What to Expect From AMD at ISSCC 2011

[Caparroz]

Is it so difficult to see he was merely showing the scaling differences between extra cores, hyperthreading and the way bulldozer does it: Extra cores which share resources inside a module, which saves die space versus extra cores which don't share any resources?

IMO it was a very informative post. Don't forget JF-AMD is under NDA and simply can't give us any performance numbers.

You do realize eXceededgoku was just kidding, right?

I'm really curious about what BD will bring in terms of multi-threaded computing. It could be really amazing for the kind of work I and my colleagues do.

[eXceeded]

You do realize eXceededgoku was just kidding, right?

I'm really curious about what BD will bring in terms of multi-threaded computing. It could be really amazing for the kind of work I and my colleagues do.

Yeah quite clearly kidding . Absolutely, can't wait for BD to come. Should be VERY useful for what I need it for as well . Also helps that my old C2D has kicked the bucket!

EDIT - at my post count

[freeloader]

cant blame amd for being secretive about a 10 year long project can we???

At first I didn't understand AMD's position on keeping quiet. But the more I hear about it, the more I believe they have a killer chip on their hands. Keeping quiet right now is preventing their current CPU lineup from sitting on store shelves forever. I'd be quiet about it too.

Purchasing a dual core C32 setup this week. Looking forward to C32 BD server variants in Q3.

[[XC] hipno650]

At first I didn't understand AMD's position on keeping quiet. But the more I hear about it, the more I believe they have a killer chip on their hands. Keeping quiet right now is preventing their current CPU lineup from sitting on store shelves forever. I'd be quiet about it too.

Purchasing a dual core C32 setup this week. Looking forward to C32 BD server variants in Q3.

i dunno if it's that in particular.... considering the number of people with any sort of real knowledge about their product is under 1% of their consumer base. only a handful of people know that BD is more then a large tractor. hell even the majority of people who work at computer shops have no idea what BD is. if they leaked results it would make no noticeable impact on sales of current CPU's. I think they are keeping the lid on things because it the least risky move to make and gets consistent results. as much as we both wish they would leak results it's just not the way the industry works...

[Musho]

You do realize eXceededgoku was just kidding, right?

I'm really curious about what BD will bring in terms of multi-threaded computing. It could be really amazing for the kind of work I and my colleagues do.

Yeah quite clearly kidding . Absolutely, can't wait for BD to come. Should be VERY useful for what I need it for as well . Also helps that my old C2D has kicked the bucket!

EDIT - at my post count

Whoops, my bad. Sorry!

[accord99]

QFT

bobcat is 90% of todays cores clock/clock

Bobcat is not even 90% clock/clock of a 2004 Clawhammer core, let alone anything today.

[god_43]

Bobcat is not even 90% clock/clock of a 2004 Clawhammer core, let alone anything today.

....you should look at the reviews again buddy.

[accord99]

....you should look at the reviews again buddy.

Loses most of the time to a 1.5GHz K8 based dual-core:

http://www.anandtech.com/bench/Product/328?vs=116

It's around 67% of a 1MB L2 version of Conroe at 1.6 GHz:

http://www.anandtech.com/bench/Product/328?vs=70

[demonkevy666]

Loses most of the time to a 1.5GHz K8 based dual-core:

http://www.anandtech.com/bench/Product/328?vs=116

It's around 67% of a 1MB L2 version of Conroe at 1.6 GHz:

http://www.anandtech.com/bench/Product/328?vs=70

1.The E-350 doesn't have a hyper transport
2. The E-350 also is sharing the memory controller with the gpu. in single channel
3 The E-350 has a gpu on die the 3250e does not.
4 The TDP is below that.
5 This doesn't compete with Conroe based cores it's competitor is ATOM.

[accord99]

1.The E-350 doesn't have a hyper transport
2. The E-350 also is sharing the memory controller with the gpu. in single channel
3 The E-350 has a gpu on die the 3250e does not.
4 The TDP is below that.
5 This doesn't compete with Conroe based cores it's competitor is ATOM.

All of which is irrelevant to the fact that Bobcat is nowhere near 90% of the clock/clock performance of "todays" core.

[Mechanical Man]

Loses most of the time to a 1.5GHz K8 based dual-core:

http://www.anandtech.com/bench/Product/328?vs=116

It's around 67% of a 1MB L2 version of Conroe at 1.6 GHz:

http://www.anandtech.com/bench/Product/328?vs=70

Naturally that 90% did not refer to conroe. AMD surealy compared that 90% to their own Athlon II processors. That afterall most likely is mainstream in market pc's.

[BeepBeep2]

All of which is irrelevant to the fact that Bobcat is nowhere near 90% of the clock/clock performance of "todays" core.

Give K8 a serious lack of memory bandwidth like you are to Zacate (for example 2GB single channel DDR3-1066 CAS 7 and have the onboard GPU use 1GB...

Things arent gonna look good compared to Bobcat.

By the way, nobody officially said Bobcat was 90% CPC of todays core.

[informal]

Actually if Bobcat didn't have half clocked L2(!) and only single channel mem. controller ,I personally suspect it would probably come very close to Athlon II at similar clock. The big performance downside is also the FPU unit which is compromised for power draw reasons.Bobcat features 64bit FPU,although it supports a lot of ISA extensions(adds SSSE3,removes 3dnow). 64bit FPU is the key performance difference between Bobcat and family 10h,since it takes 2 cycles to perform 128bit packed SSE instructions as opposed to 1 cycle on Regor/Thuban/Conroe/Penryn/Nehalem classes of cores.
So yes,Bobcat performs great considering all above performance trade offs AMD made. Bulldozer not only does not have these trade offs,it has much much more core logic investments ,even compared to single family 10h core(in both integer and FP parts of the module/cores).

[Manicdan]

theres a nice review of the HP dm1z
http://www.anandtech.com/show/4187/h...-on-the-road/7

on the dm1z you're going to find a surprisingly capable little computer that really screams past the Atom-based netbooks of old. It only felt sluggish when I was using it next to another computer with a mainstream or better processor; otherwise the dm1z was perfectly serviceable and a far cry from the waitfest that an Atom-based netbook can quickly become.

[sniper_sung]

Where is the spec for 6990?

[Shadov]

i dunno if it's that in particular.... considering the number of people with any sort of real knowledge about their product is under 1% of their consumer base. only a handful of people know that BD is more then a large tractor. hell even the majority of people who work at computer shops have no idea what BD is. if they leaked results it would make no noticeable impact on sales of current CPU's.

Ohh my, why do people write posts before they make the slightest effort to research how for example the IT sales channel works. I already have the picture (honey I cant come to bed, someone is wrong on the internet) in front of my eyes when im writing this post. But anyway, here it goes... Ill try to explain.

If AMD would give out early performance numbers the following things could potentially happen:

-market would expect new, better performing CPU's, while no actual product could be sold for quite some time stalling sales results

-distributors which sell e.g. AMD products would adjust their orders, since something new is about to come and they wouldn't want to end up with "old" stock in their warehouses

-sales channel would expect more support from AMD to move current products (since they shot themselves in the foot by announcing unavailable product and are not in the position of Intel to dictate rules)

-Intel would know exactly what to expect and how to structure price points to counter, which again would impact early reviews and market perception

Again do you think that AMD should give away early numbers behind their projects only because some people are curious and cant wait? In my opinion they should take their sweet time to make it as successful as possible...

[maxionline]

Ohh my, why do people write posts before they make the slightest effort to research how for example the IT sales channel works. I already have the picture (honey I cant come to bed, someone is wrong on the internet) in front of my eyes when im writing this post. But anyway, here it goes... Ill try to explain.

If AMD would give out early performance numbers the following things could potentially happen:

-market would expect new, better performing CPU's, while no actual product could be sold for quite some time stalling sales results

-distributors which sell e.g. AMD products would adjust their orders, since something new is about to come and they wouldn't want to end up with "old" stock in their warehouses

-sales channel would expect more support from AMD to move current products (since they shot themselves in the foot by announcing unavailable product and are not in the position of Intel to dictate rules)

-Intel would know exactly what to expect and how to structure price points to counter, which again would impact early reviews and market perception

Again do you think that AMD should give away early numbers behind their projects only because some people are curious and cant wait? In my opinion they should take their sweet time to make it as successful as possible...

If AMD shows some GOOD numbers may stop a bit SB sells...

[dartaz]

Naturally that 90% did not refer to conroe. AMD surealy compared that 90% to their own Athlon II processors. That afterall most likely is mainstream in market pc's.

It would be closer to the truth if you say that Zacate is 90% of the old K8 based Athlon

[AKM]

http://blogs.amd.com/work/2011/02/21...hats-in-a-box/

At ISSCC I will discuss how AMD’s world-class design team met these goals without resorting to a full-custom design. To allow rapid prototyping of new ideas and reduce time to market, full-custom design blocks were replaced with circuits built out of standard cells, using conventional ASIC techniques and a couple of unique twists.

Does forgoing a full-custom design mean any performance tradeoffs?

[informal]

AKM it's being already posted.

[Shadov]

If AMD shows some GOOD numbers may stop a bit SB sells...

You dont stop sells with unavailable products in a significant way especially when most companies and their product life cycles are dependant on current shipments. This approach could hurt AMD more then help.

Cmon, try to look at the bigger picture!

Anyway, enough...

[Hornet331]

5 This doesn't compete with Conroe based cores it's competitor is ATOM.

Actually it does, especial the E-series. Because you don't only find them in netbooks, but also in entrey 15" books and there it competetns pricewise against i3s.

And looking at the current prices the E-series is not attractive at all for that segmen, a i3-350m+Geforce 310M for 450€ or a E-350 for 399€.

E-based books need to be at least 100€ cheaper then compareable i3+dedicated gpu to stand a chacne in that market.

[Manicdan]

price seems to be set by the features and battery life. surly it might lose on synthetics, but most people shopping for sub 500$ laptops might not care about that, as long as it DOES whats expected.

[[XC] hipno650]

Ohh my, why do people write posts before they make the slightest effort to research how for example the IT sales channel works. I already have the picture (honey I cant come to bed, someone is wrong on the internet) in front of my eyes when im writing this post. But anyway, here it goes... Ill try to explain.

If AMD would give out early performance numbers the following things could potentially happen:

-market would expect new, better performing CPU's, while no actual product could be sold for quite some time stalling sales results

-distributors which sell e.g. AMD products would adjust their orders, since something new is about to come and they wouldn't want to end up with "old" stock in their warehouses

-sales channel would expect more support from AMD to move current products (since they shot themselves in the foot by announcing unavailable product and are not in the position of Intel to dictate rules)

-Intel would know exactly what to expect and how to structure price points to counter, which again would impact early reviews and market perception

Again do you think that AMD should give away early numbers behind their projects only because some people are curious and cant wait? In my opinion they should take their sweet time to make it as successful as possible...

im pretty sure you totally missed out on everything I said.... I don't think it was rocket science to figure out.

anyways I never once said I think it would be a good business move for AMD to release BD numbers ahead of release in fact if you would have read my whole post you would see that I said "I think they are keeping the lid on things because it the least risky move to make and gets consistent results. as much as we both wish they would leak results it's just not the way the industry works"

understand one thing. I could give a RATS @$$ how much money AMD make or what is good business practice for any of the large tech companies. I am a selfish person and I would like to see early BD numbers (as should any tech enthusiast) and really unless someone has shares in AMD I fail to see why they would care so much about there business practices... or "what is good for them as a company"

and honestly some of the points you have while being semi-true are way over done.

-The market would not even come close to stalling because of supply and demand. outside of the enthusiast community (less then 1% as I stated before) people don't know about or wait for new CPU's to be released. people and businesses need new computers and they buy them when they need them.

-Distributors will adjust their orders but ONLY really close to launch because again people are still buying things.... and AMD can always sell off the extras to OEM's and the like (which both AMD and Intel do all the time)

-leaking some numbers (does not need to be officially, could be via some Chinese Site) is not the same as announcing the pricing and marketing tactics of your new confirmed CPU

-Intel is unlikely to change prices until a few days of BD launch anyways otherwise they will loss money. which they will do regardless of an early leak or not.

and really man no need to be "I am so much smarter then everyone on the internetz...." I made no attack on you or pretty much anyone for that matter so no need to attempt at insulting my intelligence....

[Hornet331]

price seems to be set by the features and battery life. surly it might lose on synthetics, but most people shopping for sub 500$ laptops might not care about that, as long as it DOES whats expected.

Well the ASP for notebooks is/was ~730$ in 2010, so it seems people don't mind paying more.

Especial in the 15"+ segement, they arn't exactly mobile to beginn with so people look more for performance. So for a few bugs more you get better graphic performance and way better cpu performance, with same bat life time (3-4h). (e.g. the Acer 5253 - 3,2h vs Lenovo Z560 with 3.5h)

Imho if the OEM can sale this books to the consumers they will make a huge profit wiith it...