AMD Ontario APU pictured,die size ~77mm^2

[Raqia]

Those cores are incredibly small; I guess multi-threading in Atom must account for some of this. The 90% of mainstream performance claim sounds always sounded a bit sketchy, but atleast the BOINC benchmarks look very promising. All of these vague marketing claims should just be ignored until benchmarks crop up; Intel used to claim that performance of its Atom was supposed to be on par w/ the pentium m iirc.

[Manicdan]

dumb question, why not make a massive array of bobcat cores to mimic what LRB was suppose to be?

[Zibi]

32 bobcat cores + 8-12MB cache L3 + IMC & control logic, all in 32nm technology = powerfull CPU, probably like 24 cores Phenom II.

[ajaidev]

dumb question, why not make a massive array of bobcat cores to mimic what LRB was suppose to be?

In its current state its not very easy thing to do and well there was a lesson in that for Intel, AMD does not have much experience with what you are proposing and it would take a redesign "at least a minor one" and some additions have to be made.

In the end i expect a bulldozer to be better than a combo bobcat core because of the shared units and the shared L3 but the speed of the L3 has to be increased to keep up with the flow.

EDIT: What i wrote above is in respect to ray-tracing.

[trinibwoy]

dumb question, why not make a massive array of bobcat cores to mimic what LRB was suppose to be?

Bobcat is missing the most important component of LRB - the vector ALU.

[Manabu]

dumb question, why not make a massive array of bobcat cores to mimic what LRB was suppose to be?

Bobcat has relatively low vector/floating point math performance, the oposite of LRB cores. AMD already has an very good high performance graphics architeture, that they are turning more and more general purpose, with much more sucess than Intel is having with LRB. Why do you think they should abandom this winning strategy, with OpenCL, DirectCompute and so on, and copy what Intel is trying to do?

Bobcat may be suited for something like what intel is planing with it's "Single Chip Cloud". But combining so many cores in one die is not an easy task (see for how many years intel is making studies/prototipes on this), and IMHO it's is somewhat a niche market. I'm not sure that this would be the best way to spend R&D for AMD.

[Chumbucket843]

dumb question, why not make a massive array of bobcat cores to mimic what LRB was suppose to be?

1. larrabee is about data parallelism(vectorization). not task parallelism (multithreading). afaik bobcat does not support AVX. LRB already has 4x the SIMD width and better quality SIMD than AVX or SSE imo.

2. the hard part is not smacking a bunch of cores together. anyone can do that. making an efficient highly parallel processor is tricky. a lot of the issues arise from the fact that very little research as been done on parallelism compared to other areas.

[LightSpeed]

Even if its 75% of the performance of Athlon II, at 4.7mm^2 its brilliant

I expected around 6 - 7mm^2..

[mindfury]

http://www.insidehw.com/Editorials/E...pressions.html

We were one of the first (very first maybe) to go behind the scene at AMD booth at IFA Berlin and to take a sneak peak of Ontario APU platform while working. Unfortunately, we can’t share any pictures, but we can assure you that Ontario test platform is alive and kicking. Aside successfully working with Windows 7 OS, it did manage to accomplish several other things, but more on that later on. As it was stated on test motherboard, the silicon used was A0, with integrated UVD (most probably version 3.0). During the test, very small and basic cooling solution was used, but it still managed to keep APU cooled at around normal body temperature even under heaviest loads.
we naturally wanted to check how good HD video playback is. Well known and slightly irritating Big Buck Bunny video was used, 1080p resolution, and both playback and fast forwarding were very, very smooth. Next we tried Microsoft‘s N-Body Particle systems test, which showed dramatic performance difference when using GPU instead CPU for calculation, as well as full DirectX 11 compatibility. Among other things we tried, Office 2010 worked very fluidly, and we even witnessed 3Dmark06 in action. As this not being final product, with driver and possibly architecture improvements still to come, we can’t give you any numbers there, but the fact it was working and didn’t take forever to finish things, is good enough for us. There was AVP benchmark shortcut, but we couldn’t run it at the moment, hopefully we will have more luck tomorrow.

[Hornet331]

lol what.. 3dmark alway takes the same time to finish (excluding loadingtimes )...

[STEvil]

lol what.. 3dmark alway takes the same time to finish (excluding loadingtimes )...

no, but on a system that can pull more than 5-10fps.. yeah most of the time.

[AliG]

You made to many assumption here. Based on what we know about bobcat architecture, it should be in range of athlon64 perf.
Lets see.
Bobcat:
2 decoders, 2 ALU+2 AGU (max 4 uops per cycle), 64-bit fpu (max up to 2 uops ).
Athlon64:
3 decoders, 3 ALU/AGU (max up to 6 uops per cycle) 64-bit fpu (max up to 3 uops per cycle).

here we go again....

[blindbox]

here we go again....

[-Boris-]

here we go again....

Yup. Same old deliberate misinterpretation of K8s theoretical performance.

[superrugal]

Cuz bulldozer & bobcat has 2 ALU, so they have same perf.
Cuz K7 has 3 ALU, so K7 may punch Bulldozer flying to the sky.

L O L

[richierich]

http://www.insidehw.com/Editorials/E...pressions.html

sounds fun so far. Lil guy packing some punch.

[metro.cl]

1 more year to wait for decent netboks?

[Calmatory]

"Estimated 90 % of the performance of today's mainstream notebook CPU in half the area", referring to the AMD's most recent mobile CPUs?

Now someone find some numbers and compare to Atom & Nano.

Quad core Bobcat laptop with DX11 & OpenGL 4.0 capable GPU(regardless of the speed, so even Ontario would suffice, just need more cores) would be LOVELY for code toying.

[RaV[666]]

1 more year to wait for decent netboks?

It should start to trickle down before the end of the year, so wide availability in Q1 2011.So more like half a year.I myself cant wait, was waiting for a decent netbook platform for a while now.

[BatteryOperated]

1 more year to wait for decent netboks?

Yes.

Quite frankly, you should not expect a Bulldozer-related product before 1q 2012. I am not trying to dissuade anyone from buying AMD. I flat out do not expect anything from AMD on Fusion or Bulldozer until 2012.

Will Bulldozer perform well? Yes.
Will Bulldozer be released before 2012 no.

[Sn0wm@n]

Yes.

Quite frankly, you should not expect a Bulldozer-related product before 1q 2012. I am not trying to dissuade anyone from buying AMD. I flat out do not expect anything from AMD on Fusion or Bulldozer until 2012.

Will Bulldozer perform well? Yes.
Will Bulldozer be released before 2012 no.

LOL JF-AMD said it that bulldozer is dated for 2011 at least in the server market ... so why would the desktop take 1 year more ....

[savantu]

LOL JF-AMD said it that bulldozer is dated for 2011 at least in the server market ... so why would the desktop take 1 year more ....

http://www.brightsideofnews.com/news...nd-fusion.aspx

Akrout said to expect Orochi as soon as fourth quarter 2011, meaning AMD's internal roadmap slipped from second to fourth quarter - another six month delay in delivering octo-core Bulldozer-based part to the market. In our conversation with AMD's execs, we were told that Orochi will also be the base for second generation Fusion part coming after Llano, which is now in test production.

Chekib Akrout, senior VP of the AMD Technology Group.

Q4 is 2011 just like JF said.
Just 13-15 months away...

[Manabu]

Yes.

Quite frankly, you should not expect a Bulldozer-related product before 1q 2012. I am not trying to dissuade anyone from buying AMD. I flat out do not expect anything from AMD on Fusion or Bulldozer until 2012.

Will Bulldozer perform well? Yes.
Will Bulldozer be released before 2012 no.

Ontario isn't a bulldozer-related product, neither a Fusion one (this last part I'm not sure). It don't have to wait for the problematic Global Foundries's 32nm, as it will be in the already well known TMSC 40nm. So, it will be something like what RaV[666] said. I personaly would really LOVE if they become a bit ahead of the shedule, thought.

And please, stay on topic. This is not an bulldozer/orochi thread.

[SimBy]

Yes.

Quite frankly, you should not expect a Bulldozer-related product before 1q 2012. I am not trying to dissuade anyone from buying AMD. I flat out do not expect anything from AMD on Fusion or Bulldozer until 2012.

Will Bulldozer perform well? Yes.
Will Bulldozer be released before 2012 no.

Huh? Fusion AKA Ontario and Zacate Q4 2010, Llano Q1 2011. Dozer whenever GloFo gets 32nm yields where they need to be.

[ryboto]

http://www.brightsideofnews.com/news...nd-fusion.aspx



Chekib Akrout, senior VP of the AMD Technology Group.

Q4 is 2011 just like JF said.
Just 13-15 months away...

but JF corrected the news with his own comments-

The dates are:

Bobcat: producttion in Q4 2010, systems in early 2011.

Bulldozer: 2011

There is not a quarter granularity on Bulldozer, there is still an engineering checkpoint that needs to be passed before we give quarterly granularity.

Your statement that Bulldozer is Q4 2011 is not accurate.

JF-AMD

and

All I was commenting on was that Chekib did not say "Q4 2011".

The official status is: sampling to partners in Q4 2010 and production/shipping in 2011.

We have not gotten more granular than that. I asked Chekib yesterday and he confirmed that he said "2011" with no quarter.

There are still some engineering checkpoints to pass before we get down to quarter granularity in a public statement.

JF-AMD.