AMD Ryzen (Summit Ridge) Benchmarks Thread (use new thread)

[TheELF]

Doesn't intel still have to agree to give/share the licence to a second foundry or is this done with?

 

[Doom2pro]

Doesn't intel still have to agree to give/share the licence to a second foundry or is this done with?

With Intel requiring an x64 license from AMD it's at the very least mutually assured destruction.

 

[KTE]

Apple is 100% shifting to AMD...

Sent from HTC 10
(Opinions are own)

 

[TheELF]

With Intel requiring an x64 license from AMD it's at the very least mutually assured destruction.

Not really,intel has EM64T which is pretty much the same thing,it would suck for a while until apps get patched,if there would even be any that would require one,but after that it would be business as usual.
Intel just adopted amd64 to make life easier for them.

 

[scannall]

.
Intel just adopted amd64 to make life easier for them.

More because Microsoft told them to. MS didn't want to support different 64 bit x86's.

 

[nismotigerwvu]

Not really,intel has EM64T which is pretty much the same thing,it would suck for a while until apps get patched,if there would even be any that would require one,but after that it would be business as usual.
Intel just adopted amd64 to make life easier for them.

That is simply a name (not even the current one) of Intel's implementation of x64. They still need a license regardless of what they brand it as for their chips.

 

[zinfamous]

That can only be for the good. If Samsung's offerings perform better, the pressure on GloFlo to improve will be intense.

Yeah. Is it possible to determine which foundry was used based on SN or some other indicator on the box or via an online store?

 

[Doom2pro]

Not really,intel has EM64T which is pretty much the same thing,it would suck for a while until apps get patched,if there would even be any that would require one,but after that it would be business as usual.
Intel just adopted amd64 to make life easier for them.

As others have said, EM64T IS AMD64... It's just Intel's name for AMD's invention.... Other than AMD's contribution, Intel only has Itanium and we all know how that went over.

 

[Doom2pro]

Yeah. Is it possible to determine which foundry was used based on SN or some other indicator on the box or via an online store?

Hopefully, because you can't go on Diffused labels alone as Samsung has a 300mm plant with 14nm capabilities in Austin Texas.

 

[bjt2]

Not really,intel has EM64T which is pretty much the same thing,it would suck for a while until apps get patched,if there would even be any that would require one,but after that it would be business as usual.
Intel just adopted amd64 to make life easier for them.

How nice was intel... They allowed AMD to continue to use unmodified an extenstion THEY invented... They were quite generous...

Actually is INTEL that has a crosslicense agreement with AMD... Linux adopted AMD64 way before windows... INTEL could not disrupt AMD64 code base... Most servers ran Linux, at that time and now. And at that times the AMD market share in servers was not low. What do you think linux users would have done: modify the code base to be able to use INTEL Pentium 4 CPUs to heat server farms or adopt more AMD CPUs in servers?


EDIT: mine obviously was a sarcastic answer. EMT64 and AMD64 are the same thing. INTEL could not call its 64 bit implementation AMD64, you should agree...

 

[The Stilt]

Hopefully, because you can't go on Diffused labels alone as Samsung has a 300mm plant with 14nm capabilities in Austin Texas.

To AMD it is irrelevant if the dies are diffused in Korea or USA, if the manufacturing would be done by Samsung. The assembly is done elsewhere regardless (AMD has CPU assembly sites in China and Malaysia), so using the S2 fab in Austin shouldn't make any difference.

 

[Doom2pro]

To AMD it is irrelevant if the dies are diffused in Korea or USA, if the manufacturing would be done by Samsung. The assembly is done elsewhere regardless (AMD has CPU assembly sites in China and Malaysia), so using the S2 fab in Austin shouldn't make any difference.

I was talking about identifying where the CPU was fabbed, you can't go on Diffused in USA to determine it was Global Foundries Fab 8 in Malta NY because there is a 14nm Samsung fab in Austin Texas (S2).

Furthermore, it's Samsung's discretion where they fab AMDs products, either the USA Fab in Austin (S2) or the Korean one (S1) or both, it's their business and they have other customers to satisfy in addition to their own needs. You can't just blindly assume they will only Fab AMDs products at the Korea fab (which I assume you are suggesting here), that would be ill advised.

The only way to be sure a Ryzen was fabbed at Samsung is if it says Diffused in Korea, but Diffused in USA could mean either Malta NY or Austin TX... Anyway I thought my point was pretty obvious :\

 

[DrMrLordX]

GLOFO has a record of messing up and will probably continue to be, otherwise they won't use samsung's 14nm tech.

GF may not be such a bum steer in the future. Remember that they are doing all the fab work for IBM, and that they have yet to fail in that capacity (granted, thus far, all they have done is continue work on 22nm SOI for POWER8; POWER9 hasn't launched yet but there is no sign that POWER9 will be late due to poor yields on 14nm HP).

The A9 thing was a black eye on GF, but their work to improve 32nm SOI and 28nm bulk/SHP and SPP was actually pretty impressive. They are at least doing well maintaining/improving on existing processes. How well they execute with Summit Ridge may tell us a lot about GF's future.

 

[Doom2pro]

GF may not be such a bum steer in the future. Remember that they are doing all the fab work for IBM, and that they have yet to fail in that capacity (granted, thus far, all they have done is continue work on 22nm SOI for POWER8; POWER9 hasn't launched yet but there is no sign that POWER9 will be late due to poor yields on 14nm HP).

The A9 thing was a black eye on GF, but their work to improve 32nm SOI and 28nm bulk/SHP and SPP was actually pretty impressive. They are at least doing well maintaining/improving on existing processes. How well they execute with Summit Ridge may tell us a lot about GF's future.

It's likely GF's issues in the past were poor IP... Now they own IBM's IP, their fabs and their talent.

 

[Doom2pro]

As posted by Abwx on SA: Possible new Ryzen Benchmark (Looks like another A0, June stepping)

http://ranker.sisoftware.net/show_r...d5e3d5e3d1e2d0f684b989afcaaf92a284f7cafa&l=fr

 

[Sweepr]

As posted by Abwx on SA: Possible new Ryzen Benchmark (Looks like another A0, June stepping)

http://ranker.sisoftware.net/show_r...d5e3d5e3d1e2d0f684b989afcaaf92a284f7cafa&l=fr

This was posted by a japanese website days ago, including a comparison:

 

[The Stilt]

Hard to say if that's chip has Turbo enabled, since the turbo isn't really visible to OS in Zen.

 

[itsmydamnation]

Hard to say if that's chip has Turbo enabled, since the turbo isn't really visible to OS in Zen.

Are you able to query a register or something to get the current speed?

 

[The Stilt]

Are you able to query a register or something to get the current speed?

They can be read from the standard registers, but Windows doesn't use them to determine the frequency (TSC / HPET only).

 

[iBoMbY]

AMD Arica seems to be an Asus board btw., and the sample OPN seems to be identical to the Canard one.

 

[BeepBeep2]

That 6900K reads a 4.2 GHz turbo clock?
Performance is improved greatly in Sandra Multimedia for Bdw-E vs. Hsw-E

 

[Abwx]

That 6900K reads a 4.2 GHz turbo clock?

That s what some people call a "comparison", a 6950X/4GHz yield the same score, so that s way more than 4.2GHz, the 3.21GHz displayed on the pic is just fraudulous to say the least, but since it please a certain public why not...

 

[superstition]

That can only be for the good. If Samsung's offerings perform better, the pressure on GloFlo to improve will be intense.

It could be an explosive launch where the product could catch fire in the public's imagination.

 

[rtsurfer]

Lmao

Sent from my Nexus 5X using Tapatalk

 

[KTE]

GF may not be such a bum steer in the future. Remember that they are doing all the fab work for IBM, and that they have yet to fail in that capacity (granted, thus far, all they have done is continue work on 22nm SOI for POWER8; POWER9 hasn't launched yet but there is no sign that POWER9 will be late due to poor yields on 14nm HP).

The A9 thing was a black eye on GF, but their work to improve 32nm SOI and 28nm bulk/SHP and SPP was actually pretty impressive. They are at least doing well maintaining/improving on existing processes. How well they execute with Summit Ridge may tell us a lot about GF's future.

IBM still has a huge hand in developing their HP process at Albany. Even in the acquired departments, it's the same old IBMers having moved to GloFo. That is exactly why this is supposed to be a great move for GloFo.

The 14nm LP was a Samsung process, not a GloFo one. Again, workers on it from the GloFo side are ex-nanotech IBMers.

Gary Patton, CTO/SVP of R&D in May 2016:

For the first time, GloFo now reckons it can take industry process leadership using a proprietary, in-house developed process technology , and this is thanks to the IBM acquisition.

“The IBM acquisition gave us people experienced in leading edge development,” said Patton, ” the people who developed 45nm, 32nm, 22nm and 14nm are the same people who are working on 7nm.”

At the last node before the IBM acquisition – 14nm – GloFo licensed its process technology from Samsung.

“Malta (one of GLoFo’s New York fabs) is high yield on 14nm, that gives me a very solid baseline to do the next node,” said Patton.

As well as IBM’s process experts GloFo is also benefiting from the Albany Nanotechnology Centre who his still run by IBM but which is currently “focussing on Malta” says Patton.

Under the terms of the IBM-GloFo deal 50% of Albany’s effort is going to support Malta with the other 50% pursuing pathfinding.

The high performance 14nm process at Malta is used for IBM server chips using partially depleted SOI wafers. There is also an LPP 14nm finfet process at Malta using bulk CMOS.

http://www.electronicsweekly.com/ne...oundries-goes-for-process-leadership-2016-05/

At the Samsung agreement:

The new collaboration will leverage the companies’ worldwide leading-edge semiconductor manufacturing capabilities, with volume production at Samsung’s fabs in Hwaseong, Korea and Austin, Texas, as well as GLOBALFOUNDRIES’ fab in Saratoga, New York.

Developed by Samsung and licensed to GLOBALFOUNDRIES, the 14nm FinFET process is based on a technology platform that has already gained traction as the leading choice for high-volume, power-efficient system-on-chip (SoC) designs. The platform taps the benefits of three-dimensional, fully depleted FinFET transistors to overcome the limitations of planar transistor technology, enabling up to 20 percent higher speed, 35 percent less power and 15 percent area scaling over industry 20nm planar technology.

BTW what happened to GloFos 14XM process?

As you can see, Malta was already high yield in May.

Sent from HTC 10
(Opinions are own)