but instead should be expected in early 2019. There are also whispers about Intel's new GPU, Arctic Sound, appearing both on-die and in a discrete GPU designed to accelerate streaming and HPC, with Raja closely involved in the design.
"In this episode of How the Rumor Mill Churns, we address some old Intel CPUs, some new Intel CPUs, and hopefully Intel CPUs that we will never see again. End of Life for good products is often disheartening, but when EOL pertains to something that should have never existed, it goes over a lot better."
Here is some more Tech News from around the web:
- Windows 10 Spring Creators Update delayed after Microsoft finds 'blocking bug' @ The Inquirer
- Gmail is secure. Netflix is secure. Together they're a phishing threat @ The Register
- Dell claims its latest laptops are great for cheating at PUBG @ The Inquirer
- Taiwan speeding up quantum computer development @ DigiTimes
- Patch or ditch Adobe Flash: Exploit on sale, booby-trapped Office docs spotted in the wild @ The Register
- Oneadaptr EVRI 80W USB C charging station – Review @ Bjorn3d
- Wall mounting a Middle Atlantic DWR/CWR server cabinet @ Missing Remote
The big question for
The big question for Intel/Raja is where will Intel be getting its In-House GPU IP licensing from and any Unified shader patents are held by both AMD and Nvidia/others and the two biggest GPU makers have some informal agreement not to go after each others patents in court. So Raja can engineer away but any Clean-Sheet design for any Intel in-house GPU design/tapeout will see Intel still needing to obtain an IP license for Unified Shaders/any other GPU IP that AMD/Nvidia/IT/Others have the patent rights for.
Even Intel’s current Integrated In-House designs have made use of Nvidia’s Patent IP licensing in the past and maybe it will come from AMD, or already is.
Thers is also the Qusetion of any Current Patents for GPU IP that may be ready to expire and that needs to be looked at also.
Raja/Intel can not make use of any Intel/FPGA IP included on the HBM/HBM2 stacks as AMD has a patent application on that currently(Patent for FPGAs on that HBM/HBM2 stacks for On-HBM/HBM2 FPGA compute).
Intel has that FPGA accelerator IP it has just announced on a PCIe card for AI/Other usage and I also think that maybe Intel/AMD/Others should look into FPGA usage for In-Hardware(PCIe card Programmable FPGA) Ray Tracing acceleration of graphics workloads. So if Intel wants to compete with Nvidia maybe that Intel/FPGA IP should be included on the any PCIe card based Intel Discrete Graphics. AMD will have to partner Xilinx/others as Intel now owns Altera now.
Arctic Sound GPUs can not arrive this year or next unless Intel is simply scaling up its current graphis IP with Tweaks from Raja. And even with Raja/large teams their GPU designs take years to design/certify and Vett even for the likes of AMD and Nvidia. GCN and Maxwell took years and both Nvidia and AMD have been using the same basic GPU Micro-Archs with tweaks for several “generations” now and even Volta/Navi are not now or are going to be based on revolutionary new GPU micro-arch designs.
Nvidia just added that Tensor core IP for Volta with Tweaks to the Graphics parts of that design and Navi will be more Modular/Scalable for GPU dies the same way that AMD made Zen/Zeppelin Modular/Scalable for CPU dies. Navi’s Micro-arch(The Graphics parts) will be more like Vega with tweaks but with all that Modular/Scalable die design with Navi being more for easier die/wafer yields and allowing AMD to easily create the whole line of Navi low end, Mainstream, and Flagship, GPU using the Navi Modular/scalable GPU die/chiplets Building blocks.
Ta for sharing the patent
Ta for sharing the patent angle & savvy.
Intuitively, yeah right.
Intel can ~ignore serious graphics until almost 2020, and now at this late stage, airily decide to get into the powerful gpu business.
Sorry mate. For many reasons including patents, you missed the boat.
It seems incredible intel have been so slow to realise and act on the obvious reality that gpu has a vital integral role in modern computer solutions.
They have failed customers and opened a back door to now competitors, amd & nvidia.
AMD got ATI in 2005, & planned it long before.
Well if Raja is involved
Well if Raja is involved should we expect a lot of hype and then just some fizzle once the product releases like he did when he was at AMD. /s
As the other poster stated if Intel is making a totally new design did they have to license IP from AMD or Nvdia to get it done so quickly or are they just expanding upon their tired and old tech?
I also agree with the other poster about AMD’s Navi it will most likely be very close to what they got out now but tweaked and made so it can be combined together with a few GPU chips like the Zen arch is and this will be AMD’s way to get a lot more performance without having to break the bank doing a whole top to bottom new design of their GPU tech. If the Navi’s work as well as the Zen design by combining 2-4 GPU’s together and having the infinity Fabric link them together as one GPU this should get AMD back into the GPU race for a while or at least until Nvidia does the same thing with their GPU tech then we will see Nvidia take the lead again most likely Since their CEO just hates being any where near second best.
“Nvidia take the lead again
“Nvidia take the lead again most likely Since their CEO just hates being any where near second best.”
NVLink is Nvidia’s version for GPUs that is similar to what AMD’s Infinity Fabric does for AMD’s CPUs and GPUs. And Nvidia is only tops because is has a much larger market Cap than AMD and Nvidia can afford 5 different base die tapeouts each new GPU “Generation”, GP100, GP102, GP104, GP106, GP108 compared to at the time AMD’s only one Vega GPU base die tapeout the Vega 10 tapout that had to do double duty as a Professional Compute/AI base die tapout and a gaming Base die tapeout.
GP102 only beats Vega 56/64 in gameing because the GP102 based GTX1080Ti comes with 88 out of GP102’s 96 available ROPs enabled. So the GTX 1080Ti’s 88 ROPs have the higher Gpixel/s fill rates that directly translate into higher FPS.
AMD’s Vega 10 base die offers up 4096 shader cores and way more TFlops of FP performance than even the GTX1080Ti. In Fact the Vega 56 SKU has the same number of shader cores as the GTX 1080TI and the Same Number of TMUs and it’s only because that Vega 10 base die only offers 64 ROPs max that Vega 10 die based Vega 56 is not competative in gameing with the GTX 1080Ti’s 88 ROPs. Vega 64 has way more shader cores than any Nvidia consumer variants and even more than most of Nvidia’s professionl SKUs except Volta/Volta professional variants with GV100 base die which has 5376 Shader cores available.
Gamers appear to judge the Vega GPU micro-arch more on that one Vega 10 base die tapeout that is so shader heavy at the expense of ROPs but that’s not Raja fault that AMD’s management wanted Vega 10 more for compute/AI than for gaming. The Vega GPU micro-arch and its Rapid Packed math and other features are actually very efficient and just look at the Raven Ridge Mobile/Desktop APUs and see that Vega is not a bad GPU Micro-arch at all.
Nvidia wins the gaming contest with money, enough money to hit the ground runnig from the git go with 5 different Pascal base die tapouts some of which are very ROP heavy designs. So Nvidia can pull a rabbit out of its hat at the last second and beat the Vega 10 base(64 ROPs max) die based offerings in that ROP metric that directly drives the FPS metrics higher via thoes higher Gpixel/s fill rates from the GTX1080Ti’s 88 ROPs!
Nvidia’s CEO can not afford to be second best with all the money that is spent, Billions, that those 5 Pascal base die tapeouts cost to develop and AMD entire market cap would be used up just trying to compete with Nvidia’s total R&D investment. Nvidia’s consumer gaming GPU market share dwarfs AMD’s consumer/gaming marketshare.
AMD’s is more Interested in Vega for Compute/AI and the higher pro market markups than any consumer gaming marekt where AMD has such a small market share to begin with. Look at those RX 580X SKUs that AMD just introduced for the OEMs to maker use of, AND has no interest in trying to defeat Nvidia in the Discrete GPU Consumer/Gaming market. AMD needs to get its total gross sales margins above 45% or AMD’s stock prices will forever hover between $8-$15 dollars a share.
Raja’s gone and Lisa Su will not keep her job unless those Epyc and Radeon Pro WX/Radeon Instinct sales take off for the professional markets. It takes 20 Epyc 32 core CPUs/MBs and 80 Vega 10 base die based Radeon Instinct MI25 GPU Compute/AI accelerators to make one of those Project 47 PetaFlop supercomputers in a single cabinet and that’s what AMD is focused on and not any flagship cinsumer gaming crown where AMD can barely break even much less make a 45% margin on its total revenues/sales.
Lisa Su and AMD’s BOD are going after the Pro Epyc/Pro GPU markets first and formost or they are out of a job in the not too distant future!