[AT] AMD Kaveri APU Launch Details: Desktop, January 14th

[NTMBK]

Kaveri will have up to 4 CPU core (2 modules), which will be based on AMD’s latest revision of their desktop CPU architecture, Steamroller. Meanwhile the GPU will be composed of 8 GCN 1.1 CUs, which would put the SP count at 512 SPs (this would be equivalent to today's desktop Radeon HD 7750). Furthermore AMD is throwing around a floating point performance number – 856 GFLOPS – which thanks to some details found in AMD's footnotes by PCWorld gives us specific clockspeeds and even a product name. A10-7850K CPU clockspeed 3.7GHz, GPU clockspeed 720MHz.

http://www.anandtech.com/show/7507/amd-kaveri-apu-launch-details-desktop-january-14th

And can we please not have a rerun of the last thread?

 

[monstercameron]

why not wait for the mods to clean and reopen thread? or they dont do that here?

 

[NTMBK]

why not wait for the mods to clean and reopen thread? or they dont do that here?

The mod response to the last thread:

Ok, this thread is lost. Between the fanboys on both sides, the mild insults, threadcrapping, and off-topic posts, this thread is done. If someone wants to start a new one, and try to be civil, I will leave it alone. Your only other chance is for me to infract almost every poster in this thread.

Now, please can we let that be the last reference to the old thread? If any tries to carry on any off topic arguments and bickering from the last thread, I WILL ask the mods to close it.

 

[Homeles]

why not wait for the mods to clean and reopen thread? or they dont do that here?

Mark's not going to reopen it. He's opted to lock the thread and allow a new one to be opened, rather than hand out infractions.

 

[USER8000]

PCI-E 3.0 controller and onboard TrueAudio support confirmed(thanks inf64).

Possible hints to die size(thanks inf64)??

Core i7 4770K with GT630 compared to Kaveri(thanks monstercameron):

https://www.youtube.com/watch?v=HjAM2zYNqko

Some framerate figures for the demo:

http://hothardware.com/News/AMD-Ann...s-at-APU13-Confirms-Kaveri-is-A-512Core-Part/

To showcase its performance, AMD showed Kaveri up against the Haswell-based Intel Core i7-4770K paired to a low-end GeForce GT 630 discrete GPU. There are three variants of this particular card -- it's not clear which AMD utilized -- but the overall performance figures are rather stark. In the intro scene to BF4's single-player campaign (1920x1080, Medium Details), the AMD Kaveri system (with no discrete GPU, using only on-die graphics) consistently pushed frame rates in the 28-40 FPS range. The Intel system, in contrast, couldn't manage 15 FPS. Performance on that system was solidly in the 12-14 FPS range -- meaning AMD is pulling 2x the frame rate, if not more.

AMD mentioned Mantle was not used in the demo.

AMD also talked about GPU accelerated Java too.

AMD is also announcing CodeXL 1.3, which is significant for its addition of Java support. GPU-accelerated Java has been a talking point of the HSA initiative for several months, so actually pushing that option into standard tool sets is a major achievement for the company.

 

[monstercameron]

well then, the next apu13 conference is about to begin...so maybe some tidbits of information will be revealed.

 

[monstercameron]

240mm^2 is my guesstimation, the 7750 was 123mm^2 and I read somewhere that ~50% of kaveri's die is all gpu...

 

[USER8000]

240mm^2 is my guesstimation, the 7750 was 123mm^2 and I read somewhere that ~50% of kaveri's die is all gpu...

Trinity has a percentage of the CPU dedicated to the IGP too,so if you even take the TrueAudio block out of the equation,that might mean the CPU section is probably a reasonable jump in size(number of transistors).

 

[insertcarehere]

240mm^2 is my guesstimation, the 7750 was 123mm^2 and I read somewhere that ~50% of kaveri's die is all gpu...

That would be assuming that 28nm GF is equally dense as 28nm TSMC for gpus. Considering the size of igpu in richland/llano (32nm GF) vs size of comparable vilw4/5 AMD gpus (40nm TSMC), I doubt it.

 

[ShintaiDK]

240mm^2 is my guesstimation, the 7750 was 123mm^2 and I read somewhere that ~50% of kaveri's die is all gpu...

The 7770 is also 123mm2 (7770 and 7750 shares the same die.). So it could be even smaller.

 

[USER8000]

That would be assuming that 28nm GF is equally dense as 28nm TSMC for gpus. Considering the size of igpu in richland/llano (32nm GF) vs size of comparable vilw4/5 AMD gpus (40nm TSMC), I doubt it.

AMD has stated(if you look at the slide I posted),that 47% of the Kaveri chip is for the IGP.

 

[AtenRa]

PCI-E 3.0 controller and onboard TrueAudio support confirmed(thanks inf64).

Possible hints to die size(thanks inf64)??

Core i7 4770K with GT630 compared to Kaveri(thanks monstercameron):

https://www.youtube.com/watch?v=HjAM2zYNqko

Some framerate figures for the demo:

http://hothardware.com/News/AMD-Ann...s-at-APU13-Confirms-Kaveri-is-A-512Core-Part/

AMD mentioned Mantle was not used in the demo.

AMD also talked about GPU accelerated Java too.

This is a Jaguar die pic, it has nothing to do with Kaveri

 

[Asterox]

240mm^2 is my guesstimation, the 7750 was 123mm^2 and I read somewhere that ~50% of kaveri's die is all gpu...

Well today you've got an official information or confirmation of this fact.

 

[NTMBK]

The 7770 is also 123mm2 (7770 and 7750 shares the same die.). So it could be even smaller.

Plus the 7750/70 has a DDR3/GDDR5 memory controller, while in Kaveri this is shared with the CPU.

 

[Erenhardt]

There is a stream. You need to register at apu365, login, choose a key note and in the middle section click button "live now"

 

[inf64]

@ To those who asked me about GDDR5 support in IMC in locked topic

Sorry I cannot give you the source . BUT the other source that says basically the same thing is the 1st version of AMD's Software Optimization Guide (SOG) that was under NDA but was still posted in an article by BSN. Images from it are 100% genuine. The GDDR5 support is cancelled for couple of reasons (supply being number one) and the die area for it is basically wasted. It's not large die space so it's not a big deal.

 

[Homeles]

If it's not going to be used, why bring it up in the first place?

 

[inf64]

They certainly didn't plan not to use it of course. Their plans changed so they opted to cancel the GDDR5 support on the motherboard side (due to market related reasons).

 

[inf64]

Toms' has live notes coming in from day2. Some Java perf. numbers, looks good.

Searching through 4 books in multi-core mode to begin with. Task complete in 44:49s

Now running the same task, but on the GPU rather than the CPU

GPU compute is done in 9:48s, shared virtual memory in action since the Java threads could dynamically access the data whilst it was being processed

 

[dorion]

This GPU-accelerated Java, can I assume they just mean a way to perform GPU compute tasks with Java?

 

[Markfw]

Thanks to all (so far) in this thread. This is what the other one should have been. I don't have hours to "clean" that other thread, and there were problems on BOTH sides, so let this be the last reference to the other thread.

 

[Kallogan]

What you should expect without DDR4

http://www.hardware.fr/medias/photos_news/00/39/IMG0039893.gif

 

[Vesku]

Toms' has live notes coming in from day2. Some Java perf. numbers, looks good.

Are there Richland A10 versions of those java perf. numbers to compare to?

 

[Phynaz]

I have my fingers crossed this turns out really good. I'm hoping AMD breaks their track record on the CPU side of the house. We need them to do to Intel what they did to Nvidia.

 

[NaroonGTX]

What you should expect without DDR4

http://www.hardware.fr/medias/photos_news/00/39/IMG0039893.gif

Link is broken.