AMD Phenom CPUs to see further delays

[Unknown Soldier]

AMD has recently notified its partners that the launch of higher-end quad-core Phenom processors, including the 9700 and 9900, will be postponed to the second quarter of 2008 from the original schedule of early 2008, according to sources at motherboard makers.

However, whether AMD's triple-core Toliman series CPUs will also see delay will be the key decision for the company, since Toliman offers a high price/performance ratio compared with Intel's quad-core CPUs. A delay for Toliman will hurt AMD the most, noted the sources.

The sources commented that the reason for the delay of 9700 and 9900 is because AMD has not yet been able to solve the translation lookaside buffer (TLB) erratum found in the chips. However, they added in saying that, in the long-term, AMD's decision is correct since pushing products that are not ready will only hurt the company more.

News Source: http://www.digitimes.com/mobos/a20071224PD200.html

-----------------------

Ouch .. that's gotta hurt.

US

 

[K.I.L.E.R]

The current 9900 in reviews cannot flatten any of Intel's Quad Cores, and yet they will release them in the second quarter next year?

 

[Unknown Soldier]

Like I said, it's gotta hurt.

US

 

[Albuquerque]

And because of the this launch delay, all microprocessor development slows down. See also: Intel delaying Penryn quad cores because there's simply nothing anywhere near competing with the curren Conroes.

 

[Arun]

See also: Intel marketing puts fantasy spin on quad-core Penryn Errata and pretends it's an intentional delay (the original hardware.fr/behardware.com delay leak clearly indicated an Errata).

 

[AlexV]

See also: Intel marketing puts fantasy spin on quad-core Penryn Errata and pretends it's an intentional delay (the original hardware.fr/behardware.com delay leak clearly indicated an Errata).

Are you certain they were right?There's been so much spin on that particular Intel delay that for all we know it there could've been an alien plot or some similar thing behind it. The latest noise from Xbit was that 4-layered PCB mobos and the Yorkies don't work all that well together, causing FSB integrity issues, and Intel decided to do a respin(the high-end boards supposedly have no issue with the em').

Anyway, if these rumours of a further Phenom delay are true...it's like this huge overweight elephant is taking a dump on AMD's fan, and the dude has a serious case of gastrointestinal distress. This does make you wonder why are they fretting so much over such a supposedly insignificant bug(in the desktop space)?Have the TLB fix, release higher speed bins and let the users choose wheter or not to enable the fix. For server-space virtualization is very important so yes, I can see why they want fixed hardware there but...Maybe the thing isn't as trivial/impossible to reproduce as AMD and dudes like Kyle from [H] claim?Are they suicidal?Or are they simply unable to produce higher speed Phenoms within reasonable TDPs/in any significant volume?Overall, this is bad news.

 

[The_Wolf_Who_Cried_Boy]

See also: Intel marketing puts fantasy spin on quad-core Penryn Errata and pretends it's an intentional delay (the original hardware.fr/behardware.com delay leak clearly indicated an Errata).

Additional to Morgoth's post, why are Xeon 5400 series processors available (I just did a quick check on IBM and Sun's pages) if there is a flaw in the processor rather than motherboard based problem. Why would Intel risk it's reputation with no competative pressure to do so?

 

[K.I.L.E.R]

So AMD stuffs up and the whole microprocessor world is delayed by a year?
I'm seriously thinking of killing myself.

 

[AlexV]

So AMD stuffs up and the whole microprocessor world is delayed by a year?
I'm seriously thinking of killing myself.

Think from an economics point of view:it's the logical thing to do. Having frequent releases of new parts that cannibalize older ones' sales, high R&D costs associated with the need of constantly one-upping the competition, high costs of pushing the envelope on process technology in order to accomodate your constantly more complex designs simply limits your margins.

When your competitor has about jacksquat that's competitive even remotely, it's illogical to continue trimming your margins in such a fashion. You can afford slowing down, and given the rather longish cycles in the CPU business, it's not like AMD will turn things around in a month and get things rolling again...it may take quite a while:|

 

[K.I.L.E.R]

So back to the old days eh? I truly hope video cards aren't going in that direction.

 

[AlexV]

So back to the old days eh? I truly hope video cards aren't going in that direction.

Aren't they?Why would nVidia keep pushing extremely tight schedules, aggresive designs, aggresive processes, take risks, cannibalize its sales every 6 months, when they have no real competition threatening them(yes, yes, the 3870 is a nice card, but it's certainly not something to have them sweating), and when the existing competition itself isn't overly aggresive?

Going back to the old days in terms of CPUs...dunno, Intel has been fairly constant in its timelines for a new architecture ever since the 486. The difference will be is that you won't be having quick-fire salvos like the P4 and by extension the Core 2 were, with higher speed parts being pushed very aggresively, process shrinks being done on "short" notice and so on, things will fall back into a more ordered and calm scenario.

 

[Albuquerque]

See also: Intel marketing puts fantasy spin on quad-core Penryn Errata and pretends it's an intentional delay (the original hardware.fr/behardware.com delay leak clearly indicated an Errata).

Clearly? They already have quad core Penryns out there, overclocking like mad, and yet there's some unsubstantiated rumor dressed up as fact that some HUGE errata affects all of these chips? Then what about the eval units that are already out and working fine? Why didn't it affect those?

All chips have erratum, ALL of them. I don't see any major indicators of failure anywhere. Rather, I see absolutely ZERO need for Intel to push anything faster out, especially in such a "niche" market as quad cores are, and especially since they already have full product lines (quad core Conroes) that aren't selling that fast anyway.

You'll notice that the dual core Penryns are coming out on time, and they're the exact same architecture as the quads. But also keep in mind: dual core processors are still a fast market, much faster than quads. There's more product to sell and more money to make in that arena, which is why they'll sell which is why they make sense to get out.

 

[demonic]

Whilst I do feel sorry for AMD. I wish Intel knocks out the chips in JAN. I really am looking to a cool QUAD Core processor for a laptop in '08.

 

[3dilettante]

Clearly? They already have quad core Penryns out there, overclocking like mad, and yet there's some unsubstantiated rumor dressed up as fact that some HUGE errata affects all of these chips? Then what about the eval units that are already out and working fine? Why didn't it affect those?

There is an issue with chip to chip transfers over the shared bus that is, according to Intel, triggered by a certain combination of transfers. According to Intel's latest spec revision, it has only been done in a testing environment with certain hardware configurations.

If any issure or erratum matches up with the rumors, it would be that one.

A few eval units would most likely not be enough to find an issue such as this, and there are too many other factors, such as the brevity of the testing and the probability that a lot of testing is done with overclocking or preview tests, where instability would likely be chalked up to business as usual with overclocking or non-final hardware.

It's not necessarily true that there wasn't any crashing, but that it's not considered unusual for non-release hardware to not be perfect.

All chips have erratum, ALL of them. I don't see any major indicators of failure anywhere. Rather, I see absolutely ZERO need for Intel to push anything faster out, especially in such a "niche" market as quad cores are, and especially since they already have full product lines (quad core Conroes) that aren't selling that fast anyway.

Given the lead time on manufacturing, Intel would have likely already made a good number of quad-core lots.
That is one motivating factor. If the parts are in the warehouse, there's no reason to keep paying for the shelf space.

Another factor that comes up is Intel's own organization and its customers' planning. Intel would have given advance notice of product launches so its customers and board partners are not caught flat-footed. Once the momentum builds for a launch, the launch itself becomes a motivating factor.

You'll notice that the dual core Penryns are coming out on time, and they're the exact same architecture as the quads. But also keep in mind: dual core processors are still a fast market, much faster than quads. There's more product to sell and more money to make in that arena, which is why they'll sell which is why they make sense to get out.

It's also the case that the quad-cores have a feature the dual-cores don't: the shared bus over the MCM. It's that feature that most of the rumors I've seen apparently focus on, and the one Intel is apparently addressing in a specification update.

There is a signaling margin problem with MCM chips on some boards, possibly the cheap mass-market boards (which also might explain why higher-priced server boards seem unaffected).

It is unclear from this news whether the boards in question are actually out of spec, which is a key point.

If Intel's chips don't meet reliability standards for its own specifications once the product is on sale, it is a whole different can of worms than a problem on evaluation hardware.