Please Note: Blog posts are not selected, edited or screened by Seeking Alpha editors.

Sneak Preview Of AMD’s 7nm GPU


The new 7 nm Vega 20 GPU as a standalone will outperform Nvidia’s Tesla V100 by about 30% to 35%.

7 nm Vega 20 is it is not geared for the gaming.

Like AMD’s 7nm Epyc CPU will outperform anything Intel has to offer and will consume considerably less power.

As always thereis a lot of sizzle and not much steak in these sneak preview estimates.

The article by Hassan Mujtaba is a real teaser. Interesting the author shows Lisa Su revealing the top side of the new Epyc 7 nm CPU in her right hand and the top side of the 7 nm Vega 20 GPU, exposing its infinity fabric in her left hand. I believe this picture was taken at Computex last June, but is compliments of

To summarize the article, as always there is a lot of sizzle and not much steak in these sneak preview estimates, but it looks like AMD’s 7nm Epyc CPU will outperform anything Intel has to offer and will consume considerably less power.

The same is true of AMD’s 7 nm Vega 20, except the power consumption is about equal or a little higher Nvidia’s V100 GPU.

AMD Radeon Instinct 7nm Vega 20 GPU Expected To Feature 20.9 TF Compute

The images below are from Intel’s newsroom describing 8th generation that includes AMD’s Vega embedded along with High Bandwidth Memory and some other gizmos that are what I described as nothing more than souped up Core 3,4 and 5s. They have reduced the Core/GPU footprint and are claiming power savings from the shrink. I’m not sure what they have shrunk exactly as the article is not clear.

New Intel Core Processor Combines High-Performance CPU with Custom Discrete Graphics from AMD to Enable Sleeker, Thinner Devices | Intel Newsroom

Lenovo just put on US and several other markets Epyc Embedded ThinkPad E485 (14” laptop) and E585 15” Laptop) both have starting price of $562.00. It is targeted to low end mobile business market. Lenovo will have some high end versions coming out soon. I would think these two models would be excellent for the big box stores especially for up and coming back to school and holiday seasons. NO side by side comparisons with Lenovo Intel 485 and 585 which is already on the market.

ThinkPad E485 AMD

Intel may be facing headwinds in the foreseeable future.

Mostly all if not every CPU Intel has produced within the last 10 years is vulnerable to Meltdown, Spectre and a host of other newly discovered hacks. Intel has patches that in some cases substantially slow the processor down.

In the past several years Intel ventured into the GPU foundry business and tried and failed to compete in the GPU add in board market as well as in the embedded market. The gross failure ultimately forced Intel to partner with AMD in order to compete in GPU embedded market.

Intel failed to deliver 10nm technologies and may go straight for 7nm. 10nm Cannon Lake CPU’s have issues. Production costs are much higher than planned and wafer yields have been poor, making it impossible for a 2018 roll out. Without reducing wafer size and thickness Intel’s processors may still be able to compete in speed and multithreading but not in critical reductions of energy consumption and the resulting heat overproduction. The energy and heat overproduction issues mentioned before knocks Intel out of the driver’s seat in the server, Cloud and AI markets for an undetermined time.

As if Intel didn’t have enough problems, Intel has been suffering with incompetent leadership for many years. All of Intel’s major failures have happened on Brian Krzanich watch. If it hadn’t been for his misbehaviors his head would have been rolling soon anyway.

In the meantime AMD, Intel’s arch rival rolled out new technologies Q1 2017 and has since released 2nd and is getting ready to deploy 3rd generation 7nm iterations of Epyc, Threadripper and Vega, not to mention embedded solutions. All of AMD’s products outperform Intel in core count, speed, multithreading, memory channel bandwidth, energy efficiency, and has built in AI that tells the CPU what the next task will be and sets up the instruction set in advance, all operations work seamlessly using AMD’s proprietary infinity fabric and high bandwidth low latency DRAM, giving the CPU multiple lanes through the CPU and multiple lanes out of the CPU into both memory and the GPU making multiple simultaneous computations possible. This takes multitasking to new levels.

Intel’s best efforts to combat AMD will have to be a CPU next year that will have code permanently burned into the CPU that will patch its vulnerabilities. The CPU will possibly have 28 Cores based on the Zeon Platinum 8180, have competitive multithreading capability and might be able to overclock close to 5.0 GHz. The down side being without a reduction in wafer size and thickness this technology will have limited application in the server, AI and cloud compute market due to the reasons discussed above.

Many are aware of the Intel’s dismal showing at Computex. A spokesperson form Intel gave the following explanation when questioned by Joel Hruska, a writer for “The 28C demo at the keynote is a real product in development targeted at the high end prosumer and enthusiast audience. Intel continues to optimize design and process across its products, and the demo showcased an upcoming Intel product having the capability of 5.0 GHz overclocking across all 28 cores.”

At best Intel may be able to develop a temporary fix until a completely new architectures can be engineered that will open up the bottlenecks in existing technologies and will incorporate superior internal CPU to DRAM data encryption similar to AMD’s. Time to market will most likely be two to three seemly light years away.

In the interim Intel will of course likely offer tamed down versions of the energy gobbling 8180 demoed at Computex or more likely will soup up versions of its consumer grade CPUs in an attempt to prop up its market dominance.

On the bright side Intel has several tricks up its sleeve that may be its saving grace. First Intel partnered with AMD and will produce Intel CPUs embedded with AMD’s Vega GPUs. This will keep Intel’s foot in the market for HPC desktops, laptops and possibly mobile if they can edge the wafer size down a few nm.

Second Intel’s acquisition of Mobileye combined with the Intel/MU 3D X point NAND (Intel’s trade name is Optane) and the Intel/AMD embedded partnership may produce an APU (Accelerated Processing Unit) that is capable of competing with Nvidia’s aspirations to own the autonomous automotive segment. At this point, this is only conjecture on my part, but I think a slightly likely outcome.

In summary, there is no doubt Intel is having its battle of a lifetime for supremacy. Intel is not in any danger of collapse and is working diligently in many areas to bring new products to the market. In the short term, the stock may decline due to lackluster product development.

In the long term Intel looks strong with excellent innovations in the pipeline, especially considering Intel’s exceptional partnerships combined with the acquisition of Mobileye. In my opinion Intel is positioned to be a real technology mover and shaker in the horizon ahead.

Disclosure: I am/we are long AMD.