AMD Zambezi news, info, fans !

[-Sweeper_]

Ouch, this doesn't look very good, BD 8C being positioned slightly lower than 4C Core i7, though I guess that makes sense, most desktop workloads use 4 cores at most, and isn't one SNB core as big as 2 BD cores? So it'd be a miracle if per core perf. would be close, turbo can mitigate that somewhat but not enough. I hope they do better in server, that's hugely more important for AMD.

no, 1 SB core with SMT is ~18.4mm² while each bulldozer module is 30.9mm²

2x IB (22nm) cores with SMT should be smaller than a BD module (2x ~13-14mm²)

[informal]

One BD module has also 2MB of L2 while 1 SB core has only 256KB. So without L2 ,one bulldozer core(if possible to be viewed like that) would be around 7-8mm^2 and with 1MB of L2 it would be around 14mm^2. But aside of that,2 IB cores will probably just be slightly smaller than 1 BD module.How much performance you expect from 2 IB cores?It is basically SB shrank with probably same amount of L2 per core.Clock will be boosted probably,but at the expense of TDP. By that time AMD will have Terramar ,20/10 core server and 10 core desktop chip based on improved Orochi design. 25% more cores and 25% more TDP headroom for Turbo boost clock on improved GF 32nm SOI hk/mg process. That one module is just 30mm^2 more Vs Zambezi/Valencia and it brings 25% more performance. That's the beauty of Bulldozer design .

[Sn0wm@n]

Bulldozer will be late ... Expect more august ...

now why would you say that

[-Sweeper_]

One BD module has also 2MB of L2 while 1 SB core has only 256KB. So without L2 ,one bulldozer core(if possible to be viewed like that) would be around 7-8mm^2 and with 1MB of L2 it would be around 14mm^2. But aside of that,2 IB cores will probably just be slightly smaller than 1 BD module.How much performance you expect from 2 IB cores?It is basically SB shrank with probably same amount of L2 per core.Clock will be boosted probably,but at the expense of TDP. By that time AMD will have Terramar ,20/10 core server and 10 core desktop chip based on improved Orochi design. 25% more cores and 25% more TDP headroom for Turbo boost clock on improved GF 32nm SOI hk/mg process. That one module is just 30mm^2 more Vs Zambezi/Valencia and it brings 25% more performance. That's the beauty of Bulldozer design .

why take out L2, it doesnt belong to the bulldozer module?

if there is 2mb L2 it is because it probably needs that 2mb L2 to perform well and AMD has to pay for that extra space...

by the time AMD will have terramar intel will have IB-EP with 12c/14T (50% more cores) on a 22nm tri-gate process, possibly on higher stock clocks than SB-EP plus a more agressive turbo mode (yes, a lot more TDP headroom to play with too)...

and lets see if the beaty of that design is enough to match/beat a 4ghz+ IB 4C/8T (probably with minor IPC boost 0-5%) in daily desktop workloads, cause thats what consumers (even intel fans) want, to bring prices down...

[informal]

When do you expect IB-E will launch? Terramar is scheduled for Q3 next year so let's see when is the IB-E scheduled to launch. Komodo for desktop(10 BD+ cores,no IGP) is Q2 2012.
IB mainstream will be a Fusion-type chip,so it will go against 2 module improved Orochi with improved Radeon cores(more SPs,improved IMC). Do you really think it is going to have hard time against IB+16EU GPU? It may be a bit slower in CPU tasks,but it will blow away that HD4000 to another world . Then you put a discrete Radeon card and get another 60% on top of that mountain of GPU advantage... So it's not AMD who will play catch up game in mainstream desktop segment. One thing intel will probably use as a marketing point and with a reason, is QuickSync. Great feature indeed but we shouldn't forget OpenCL and what GPGPU will be able to do one year from now.

[-Sweeper_]

we're talking about x86 CPU performance, fusion should have higher IGP performance and I dont expect that to change anytime soon.. it may surprise you but most people dont buy computers to player hardcore games and SB IGP is pretty fine for those, even casual gamers... telling intel will play catch up in mainstream desktop segment because of graphics performance deficit sounds a little desperate..

the openCL bennefit of those stream processors is yet to be seen by the average joe...

[informal]

Average Joe has no idea how fast his CPU is . He just knows that when he or his son runs a Counter Strike or Half Life 3/4/whatever, he doesn't want a slide show,he wants a playable game.And don't tell me Average Joe will be able to see a difference in CPU performance between QC 3.5-4Ghz BD+ X4 and QC 4GHz IB because I highly doubt this .

[-Sweeper_]

Average Joe has no idea how fast his CPU is . He just knows that when he or his son runs a Counter Strike or Half Life 3/4/whatever, he doesn't want a slide show,he wants a playable game.And don't tell me Average Joe will be able to see a difference in CPU performance between QC 3.5-4Ghz BD+ X4 and QC 4GHz IB because I highly doubt this .

well, thank god SB/IB IGP doesnt/wont run many popular games like a slide show... the same way you think BD+'s CPU performance might be more than enough to most consumers, I say the same about SB/IB IGP...

I'd also say that if the average joe ever notices a difference between them, it would be while doing CPU tasks, not watching his son play games at higher resolutions

By that time AMD will have Terramar ,20/10 core server and 10 core desktop chip based on improved Orochi design.

I smell a contradiction here, you just told us terramar & komodo would launch in Q3 & Q2, so they will be out months after the first IB chips arrive (IB-EP is the one who comes later H2/2011)

[informal]

The only problem is that average joe won't notice a few seconds here and there in CPU performance while he may notice 2-4x lower frame rates . Being 10-20% slower in CPU department is not bad and you know it,while being 2-4x slower in games and GPGPU is more than bad. It's plain awful when you think of it. With Fusion type of CPUs it's all about the complete package Joe gets,and he will get much better package with Bulldozer Fusion.

edit:

I smell a contradiction here, you just told us terramar & komodo would launch in Q3 & Q2, so they will be out months after the first IB chips arrive (IB-EP is the one who comes later H2/2011)

Where is the link for IB-E launch date?? As you can read my post,I never mention IB since we all know it's Q1 2012 or Q4 this year if stars align. IB-E news on the other hand are nowhere to be found.

[-Sweeper_]

wow, someone already has numbers comparing IB's IGP with the next gen fusion, could you share more with us informal?

you could also point out why you think most people around the world care about hardcore gaming while buying a computer
(specially without discrete graphics card for those few % who actually really care)...

[informal]

wow, someone already has numbers comparing IB's IGP with the next gen fusion, could you share more with us informal?

you could also point out why you think most people care about hardcore games while buying a computer without discrete graphics card...

We know it's 16EU,it's been already posted in the news. How much better performance can it get? Ontario's weak CPU with 80SPs is posting fps numbers in the realm between HD2000 and HD3000.If IB matches Llano's 400SPs I would be hugely surprised . BD Fusion is going to ramp up Llano's SP count,so if that impossible task of matching Llano happens(which it won't since intel needs 2x-3x more GPU power to do that), BD Fusion will stomp it again.

As for hardcore gaming comment,where did I mention HC gaming?? BD Fusion (and Llano) will have much better GPU,no matter what settings you use on both competing products.The point is it will give you either much better fps at any resolution used or much higher in-game image quality with similar fps (AF/AA on HD intel solution I won't even start mentioning). No matter how you turn it around ,with Fusion it's not only about CPU cores but about the whole hybrid chip.While IB may be a bit faster in CPU tasks(by how much and what is noticeable?),a huge difference in games and GPGPU will make BD Fusion much better product.

PS still waiting for your IB-E roadmap link and H2 '12 launch date

[-Sweeper_]

Fusion (2012) will be a much more balanced package, offering fast BD cpu cores and good graphics... but will that make it a better product? what I'm saying is that it actually depends on your specific tasks.. IB should offer very fast cpu cores + decent graphics... if you rely on the cpu side (and a lot of people do) get IB or komodo... if you rely on graphics get fusion (or simply buy a discrete graphics card)... now the question is, if you rely on both, what chip to get?
I say its pretty soon to answer this...

ps: I saw it in one of those pcwatch roadmaps (sent you a PM with the link), but both of us know its no surprise that IB-EP is coming around 1 year after SB-EP and zambezi will have to fight LGA 1155 IB for some time before komodo arrives...

[Nintendork]

It's not about hardcore gaming, just play games at a decent level "similar to what i see on consoles".

SB/IB IGP can run modern and popular games "smoothly" at low details, HD6450 is twice as fast as HD3000 but it's also slow if they want to see that "wow factor" of pc games. In low they look more or less like a Wii/PS2.

Fusion Llano brings that "HD5570/5670" hall of famer gpu quality. Not expensive, no need of a quality PSU and can play any game with mid-high/high details as long as they don't force resolution (1366x768/1400x900/1600x900).

And no, SB IGP are not enough for gaming(low cost are HD2000 which is even worse). Now if Joe listens to what the store salesman says or just buy blindessly a brand pc, they'll probably be disappointed the moment they ask some friend or forum.

[undone]

BD Fusion is going to ramp up Llano's SP count,

Don't forget not only cpu part can be alter to new arch, the gpu part could be, too.
In 2012 the HD6000 series will be replaced by the new HD7000 architecture, nobody can sure AMD won't do the same in its APU.

[informal]

Well none of those products are available so we don't really know what will happen.But the fact is that a whole package will be judged and evaluated,that's all.

SB-E + 1 year should indeed equal launch date for IB-E,but there is no really data and SB-E will be a Q4(late or early Q4?).

@ undone

Yes,I expect them to use VLIW4 with Trinity Fusion,a move from present 5way SPs.

[undone]

Yes,I expect them to use VLIW4 with Trinity Fusion,a move from present 5way SPs.

Does amd ever announced the Llano is VLIW5 based?

[Formula350]

now why would you say that

I think he's just trying to stir the pot :P No proof, or even link to speculation implies a hollow statement.

I smell a contradiction here

I smell an Intel troll trying to get a rise out of people, simply due to his own boredom

[Warwian]

Good analysing. If it's true it will be a good sign for amd. Even if the Zambezi single core abillity can only reach Westmere, the higher frequency will be more than enough to fill the gap. 95w with 8 core @3.6-4.0ghz is awesome.

Such high frequencies would indeed be very promising for Bulldozer performance if true. Though I am bit worried that the high frequencies will end being Bulldozer's only defense against higher IPC efficiency of SB.
What I'm saying is that I'm hoping Bulldozer is able to compete with SB in terms of IPC efficiency aswell and not just by having a very high frequency. But I have a feeling it might end up that way. But as always we'll have to wait and see.

Selling CPUs with a potential frequency of above 4GHz should be great from a marketing point of view though. 4GHz is after all something rather magically to be able to pass, especially from a processor in stock configuration.

[Sn0wm@n]

I think he's just trying to stir the pot :P No proof, or even link to speculation implies a hollow statement.

probably what it is

[informal]

Does amd ever announced the Llano is VLIW5 based?

Not really but they said,back in the day of HD5000 series, they will use current Radeon cores in their upcoming Llano device.That's why I concluded it will be vliw5.It may end up being something else ,we don't know. But a switch to vliw4 in BD Fusion will cut the GPU die area a bit(10% according to HD6800 available info) while maintaining same performance.

Such high frequencies would indeed be very promising for Bulldozer performance if true. Though I am bit worried that the high frequencies will end being Bulldozer's only defense against higher IPC efficiency of SB.
What I'm saying is that I'm hoping Bulldozer is able to compete with SB in terms of IPC efficiency aswell and not just by having a very high frequency. But I have a feeling it might end up that way. But as always we'll have to wait and see.

Selling CPUs with a potential frequency of above 4GHz should be great from a marketing point of view though. 4GHz is after all something rather magically to be able to pass, especially from a processor in stock configuration.

We don't know the IPC story in Bulldozer's case,yet.

[Nintendork]

Does amd ever announced the Llano is VLIW5 based?

Fusion Llano
Redwood core (High End) HD5000, VLIW5
Caicos core (Low End), HD6000 VLIW5 (only HD6900 are VLIW4)

[Warwian]

We don't know the IPC story in Bulldozer's case,yet.

Yes, I realise this but it doesn't make me less worried.

Anyways, anyone else have any thoughts on my proposed frequencies? Do they even seem remotely likely?

[tifosi]

Yes, I realise this but it doesn't make me less worried.

Anyways, anyone else have any thoughts on my proposed frequencies? Do they even seem remotely likely?

3.7 Ghz stock on a 45nm node (although matured process) will make me think that AMD will do something ridiculous sometime soon Ghz wise

Quad cores...i expect them to hit north of 4 Ghz on 32nm and anything less on launch means... higher speeds will follow just like with Phenom II... from 3 Ghz at the top end to 3.7 Ghz at the top end.

Six core processors... they started with 3Ghz and are at 3.3Ghz (and then some with turbo)... I'd expect them to launch with about 3.5 Ghz at the least with the TDP in mind.

So yes... if i'm looking forward optimistically towards this smaller node... i can't be faulted...

There's only one thing though... marketing may want to use some of the capability clocking wise for advertising turbo... :P

[Olivon]

probably what it is

Nope. I'm really serious ...

But we will see

[FlanK3r]

give reason, its not reason...