Bulldozer "delayed" until September 2011 (Rumor)

[Nemesis 1]

Erm, AMD is one of GF's big customers, if not the biggest. Volume of business given does makes a difference in how you charge the potential customer. This is the norm across many industries...

On topic now... there was/ is a guessing game going on about what, how and when... at SA forums... Now with Charlie also hinting that Server is being ramped up... i look at my previous post here... where shipping of server BD products starts in Q2, 2011. I'd not blame AMD for playing along with OEM's on this one. MagnyCours was a good chip for its time and well, nearly a bloody miracle for its size... and the fact that it was on 45nm. Now that 12 core chip on a 45nm node is doing 2.5 Ghz. However, as brilliant as that chip was, there were not so many systems from partners which came along initially. Infact i read more articles about systems built with MC chips in this year of BD launch than i read in the launch year of MC chips.

Given that they're doing 2.5 Ghz on a 12 core chip on 45nm node, i'm certain that AMD could easily manage upto 3 Ghz with 16 cores in the same, or lower thermal envelope. How is that possible? BD cores are smaller, and well on a smaller node. It is that simple. What does that mean? OEM's will want more of those and AMD will have to supply. I suspect that OEM's and AMD are working together on the server chips, even as we bicker on forums

Desktops, sure its an important market, but Llano will do AMD's bidding with OEM's for a bit... Oh yes, they're to sell 8+ million of those chips this year... And i honestly didn't expect to see Trinity this early on. I mean on Llano intro day... It was a little silly i guess... It is like, 'hey you want this candy, but wait! we have better one... come in a week again.' They sure could and may be in hindsight should have waited till their E3 conference. But hindsight's a bitch To think about it... i'd have loved to read about what socket it fits into (if it is the same as FM1, it would have built confidence in people's heads)...mem support... graphics part... all that at once... As that would have laid the smackdown proper, on Intel! Not to say that they aren't crapping their pants already... but just that AMD could have played their hand better. Gotten more mileage out of the coverage.

AMD has lost even more server market share after its released . SO MC did nothing for AMD. A miracle when intel did this you guys called it a hack job , Now 6 years later its a miracle . OH my.

 

[Nemesis 1]

Did that AMD guy really just show a heatspreader and say "look TRINITY"? Seems like a half-hearted attempt to distract from the delay news. I mean they are really pulling a "Pork Shoulders" nVidia move, they steal nVidia's marketing manual? Unlike Nemesis I do think AMD really has a nice thing with these hybrid chips now out of development hell. I can only hope for competition's sake that AMD's marketing department is ready to exploit this 6-9 month lead on hybrid products. I believe Intel confirmed IB will have up to 32 EUs and the clocks will probably be impressive on their 22nm process. That's where I think this desire to shout about Trinity comes from, they need that selling in volume several months before IB launch.

But come on, shouting about a hybrid (APU) product that's based on the cores of the chip you just announced as delayed, I just don't come away with the positive impression AMD wants me to have. Show me BD numbers then talk up Trinity.

Intels chip is the First High bread . It lacks CL and DX11 . Or simply stated it lacks DX11. Thats not a big deal at this time . Point is Intel got to SoC first on the desktop

 

[Dravic]

Wasn't this Fusion is a wild success even beyond AMD's dreams news already talked about by AMD.

AMD's Raymond Dumbeck announced this week that AMD sold 5 million "Fusion" chips in the first five months of the year. "

Brazos has clearly exceeded AMD's wildest expectations. With support from Dell Inc. (DELL), Sony Corp. (TYO:6758), Acer Inc. (TPE:2353), Hewlett-Packard Company (HPQ), and Lenovo Group Ltd. (HKG:0992) AMD's Fusion project is seriously threatening Intel's netbook and notebook processor offerings.

http://www.dailytech.com/AMD+Fusion...el+in+the+Notebook+MidMarket/article21763.htm

I too would ramp up the successor of chip I can't keep stocked over the one that hasn't launched yet and is still going though refined steppings.


If three months ago your product meetings went like this what would you have said..

Hey boss, Brazos sold out, and we cant make them faster enough. Also, Bulldozer needs a respin or two…

Where do you put your desktop eggs.. 32nm successor to Brazos, or BD?

Sure sucks for us, but it's the more sound company move. IMHO

 

[Khato]

Sorry that list is flawed. Chipset changes that effect processors should be added (i.e. FSB), server sockets added on one side, and other minor errors.

Why should chipset changes be added? Calling it flawed for a reason which wasn't part of its stated intent is nonsensical. That said, chipset changes actually didn't drive any of the incompatibilities between LGA 775 'generations' - it was entirely VRM issues. Why can I say that? Because the chipsets easily supported FSB speeds a generation ahead. Further, as was stated, it's a list of "sockets targeted towards consumers/enthusiasts" - while socket 940 and socket F were primarily intended for the server market, they were also used in AMD's high-end enthusiast platform. (Yes, I love how some complain about Intel now using the same strategy while turning a blind eye to the fact that it was AMD that introduced it.)

Some AM2 boards can run AM3 processors, show we then consider AM2, AM2+, and AM3 as the same? No. Cause not all could run.

Yup, because some AM2 boards were made alongside AM3 boards. But some original AM2 boards can't support AM2+ processors... aka, it's a mess every bit as bad as the 'generations' of VRM support with LGA 775. As said in my original post, I freely acknowledge that LGA 775 could be considered to be the equivalent of 3-4 sockets, though at least there newer versions were compliant with older processors, unlike some of the AM* progression.

Other thing to keep in mind are the reasons for all the socket changes... Here's a hint, compare the performance improvement from LGA 775 intro to LGA 1155 against that of socket 939 to AM3. It's easy to just make minor alterations to the socket when you're not changing much with the processor.

 

[Nemesis 1]

Well lets wait on 2nd Qt earning results the numbers that matter. Intels 1st qt was great . Amds was were AMD plays not really good for selling so many cpus as you stated .

 

[Bearach]

Why should chipset changes be added?

Because chipsets also played a part in the processors supported. I only said chipsets that were a fundamental reason the older platforms were not capable of running a processor, not all. So, yes, your chart is flawed on the basis that you used it.

 

[Vesku]

Don't think you want to go down that road Nemesis1, do you recall as AMD moved to their own chipset/motherboards, because of the Intel slot patent shenanigans, motherboard makers were worried about displaying AMD boards in public for fear of upsetting the Intel gorilla (losing "ad" money, preferred pricing?). Intel already got rid of that baggage in the US for some pocket change, no need to try to reverse jab with it.

 

[Nemesis 1]

IBM insisted on a second X86 supplier . Than AMD reverse engineers intels design and steals it . The socket change and cash per amd unit sold is all intel won other than AMD couldn't out source . Thats all changed now . GF has the foundries and AMD doesn't have to pay intel for every chip it sells. If I was intel I wouldn't want a leach sucking off my chipsets

 

[Khato]

Because chipsets also played a part in the processors supported. I only said chipsets that were a fundamental reason the older platforms were not capable of running a processor, not all. So, yes, your chart is flawed on the basis that you used it.

Which doesn't work out all too well as an argument since the only place where chipset support played a part was on LGA 775. And in that case, it wasn't chipset support that would cause a newer processor to not function in an older motherboard, as my underclocked E8600 running at an 800 FSB in my home server would imply. That said, the VRM issues with LGA 775 definitely are a valid reason to consider that socket as having multiple iterations of the same form as the AM* sockets... which I acknowledged in the initial post... and which merely bring Intel up to having the same number of sockets as AMD since they diverged (assuming you count 3 iterations of LGA 775, Intel would have one more if you count there as being 4 iterations of it, which I believe is more correct. Oh, and that's counting the pending socket introductions of this year too.)

Point being, the argument that AMD is so nice and considerate of their customers while Intel loves forcing everyone to buy a new motherboard with every processor is... not very factual.

 

[Vesku]

Intels chip is the First High bread . It lacks CL and DX11 . Or simply stated it lacks DX11. Thats not a big deal at this time . Point is Intel got to SoC first on the desktop

Sure their resources got them there first, but AMD does have an edge in the related tech. AMD needs very much to avoid any further delays in APU releases at this critical juncture. I think Bulldozer must have some teething issues but I also don't think the Llano demand PR spin is completely hollow. Making a big impression on consumers before IB shows up in volume is critical. Intel might have started a bit later but they execute much faster. Brazos was a good start and all signs point to heavy interest by OEMs in Llano and even knocking on AMD's door for tablet specced Brazos chips. Too bad for AMD that the global finance market is still recovering, they could really use some extra cash for a PR blitz on Llano launch. Instead they have to rely partly on partners advertising budgets which is the opposite situation of Intel.

It's actually pretty amazing AMD hasn't gone belly up.

 

[sandorski]

It could all be Lowering Expectations, then releasing an amazing product. My inner Fanboy hopes anyway. Way back in the before time, I recall how the negative rumours increased in volume as the release of Athlon drew nearer. Then it was released and everyone was thoroughly impressed.

Will that happen again? Meh, I doubt it, even if BD/Lano turn out to be very good, the era around the release of the Athlon was a very different time and AMD was still very poorly established as a solo player in the Industry. It's very possible that the true affect of a successful BD/Lano lineup may not be truly known for a few years. Simply because the PC landscape is in a state of transition, that being from CPU Performance to APU Performance.

 

[Vic Vega]

I don't care that it's delayed. These CPU makers are out of touch.

 

[Nemesis 1]

IF you look at AMDs entire history . They haven't ever really made any money. They did what they could with one fab but just couldn't produce enough at that node to make hammer the success it should have been the second fab was to little to late. If APU drives AMD into the future the company really is ATI.

 

[Vesku]

Point being, the argument that AMD is so nice and considerate of their customers while Intel loves forcing everyone to buy a new motherboard with every processor is... not very factual.

The AMD socket good will comes mostly from the AM2->AM2+->AM3 change being very new CPU in old Mobo friendly with AM2+->AM3 being especially well supported. If you take into account AMD's lag behind Intel and compare the AM platform to the socket 775 platform I think the AM platform was clearly more drop in CPU friendly. It is only now that AMD is reaching the same point in their product line as Intel did with having to support chips with on die IGP and ones with no on die IGP with the Llano socket and AM3+ socket AMD is finally reaching that same point of divergence. You can bet if AMD thought they could produce a good enough CPU they'd also have a performance socket in the pipeline equivalent to 1366, 2011.

TLDR: Socket 775 had more compatibility issues than the AM socket, enough to make AM owners chuckle a bit at all the "help me figure out what my 775 board can handle" threads.

 

[Nemesis 1]

It could all be Lowering Expectations, then releasing an amazing product. My inner Fanboy hopes anyway. Way back in the before time, I recall how the negative rumours increased in volume as the release of Athlon drew nearer. Then it was released and everyone was thoroughly impressed.

Will that happen again? Meh, I doubt it, even if BD/Lano turn out to be very good, the era around the release of the Athlon was a very different time and AMD was still very poorly established as a solo player in the Industry. It's very possible that the true affect of a successful BD/Lano lineup may not be truly known for a few years. Simply because the PC landscape is in a state of transition, that being from CPU Performance to APU Performance.

Who uses APU other than AMD . Intel not NV not . So APU is not the future. Its a name AMD uses nothing more.

 

[Bearach]

Which doesn't work out all too well as an argument since the only place where chipset support played a part was on LGA 775. And in that case, it wasn't chipset support that would cause a newer processor to not function in an older motherboard, as my underclocked E8600 running at an 800 FSB in my home server would imply. That said, the VRM issues with LGA 775 definitely are a valid reason to consider that socket as having multiple iterations of the same form as the AM* sockets... which I acknowledged in the initial post... and which merely bring Intel up to having the same number of sockets as AMD since they diverged (assuming you count 3 iterations of LGA 775, Intel would have one more if you count there as being 4 iterations of it, which I believe is more correct. Oh, and that's counting the pending socket introductions of this year too.)

Point being, the argument that AMD is so nice and considerate of their customers while Intel loves forcing everyone to buy a new motherboard with every processor is... not very factual.

Okay, so all the 440 series ran every Pentium II? I'm not making an argument at all in AMD's favour, just that your chart is by far inadequate to show the real changes.

440BX ran some slot 1, and some socket 370. To say a chipset makes no difference is wrong, the 440LX could only run up to 66Mhz bus processors but used slot 1, the 440BX did up to 100Mhz. It does suggest there has been more changes both sides that you're not taking into account.

I'm only showing one flaw, there are plenty more. I'm not arguing about your original argument being that AMD are not better than Intel, I'm arguing the facts that you've left out.

Slot 1 saw more than 2 chipsets, but at least 2 changes, because some chipsets couldn't run above 66Mhz FSB processors. Does that not count as a change in platform also? How does chipsets that actually effect you running a CPU not count?

It is flawed no matter how you look at it, unless you count chipsets that changed something key in the support of a new-gen processor.

 

[sandorski]

Who uses APU other than AMD ..

Exactly. If AMD is right, Fusion is the Future.

 

[sandorski]

Okay, so all the 440 series ran every Pentium II? I'm not making an argument at all in AMD's favour, just that your chart is by far inadequate to show the real changes.

440BX ran some slot 1, and some socket 370. To say a chipset makes no difference is wrong, the 440LX could only run up to 66Mhz bus processors. It does suggest there has been more changes both sides that you're not taking into account.

I'm only showing one flaw, there are plenty more. I'm not arguing about your original argument being that AMD are better than Intel, I'm arguing the facts that you've left out.

Slot 1 saw more than 2 chipsets, but at least 2 changes, because some chipsets couldn't run above 66Mhz FSB processors. Does that not count as a change in platform also? How does chipsets that actually effect you running a CPU not count?

It is flawed no matter how you look at it, unless you count chipsets that changed something key in the support of a new-gen processor.

Indeed. Intel was Chipset Happy, to be kind. They got that reputation for good reason and AMD has the opposite reputation for good reason as well.

 

[Vesku]

Who uses APU other than AMD . Intel not NV not . So APU is not the future. Its a name AMD uses nothing more.

Are you just referring to the name "APU" now or heterogeneous computing in general, because Intel clearly thinks heterogeneous computing is the most likely future path and nVidia seems to be staking their existence on it? Pretty petty to harp on their choice of branding if you are really just referring to the "APU" designation.

 

[Arachnotronic]

Also, it's interesting to note that Intel has their MIC stuff marketed as a "co-processor" for server. So they definitely believe in this kind of heterogeneous environment.

 

[Nemesis 1]

Sure their resources got them there first, but AMD does have an edge in the related tech. AMD needs very much to avoid any further delays in APU releases at this critical juncture. I think Bulldozer must have some teething issues but I also don't think the Llano demand PR spin is completely hollow. Making a big impression on consumers before IB shows up in volume is critical. Intel might have started a bit later but they execute much faster. Brazos was a good start and all signs point to heavy interest by OEMs in Llano and even knocking on AMD's door for tablet specced Brazos chips. Too bad for AMD that the global finance market is still recovering, they could really use some extra cash for a PR blitz on Llano launch. Instead they have to rely partly on partners advertising budgets which is the opposite situation of Intel.

It's actually pretty amazing AMD hasn't gone belly up.

Only in so far as they got ATi which gives them a head start . But to make all this work requires expertize in compute VLIW which intel leads . For CL c99 is the standard Intels C++ compiler is better than AMDs 64 compiler. Intel has vec for AVX and how avx relates to Cl well find out soon enough . I could easily and rightfully so after looking at intel roadmaps infer that enhanced AVX for IB is a move from 256 bit to 512 . As you can read intels AVX white paper intel clearly states that AVX can scale from 256bit all the way to 1024 bits .

So I see it like this .

SB 256 bit
IB 512 bit
HW 1024 bit

 

[Genx87]

Exactly. If AMD is right, Fusion is the Future.

Fusion type products arent the future, they are the now. Intel owns 60% of the graphics market for a reason.

 

[Khato]

I'm not arguing about your original argument being that AMD are not better than Intel, I'm arguing the facts that you've left out.

I'll certainly cede the point that that chart is leaving out a number of other issues that would cause incompatibility/performance degradation within a single socket. After all, an initial socket A motherboard with its 100MHz base clock EV6 FSB wouldn't play nice with a Palomino and it's 133 MHz base clock FSB, which in turn would have issues with 166 MHz Thourorughbreds, which again didn't work with the 200 MHz Bartons. I think we're all glad that the FSB has fallen to the wayside eh?

Anyway, my original point was just to provide some semblance of a history of socket changes to show that it's foolish to not be worried about future compatibility with AMD while expecting it of Intel. If anything, current indications are that the Intel platform is going to be more stable given that Ivybridge will supposedly work in current Sandybridge motherboards. (Yeah, I doubt it'll actually happen, just like I doubt that the forthcoming FM1 socket will work with anything that comes after Llano.)

 

[GammaLaser]

Bulldozer not delayed after all?

 

[Vesku]

No doubt about it, AMD could use more compiler expertise. Same reason nVidia was able to keep ahead of ATI until AMD caught a good headwind while nVidia hit a calm patch, Another limitation of being the much smaller company in a two player market.