German site PC Games Hardware today posted an interview with AMD CEO Dr. Lisa Su. The interview was conducted as part of a roundtable discussion following her CES keynote last week.
Dr. Su addresses questions about the current and future role for Vega for both professionals and consumers, the outlook for Ryzen 3000 and EPYC, ray tracing and FreeSync, AMD’s 2019 product roadmap, and the future of chip design.
Check out the interview over at YouTube or via the player embedded above.
AMD really lucked out with
AMD really lucked out with spare 7nm capacity at TSMC.
Not really since the 7 nm
Not really since the 7 nm process from TSMC didn’t bring any significant power efficiency gain for AMD products…
Actually nVidia did better products than AMD using the 16 nm process from TSMC which lead to question the so called power efficiency gain from the latest nodes from any founder!
IMHO there’s no more significant gain to expect from foundry processes under 14 nm but there’s room for new materials such as silicon-germanium alloyS that can be used for specific purposes (e.g. SiGe20 for desktop processors and SiGe10 for server processors).
And how do you know this
And how do you know this without the power usage benchmarks available on Radeon VII?
And don’t even think of using any of AMD’s TDP specs to estimate power usage as AMD’s TDP is for folks to estimate the required cooling solution to keep the chip from overheating. You do Know that at 7nm and using denser design/layout libraries the Vega 20 Die area is rather smaller and more densely packed with transistors so that’s going to increase the heat density/per-unit-area also and all the heat will have to be transmitted/dissipated over that smaller Vega 20 die/area so TPD will go up/remain high at 7nm.
You have no idea what the power usage on Vega 20/Radeon VII will actually be on TSMC’s 7nm process node.
Most certianly the clocks on Radeon VII are higher so the only way to test to some degree TSMC’s 7nm efficiency gains is going to require clocking Radeon VII/Vega 20 at the same clock rate as Vega 10/64. But Radeon VII has only 60 nCUs enabled while Vega 64 has 64 nCUs enabled.
Folks need to stop using AMD’s TDP metrics for any thing other than estimateing the cooling solution required for that specific DIE/Area(Radeon VII at 7nm) to properly cool the die and waiting for the actual power usage benchmarking(GN/Other sites) before attempting to make any foolish claims like chipman often spews!
TSMC claims that:
“Compared to its 10nm FinFET process, TSMC’s 7nm FinFET features 1.6X logic density, ~20% speed improvement, and ~40% power reduction.” (1)
So that’s just the improvment over TSMC’s 10nm process and probably a good bit more than TSMC’s own 16nm process and GF’s 12nm process. Really TSMC’s 7nm density over even its 10nm process density and Vega 20’s smaller die size is really going to be hot just look at the die slze of Vega 20 relative to Nvidia’s giant DIEs. But Any power usage metrics will have to wait for Radeon VII’s power testing unless someone in the Press wants to purchase an MI50 to test and get some idea!
(1)
“7nm Technology”
https://www.tsmc.com/english/dedicatedFoundry/technology/7nm.htm
Yet another AMD fanboy that
Yet another AMD fanboy that takes as a poor excuse the Thermal Design Power definition to hide the fact that the electrical yield of a processor is equal to Z E R O…
On other words, 100 % of the power consumption is converted to heat like any electric radiator.
Once again, TSMC claims are pure marketing without any R E A L proof!
Well Nvidia uses that TSMC
Well Nvidia uses that TSMC 12nm node for Turing and you find no fault with that. And Nvidia also uses lower density libraries and transistors/cell sizes that use more fins per transistor in order to push the clocks higher but Nvidia’s DIEs are ginormous. So really the TDPs on Nvidia’s dies are lower beacuse there is more die/area in which to dissipate the thermal loads.
And really your constant FUD is known to all. I’m not the only one that smells chizow in your style of FUD casting. If TSMC’s process nodes are so crappy why does Nvidia make use of TSMC’s services.
Really this is not about AMD this is about TSMC’s 7nm node and you are so Butthurt because AMD is using it for Vega 20/Radeon VII so you just have to take a swipe at TSMC! The Very Same TSMC that Nvidia makes use of and you really are an egregious fool there with chizow’s hand up your nether region Mr. chipman-SP!
AMD uses TDP so the GPU gets fitted with the proper cooling solution and until the Power/Benchmarks arrive you are just hallucinating from snorting too much of that green coolaid powder. You see anyone not fanatic about Nvidia as some AMD fanboy but that just outs you more as a fanatic for Nvidia at the cost of all reason. But it’s not working and chipman is Known as a possible SP of chizow with that lust for the FUD and loads of green powder caked around and almost clogging up both nostrils!
How’s that Nvidia Share Value and market CAP lately! Things have gone down more than 50% since OCT 2018 and that crazy coin market will not matter to AMD what with all those Epyc/Naples and Epyc/Rome revenues that will be getting rather larger with each increasing percentage point of Server CPU market share gained.
AMD is sure making use of loads of TSMC’s 7nm including some extra TSMC 7nm capacity that’s come online since the smart phone market reached peak levels and starting falling. Nvidia has only got GPUs mostly and that market is sure sensitive to those fickle miners. And now Nvidia’s stuck with loads of Pascal that are not very competative performance wise against Radeon VII, or even the RTX 2080 that the Radeon VII is targeting.
Just look at chipman hissing at TSMC now! Hissing and foaming at the mouth with green coolaid powder almost clogging up both nostrils. Hissy Fiting away, hissssssss! Hallucinating away also and babbling incoherently!