AMD CPU speculation... and expert conjecture

[8350rocks]

The CPU clocks don't have to be limited to 2.1/2.9Ghz. Glofo claims in his website that its High performance 28nm bulk can go above 3 Ghz. I also read in some part that Glofo claimed that they had perfected the bulk process closing the gap with SOI.

Ok, I scoured high and low and found ONE bulk chip made by TSMC that broke 3.5Ghz. It's the Sparc T5 which has a max clock of 3.6Ghz. If GF can catch up and match that's not too too bad, but still a far cry from the 4.4Ghz turbo of the A10-6800K. That's about 20% slower which would eat away a sizable chunk of a 30% performance gain (being optimistic) for SR.

http://en.wikipedia.org/wiki/SPARC_T5

Interesting. I wait a reduction on clocks compared to Richland, otherwise I am unable to explain the official 1050 GFLOPS claimed by AMD for Kaveri.

up the gpu clocks and down the cpu clocks.

I was referring to the CPU; yes, the GPU would be clocked higher than Richland. I wait ~900 MHz.

90% of that 1TF comes from the gpu.

I wait 921.6 GFLOPs for the GPU, i.e. 88%

That doesn't help me with my compiling or any of my other heavy stuff...sorry...that APU is not for me with less than 6 cores and/or on bulk process.

 

[hcl123]

Therefore, as I said FD-SOI 28nm isn't ready, contrary to claims made here by others.

Kaveri presentation at Computex and samples used by mobo makers have to be bulk. What consequences has, if any, that Gigabyte released a kaveri mobo using a bulk model for testing if the final Kaveri is FD-SOI?

Well it depends on what you mean by ready. STMicro is already in full ramp up *production* (not tests) of their ARM eQuad on 28nm FD-SOI. Products based on ST/ST-Ericsson "entourage" most probably (almost 100% sure) will be here before cristmas.

Glofo is a little late, since the tech is based on STMicro, and they have to adapt. But if FD-SOI is late as in a general idea, than that contradicts the general plan that is to be first at 14nm than even intel (edt) big LOL ... http://www.advancedsubstratenews.com/2013/06/which-will-hit-the-14nm-jackpot-first-fd-soi-or-finfet-gauntlet-down-race-on-2/

Everyone on the side of the EU is pushing this very hard, the billions of intel are more than balanced
http://www.advancedsubstratenews.com/2013/08/fd-soi-gets-big-boost-from-nano2017-program-and-impresses-arm-ceo/

Everything is ready, all the big wafer suppliers ( one pertinent critic of intel against FD-SOI) are on board.
http://www.advancedsubstratenews.com/2013/07/ready-for-fd-soi-says-worlds-largest-silicon-wafer-company/

We don't know for sure what the kaveri at computex was made of... probably bulk... but that has none consequence for FD-SOI. Kaveri could be all bulk, or only the first SKUs, the mobile parts, being the "desktop" parts (and some mobile for high end notebooks) FD-SOI... or all be already FD-SOI (not likely if they are to be here at cristmas time -> not late AMD stated).

Its so easy to transition from bulk to FD-SOI that all scenarios are perfectly possible, specially if the base bulk design to be ported is from the same fab, which is the case.

 

[juanrga]

Interesting. I wait a reduction on clocks compared to Richland, otherwise I am unable to explain the official 1050 GFLOPS claimed by AMD for Kaveri.

up the gpu clocks and down the cpu clocks.

I was referring to the CPU; yes, the GPU would be clocked higher than Richland. I wait ~900 MHz.

90% of that 1TF comes from the gpu.

I wait 921.6 GFLOPs for the GPU, i.e. 88%

That doesn't help me with my compiling or any of my other heavy stuff...sorry...that APU is not for me with less than 6 cores and/or on bulk process.

I predict the CPU will have 128 GFLOP, so that 128 + 921.6 = 1049,6, which AMD rounds to 1050 GFLOP. Guess what? Richland CPU is 131.2 GFLOP; therefore a part of the Steamroller improvement over Piledriver will be lost due to lower clocks.

There is no 6-core Steamroller in any known roadmap, public or leaked. Top server parts are 4-core Steamroller. Top desktop parts are 4-core Steamroller. AMD will be selling 2/4-cores Steamroller products. Again if anyone has new info she/he can share with us.

I am leaving open the possibility that Kaveri is finally SOI, but I maintain in mind that AMD said at CES 2013 that Kaveri is bulk.

Therefore, as I said FD-SOI 28nm isn't ready, contrary to claims made here by others.

Kaveri presentation at Computex and samples used by mobo makers have to be bulk. What consequences has, if any, that Gigabyte released a kaveri mobo using a bulk model for testing if the final Kaveri is FD-SOI?

Well it depends on what you mean by ready. STMicro is already in full ramp up *production* (not tests) of their ARM eQuad on 28nm FD-SOI. Products based on ST/ST-Ericsson "entourage" most probably (almost 100% sure) will be here before cristmas.

Glofo is a little late, since the tech is based on STMicro, and they have to adapt. But if FD-SOI is late as in a general idea, than that contradicts the general plan that is to be first at 14nm than even intel (edt) big LOL ... http://www.advancedsubstratenews.com/2013/06/which-will-hit-the-14nm-jackpot-first-fd-soi-or-finfet-gauntlet-down-race-on-2/

Everyone on the side of the EU is pushing this very hard, the billions of intel are more than balanced
http://www.advancedsubstratenews.com/2013/08/fd-soi-gets-big-boost-from-nano2017-program-and-impresses-arm-ceo/

Everything is ready, all the big wafer suppliers ( one pertinent critic of intel against FD-SOI) are on board.
http://www.advancedsubstratenews.com/2013/07/ready-for-fd-soi-says-worlds-largest-silicon-wafer-company/

We don't know for sure what the kaveri at computex was made of... probably bulk... but that has none consequence for FD-SOI. Kaveri could be all bulk, or only the first SKUs, the mobile parts, being the "desktop" parts (and some mobile for high end notebooks) FD-SOI... or all be already FD-SOI (not likely if they are to be here at cristmas time -> not late AMD stated).

Its so easy to transition from bulk to FD-SOI that all scenarios are perfectly possible, specially if the base bulk design to be ported is from the same fab, which is the case.

I did mean "ready at Glofo". Glofo is not still at risk production. They hope to be risk producing in Fab 8 at the end of 2013.

Kaveri desktop comes first. Kaveri mobile comes latter.

 

[de5_Roy]

XBox One’s sound block is much more than audio
http://semiaccurate.com/2013/09/03/xbox-ones-sound-block-is-much-more-than-audio/
last part of in depth look into xbone(R)'s hardware.

 

[gamerk316]

Notice, they say "support"...again...anyone developing specifically for consoles would be an idiot to use a less optimal physics engine. The only edge case I could see would be a PC game that was later ported to consoles, but even then...that doesn't make a lot of sense.

PhysX is the ONLY physics engine that does a good job at multi-object dynamics; Havok falls flat on its face once you start to interact with more then one or two objects at a time. That's why PhysX is typically used for a lot of effects that involve lots of small objects [particles, wind, etc].

 

[hcl123]

XBox One’s sound block is much more than audio
http://semiaccurate.com/2013/09/03/xbox-ones-sound-block-is-much-more-than-audio/
last part of in depth look into xbone(R)'s hardware.

Not that bad analyses, but i think some imprecisions are remarkable

As example the ESRAM, tough being kind of an "extension" of GPU DRAM, has a direct link to the CPU block

http://semiaccurate.com/assets/uploads/2013/08/XBox_One_SoC_diagram.jpg
(what is that black narrow line from bottom (ESRAM) to top (CPU) block ?)

In the general conclusion i think Charlei hinted/suspected right... coherency is prevalent pervasive... but seems "memory coherency" not cache coherency, the "unified" view is "virtual memory" not physical memory, as was always the case since the HSA first hints in 2011.

http://mygaming.co.za/news/wp-content/uploads/2013/05/AMD-Kaveri-hUMA-shared-memory-600x337.jpg

In that strong suspicion, is even more to suspect XBO is hUMA to the core, and being so, is intrinsically dependent of the OS for these hUMA ability (as all does), and so needs a "runtime/*driver*" for this virtual memory capability, since the main and only real VM (virtual memory) mechanism is in the OS. And this "virtual view" would be applicable with everything tightly coupled together in a single chip and single DRAM pool, or if everything was in discrete parts.

Having 15 "processors" offloading CPU and GPU, also means that the GPU is free to concentrate more on graphics, and for this processors the definition of co-processors is blurred, as all systems under hUMA/HSA will be, it will be easier and more flexible and its not exactly ccNUMA on the cache/memory physical front.

Though the difference is too great to compensate compared with PS4 for graphics, the difference in performance should not be that big or notorious as charlei suggests (game dependent). At least XBO seems a "compute" monster (physics and effects should be its domain), not only an AV very elaborated chip, apparently quite better in this department (compute) than PS4... lets see if games take advantage of that.

 

[jdwii]

XBox One’s sound block is much more than audio
http://semiaccurate.com/2013/09/03/xbox-ones-sound-block-is-much-more-than-audio/
last part of in depth look into xbone(R)'s hardware.

Why are so many talking about the crappy xbox it sucks compared to the PS4 which is cheaper and more powerful, even if we include HSA. And exclude the crappy 32mb of embedded ram on the xbox(which barely has more bandwidth than the PS4 actually 8GB of gddr5 memory and for graphics bandwidth over latency anyday).

 

[hcl123]

Perhaps Sony should make a similar presentation... all wold be more clear (sometimes appearances can be deceiving).

UPDATE:

I did mean "ready at Glofo". Glofo is not still at risk production. They hope to be risk producing in Fab 8 at the end of 2013.

Kaveri desktop comes first. Kaveri mobile comes latter.

Risk production supposedly 2H 2013, or more precisely 3Q... which should be about now...
[ EDIT: i don't remember were it was hinted, but i think risk happened already beginning of August, and if everything follows the usual, one more step(3 months), pre-production/production by November, another step to iron out everything (more 3 months) and full production ready by Feb/Mar14, and on the streets before summer 14... perhaps in here the broooaa of late... ]


Kaveri desktop first only if they reversed it... since Llano mobile has been first, matter of fact Richland desktop is only a couple of months old, while mobile was since March... kind of makes no sense to have kaveri desktop now by cristmas.

 

[hcl123]

now!.. if true this completely re-shuffles the deck

[RUMOR]Microsoft Xbox One Could Possibly Feature a Powerful Discrete GPU Core Stacked Inside the APU

http://wccftech.com/rumormicrosoft-xbox-possibly-feature-discrete-gpu-core-stacked-main-soc/

http://misterxmedia.livejournal.com/

"Stacked" i don't believe it... but a kind of "mobile" similar GPU (768sp) part at 853Mhz in Xfire config is possible ... MSFT hasn't disclosed nothing about their "move" engines inside the "Main SoC" either, and for those that didn't pay attention, one of the best possibilities that hUMA/HSA has besides easier programing, is that it makes a pretty good seamless Xfire (half) protocol only by itself.

But of course PS4 could have it to( its also hUMA/HSA in large extent)... only doubt seriously it will be only ~$400 if it does.

 

[Cazalan]

[RUMOR]Microsoft Xbox One Could Possibly Feature a Powerful Discrete GPU Core Stacked Inside the APU

There may be a separate video decode unit for simple things like DVD playback in low power mode. Those are rather common and cheap/small cores.

Stacked die has proven quite challenging beyond just memory so far. AMD, NVidia and Intel all have that on their roadmap for discrete GPUs. Intel is adding it in their socketed Phi part.

http://sites.amd.com/us/Documents/TFE2011_006HYN.pdf

AMD slides on why they have no choice but to stack DRAM. Probably not until 2015/2016 though.

 

[esrever]

XBox One’s sound block is much more than audio
http://semiaccurate.com/2013/09/03/xbox-ones-sound-block-is-much-more-than-audio/
last part of in depth look into xbone(R)'s hardware.

Why are so many talking about the crappy xbox it sucks compared to the PS4 which is cheaper and more powerful, even if we include HSA. And exclude the crappy 32mb of embedded ram on the xbox(which barely has more bandwidth than the PS4 actually 8GB of gddr5 memory and for graphics bandwidth over latency anyday).

Why would you need so much bandwidth with such low shader units? The ESRAM is much more powerful than just a simple GDDR5 bus. The possibilities that devs can use it for will be quite cool. Currently it looks like the xbox will have a better cpu and multimedia processing and the ps4 will have more gpu processing.

 

[esrever]

now!.. if true this completely re-shuffles the deck

[RUMOR]Microsoft Xbox One Could Possibly Feature a Powerful Discrete GPU Core Stacked Inside the APU

http://wccftech.com/rumormicrosoft-xbox-possibly-feature-discrete-gpu-core-stacked-main-soc/

http://misterxmedia.livejournal.com/

"Stacked" i don't believe it... but a kind of "mobile" similar GPU (768sp) part at 853Mhz in Xfire config is possible ... MSFT hasn't disclosed nothing about their "move" engines inside the "Main SoC" either, and for those that didn't pay attention, one of the best possibilities that hUMA/HSA has besides easier programing, is that it makes a pretty good seamless Xfire (half) protocol only by itself.

But of course PS4 could have it to( its also hUMA/HSA in large extent)... only doubt seriously it will be only ~$400 if it does.

all of that is BS from both sites

 

[juanrga]

I did mean "ready at Glofo". Glofo is not still at risk production. They hope to be risk producing in Fab 8 at the end of 2013.

Kaveri desktop comes first. Kaveri mobile comes latter.

Risk production supposedly 2H 2013, or more precisely 3Q... which should be about now...
[ EDIT: i don't remember were it was hinted, but i think risk happened already beginning of August, and if everything follows the usual, one more step(3 months), pre-production/production by November, another step to iron out everything (more 3 months) and full production ready by Feb/Mar14, and on the streets before summer 14... perhaps in here the broooaa of late... ]


Kaveri desktop first only if they reversed it... since Llano mobile has been first, matter of fact Richland desktop is only a couple of months old, while mobile was since March... kind of makes no sense to have kaveri desktop now by cristmas.

According to GLofo 28 nm FD-SOI will not be ready for risk production until 4Q 2013 aka "not ready still"

According to AMD kaveri desktop comes first then mobile.

 

[juanrga]

Notice, they say "support"...again...anyone developing specifically for consoles would be an idiot to use a less optimal physics engine. The only edge case I could see would be a PC game that was later ported to consoles, but even then...that doesn't make a lot of sense.

PhysX is the ONLY physics engine that does a good job at multi-object dynamics; Havok falls flat on its face once you start to interact with more then one or two objects at a time. That's why PhysX is typically used for a lot of effects that involve lots of small objects [particles, wind, etc].

Just by curiosity I searched a little and there is ~470 games using PhysX, including titles such as Batman and Metro Last Light. Engines such as the Unreal engine 4 use PhysX.

 

[gamerk316]

Notice, they say "support"...again...anyone developing specifically for consoles would be an idiot to use a less optimal physics engine. The only edge case I could see would be a PC game that was later ported to consoles, but even then...that doesn't make a lot of sense.

PhysX is the ONLY physics engine that does a good job at multi-object dynamics; Havok falls flat on its face once you start to interact with more then one or two objects at a time. That's why PhysX is typically used for a lot of effects that involve lots of small objects [particles, wind, etc].

Just by curiosity I searched a little and there is ~470 games using PhysX, including titles such as Batman and Metro Last Light. Engines such as the Unreal engine 4 use PhysX.

There's two different types: The CPU bound PhysX, which is more or less your basic run of the mill physics engine. The other is the HW accelerated PhysX, which is run on the GPU. People tend to complain about the crappy CPU performance of this model, blaming it on various factors (initially written in X87 prior to version 1.2, etc), rather then the fact that CPU's SUCK at this type of work.

I'll say it again: Multi-object dynamic physics requires its own dedicated co-processor. Once you get to about a hundred objects, its more complex to compute then even rendering. That's why we still have excessive amounts of ragdoll physics and REALLY simple physics implementations: Because its too hard to do on the CPU, and the GPU is busy rendering. That's why I love how much more powerful iGPU's are getting, as this opens the door to using them as co-processors, driving better physics designs in games.

 

[gamerk316]

SK Hynix Inc, the world’s second-largest maker of computer-memory chips, suspended operations at a factory in China after a fire.

The company is ascertaining if there are any casualties and investigating the cause of the fire, which occurred at a plant in Wuxi, China, Son Hee Young, a spokesman of Icheon, South Korea-based SK Hynix, said in an telephone interview. The fire, which started at 3:50 p.m. local time during deployment of equipment, was extinguished by 5:20 p.m., the company said in a statement today.

The factory makes dynamic random-access memory chips for mobile phones and personal computers.

SK Hynix, which supplies chips to global smartphone manufacturers including Apple Inc., said in July there was limited second-half supply growth for memory chips. Lower-end smartphones will drive demand for dynamic random-access memory, or DRAM, chips in the second half, it had said.

Hynix shares rose 0.7 percent to 28,650 won at the close in Seoul. The stock has gained 11 percent this year, compared with a 3.2 percent decline in the benchmark Kospi index.

Expect price hikes on everything with installed RAM (GPU's, for instance) in the near future.

 

[de5_Roy]

desktop ddr3 ram prices are already on the rise. this'll only make it faster....

 

[Cazalan]

That memory cube stuff is getting closer.

http://www.techpowerup.com/190131/altera-and-micron-lead-industry-with-fpga-hybrid-memory-cube-interoperability.html

 

[de5_Roy]

Microsoft Xbox One Enters Mass Production, Receives CPU Frequency Boost.
Microsoft Further Speeds Up Xbox One’s System-on-Chip
http://www.xbitlabs.com/news/multimedia/display/20130904112824_Microsoft_Xbox_One_Enters_Mass_Production_Receives_CPU_Frequency_Boost.html
1.75 ghz..

AMD to ship 64-bit ARM chips to server makers next year
http://www.pcworld.com/article/2048069/amd-to-ship-64bit-arm-chips-to-server-makers-next-year.html

 

[blackkstar]

Notice, they say "support"...again...anyone developing specifically for consoles would be an idiot to use a less optimal physics engine. The only edge case I could see would be a PC game that was later ported to consoles, but even then...that doesn't make a lot of sense.

PhysX is the ONLY physics engine that does a good job at multi-object dynamics; Havok falls flat on its face once you start to interact with more then one or two objects at a time. That's why PhysX is typically used for a lot of effects that involve lots of small objects [particles, wind, etc].

http://www.youtube.com/watch?v=143k1fqPukk

Bullet would like to have a word with you.

If you have to choose between PhysX on CPU and Bullet on GPU, which do you choose?

Also, why are you guys getting so worked up over 28nm FD-SOI when IBM has 22nm SOI?

http://www.advancedsubstratenews.com/2013/08/pcmags-michael-miller-called-ibms-22nm-soi-power8-the-most-fascinating-of-the-high-end-processors/

I wouldn't be surprised if 28nm SOI was cancelled at GloFo and replaced with 22nm and some of these delays are for porting to 22nm SOI.

 

[de5_Roy]

^^ what would happen if kaveri was not designed/optimized for different foundries like jaguar?

 

[Cazalan]

^^ what would happen if kaveri was not designed/optimized for different foundries like jaguar?

Jaguar is designed for TSMC bulk. They can't just take that anywhere else. TSMC has their own cell libraries. It may be easier to port it to other bulk process but it's still considerable time and effort.

Now if it were GF bulk, they could also take that to Samsung or IBM.They're part of the Common Platform alliance.

http://www.commonplatform.com/

 

[gamerk316]

Notice, they say "support"...again...anyone developing specifically for consoles would be an idiot to use a less optimal physics engine. The only edge case I could see would be a PC game that was later ported to consoles, but even then...that doesn't make a lot of sense.

PhysX is the ONLY physics engine that does a good job at multi-object dynamics; Havok falls flat on its face once you start to interact with more then one or two objects at a time. That's why PhysX is typically used for a lot of effects that involve lots of small objects [particles, wind, etc].

http://www.youtube.com/watch?v=143k1fqPukk

Bullet would like to have a word with you.

If you have to choose between PhysX on CPU and Bullet on GPU, which do you choose?

Also, why are you guys getting so worked up over 28nm FD-SOI when IBM has 22nm SOI?

http://www.advancedsubstratenews.com/2013/08/pcmags-michael-miller-called-ibms-22nm-soi-power8-the-most-fascinating-of-the-high-end-processors/

I wouldn't be surprised if 28nm SOI was cancelled at GloFo and replaced with 22nm and some of these delays are for porting to 22nm SOI.

Ok, I've not played with Bullet much, but it looks a little more limited then PhysX. But if its significantly faster, that alone may be worth it. Both have the same concept though: Use the power of the GPU to allow for much more realistic physics simulations.

 

[griptwister]

Notice, they say "support"...again...anyone developing specifically for consoles would be an idiot to use a less optimal physics engine. The only edge case I could see would be a PC game that was later ported to consoles, but even then...that doesn't make a lot of sense.

PhysX is the ONLY physics engine that does a good job at multi-object dynamics; Havok falls flat on its face once you start to interact with more then one or two objects at a time. That's why PhysX is typically used for a lot of effects that involve lots of small objects [particles, wind, etc].

http://www.youtube.com/watch?v=143k1fqPukk

Bullet would like to have a word with you.

If you have to choose between PhysX on CPU and Bullet on GPU, which do you choose?

Also, why are you guys getting so worked up over 28nm FD-SOI when IBM has 22nm SOI?

http://www.advancedsubstratenews.com/2013/08/pcmags-michael-miller-called-ibms-22nm-soi-power8-the-most-fascinating-of-the-high-end-processors/

I wouldn't be surprised if 28nm SOI was cancelled at GloFo and replaced with 22nm and some of these delays are for porting to 22nm SOI.

That would be bloody awesome! I really hope you're right about 22nm SOI!

 

[hcl123]

According to GLofo 28 nm FD-SOI will not be ready for risk production until 4Q 2013 aka "not ready still"

According to AMD kaveri desktop comes first then mobile.

Any source ? (link please)