AMD, present time.

Thanks for the read, I appreciate it, does seem as if Phenom is not up to par, there goes another year for AMD as the second processor of choice, some in 2006, 2007 now 2008 ugh:mad:
 
Kyle has stated that he thinks 2008 will be a turn around year for AMD,I dont think he's right.I see more of the same.

If you want to know the future,look to the past.
 
I think AMD is comming but slower than we want. Personaly I am waiting for the Octal core cpu's @8ghz and the Octal core graphic cards. Then it's time for a new build.:eek:
 
Kyle has stated that he thinks 2008 will be a turn around year for AMD,I dont think he's right.I see more of the same.

If you want to know the future,look to the past.

so nvidia's going to release more FX's ?
 
Kyle has stated that he thinks 2008 will be a turn around year for AMD,I dont think he's right.I see more of the same.

I agree.
Nothing on the roadmap of AMD is anything more than just a respin of current technology (and things just seem to disappear off the roadmap...).
Phenom isn't magically going to become an architecture with superhigh IPC, excellent clockspeed scaling and very low power consumption with just some more steppings or even a 45 nm shrink. As with all respins/dieshrinks, IPC will be virtually unchanged, clockspeeds will increase, but not dramatically, and power consumption will decrease, but again not dramatically.

In the meantime, Intel will also be respinning and shrinking its technology... AND they will introduce Nehalem. So by the looks of it, AMD isn't going to become more competitive in 2008. Phenom was their shot at regaining their competitive edge, and it failed. They're now in damage control mode, waiting for the inevitable to happen: quadcores in the mainstream and budget market.
(I'm not talking about performance here, Intel has the fastest processors, but things are going way further than that... Intel can deliver more bang for the buck with their quadcores because production costs are lower, mainly because the design is more efficient).

As I've said before, AMD makes nice chipsets, but as a platform it doesn't mean that much when you don't have a killer CPU to go with that chipset. I'm not going to buy an inferior CPU just because I like a chipset. I select my CPU first, and then I choose the best chipset to go with that, for my needs. I think most people work that way.

As for ATi... even if they do come up with a winner GPU, the GPU market is far smaller than the CPU market, so I don't think it can save the company as a whole.

I think AMD should reinvent Phenom. Not as a high-end part... They failed miserably there, and the videos on Youtube are a painful testament to AMD's gross underestimation of Intel.
They should reinvent it as a good pricefighting part. Perhaps they should drop the native quadcore. The dualcore version may stand a good chance, as long as dualcores last in the mainstream... Perhaps a native triplecore would also be okay, because it would solve some of their production and scaling problems with the smaller die size. Other than that, they have to get MCM going, and ofcourse 45 nm.
The thing I don't get... Why didn't AMD see this coming? Many people saw it coming. The moment AMD caught their first glimpse of the new Core2 line, and especially the MCM-based quadcores, alarm bells should have been ringing all through AMD's headquarters, and AMD should have made the decision there and then to go pricefighting instead of high-performance. They must have known that Core2's IPC was hard to beat, and that Intel would have its 45 nm process underway by the time AMD finished its first quadcore design on 65 nm. It has mismanagement written all over it.
I don't see this 'reinvention as pricefighter'-trend on the roadmaps yet though.
 
I agree.
Nothing on the roadmap of AMD is anything more than just a respin of current technology (and things just seem to disappear off the roadmap...).
Phenom isn't magically going to become an architecture with superhigh IPC, excellent clockspeed scaling and very low power consumption with just some more steppings or even a 45 nm shrink. As with all respins/dieshrinks, IPC will be virtually unchanged, clockspeeds will increase, but not dramatically, and power consumption will decrease, but again not dramatically.

In the meantime, Intel will also be respinning and shrinking its technology... AND they will introduce Nehalem. So by the looks of it, AMD isn't going to become more competitive in 2008. Phenom was their shot at regaining their competitive edge, and it failed. They're now in damage control mode, waiting for the inevitable to happen: quadcores in the mainstream and budget market.
(I'm not talking about performance here, Intel has the fastest processors, but things are going way further than that... Intel can deliver more bang for the buck with their quadcores because production costs are lower, mainly because the design is more efficient).

As I've said before, AMD makes nice chipsets, but as a platform it doesn't mean that much when you don't have a killer CPU to go with that chipset. I'm not going to buy an inferior CPU just because I like a chipset. I select my CPU first, and then I choose the best chipset to go with that, for my needs. I think most people work that way.

As for ATi... even if they do come up with a winner GPU, the GPU market is far smaller than the CPU market, so I don't think it can save the company as a whole.

I think AMD should reinvent Phenom. Not as a high-end part... They failed miserably there, and the videos on Youtube are a painful testament to AMD's gross underestimation of Intel.
They should reinvent it as a good pricefighting part. Perhaps they should drop the native quadcore. The dualcore version may stand a good chance, as long as dualcores last in the mainstream... Perhaps a native triplecore would also be okay, because it would solve some of their production and scaling problems with the smaller die size. Other than that, they have to get MCM going, and ofcourse 45 nm.
The thing I don't get... Why didn't AMD see this coming? Many people saw it coming. The moment AMD caught their first glimpse of the new Core2 line, and especially the MCM-based quadcores, alarm bells should have been ringing all through AMD's headquarters, and AMD should have made the decision there and then to go pricefighting instead of high-performance. They must have known that Core2's IPC was hard to beat, and that Intel would have its 45 nm process underway by the time AMD finished its first quadcore design on 65 nm. It has mismanagement written all over it.
I don't see this 'reinvention as pricefighter'-trend on the roadmaps yet though.


Hell of an outlook.
 
Phenom was their shot at regaining their competitive edge, and it failed. They're now in damage control mode, waiting for the inevitable to happen: quadcores in the mainstream and budget market.

Perhaps Phenom is that mainstream quad core part?

As for ATi... even if they do come up with a winner GPU, the GPU market is far smaller than the CPU market, so I don't think it can save the company as a whole.

The integrated GPU market is huge, and that's where ATi is doing fine.

Why didn't AMD see this coming?

They've been busy with mergers and game consoles. Sorta like how nvidia was busy with the xbox in the FX days. Of course they saw it coming, they just can't do much about it until they've freed up some resources.

They should reinvent it as a good pricefighting part. Perhaps they should drop the native quadcore. The dualcore version may stand a good chance, as long as dualcores last in the mainstream... Perhaps a native triplecore would also be okay, because it would solve some of their production and scaling problems with the smaller die size.

Way ahead of you. They announced tri-core cpus a couple months ago, where its pretty much a quad with a disabled core (so they can pick out the quads with 1 bad core and still sell them). Arriving march-ish? I don't remember.
 
Don't forget that ATi is leading nVidia by a very huge margin in the console market, both 360 and Wee uses ATi graphic.
 
Nvidia was working on the PS3 graphics but they had no trouble releasing an 8800 GTX/GTS while ATI was still using 1900XTX to compare
 
Kyle has stated that he thinks 2008 will be a turn around year for AMD,I dont think he's right.I see more of the same.

If you want to know the future,look to the past.

Did he actually say that? Got a quote, link etc? I remember him having high hopes for Phenom (as did most of us) but it would take a brave soul to predict AMD reversing their fortunes in 2008 based on their current roadmap. In fact, I think by the end of 2008 AMD would be further behind than they are now, due to Nehalem.
 
I think that AMD might be able to tweak IPC somewhat with the 45nm shrink, but I don't think it will be a miraculous turnaround putting them ahead of Wolfdale and Yorkfield. Best case scenario I think is that they will be within a stone's throw of those parts, while worst case is that it will be just that - a die shrink. But AMD would be foolish not to try some IPC tweaks on their next shrink.
 
I think that AMD might be able to tweak IPC somewhat with the 45nm shrink, but I don't think it will be a miraculous turnaround putting them ahead of Wolfdale and Yorkfield. Best case scenario I think is that they will be within a stone's throw of those parts, while worst case is that it will be just that - a die shrink. But AMD would be foolish not to try some IPC tweaks on their next shrink.

Even if AMD can approach Yorkfield levels of IPC (which I highly doubt) they would still have a raw clockspeed deficit to overcome. Even at this early stage, Yorkfields are overclocking to 4GHz+ easily, and yields will only improve with time. I'm not saying there will be any Yorkfields clocked 4GHz at stock, but the potential is there should Intel feel threatened. Right now it feels as if they are letting their foot off the pedal a little and waiting for AMD's next move to determine what they have to do to maintain the 'status quo'.
 
For the low low price of $1100

I'm talking about the scalability of Intel's 45nm process, and the implications for AMD as a result. What does the price of the Extreme Edition CPU have to do with the discussion? :rolleyes:
 
Don't forget that ATi is leading nVidia by a very huge margin in the console market, both 360 and Wee uses ATi graphic.


Thats not making much of a difference in the one area that counts the most !

Profits.


Did he actually say that? Got a quote, link etc? I remember him having high hopes for Phenom (as did most of us) but it would take a brave soul to predict AMD reversing their fortunes in 2008 based on their current roadmap. In fact, I think by the end of 2008 AMD would be further behind than they are now, due to Nehalem.


Dont you read the articles he publishes ? I mentioned weeks ago,how he stated that Intel is out to 'kill' Nvidia and several said 'got a link'.... Dig around and you'll find it.
 
I think that AMD might be able to tweak IPC somewhat with the 45nm shrink, but I don't think it will be a miraculous turnaround putting them ahead of Wolfdale and Yorkfield. Best case scenario I think is that they will be within a stone's throw of those parts, while worst case is that it will be just that - a die shrink. But AMD would be foolish not to try some IPC tweaks on their next shrink.
Unfortunately, die shrinks historically haven't done much to improve the IPC of either company's processor lines. There may be some exceptions but I would have to concur with Scali2 with this prediction. Even in a best case scenario of let's say 10% IPC increase with the 45nm die shrink, would this be sufficient to reach the IPC performance levels of Conroe, let alone Penryn? I doubt it, and as others have stated here and elsewhere, Intel could easily bump up the frequency since their 45nm process is very solid. Intel is easing off the pressure for now, slightly, but they can easily reapply and increase the competitive pressure in the market if they will feel threatened again in the future. They have the resources to do as they will.

So, where does this leave AMD currently? I must reluctantly conclude between a rock and a hard place. I don't see a bright future ahead for their 65nm process since it's so late in coming. The 45nm shrink remains to be seen, I will not judge it prematurely before its release. It may emerge as AMD's savior for its CPU division but if it gets delayed, even a quarter or two, AMD will be left still fighting Intel with 65nm parts while Intel will be nearing the launch of their 32nm products. Not good, especially when compounded with recent rumors floating around that Bulldozer might get delayed indefinitely...

Another point where Scali2 is correct is the expectation of many people that manufacturing a native quadcore instead of a MCM chip was an error in architectural decision for the next processor generation. Many of you will probably disagree with me here. However, if I not being an engineer, had foreseen potential problems several quarters before its release, why didn't AMD? Were they so far advanced with the engineering that they couldn't change their course? It's likely the answer an engineer would reply with, but I'm still left with desiring more information as to why AMD stayed the course when the road ahead had more than its share of potholes. Once again, this leads us back to failed administrative policies.
 
^^^


What are the limits of silicon? If intel proceeds with shrinking won't they hit a wall, and have to figure out what to use in order to shrink their chips even more? Will this give AMD some time to catch up, improve? My guess is that their going to milk it, take it slow till the end of silicon unless AMD amazingly gets it together anf forcing them to fight back. I mean, we did have Intel Delay 3 lines of processors, right?

I heard that after silicon has reached it its limit that discoveries with carbon have proven to continue CPU development. Is this true,have you heard otherwise?
 
Perhaps Phenom is that mainstream quad core part?

No, it's clearly designed as a high-end part. Which means that it's expensive to produce (large die, low yields, high power consumption etc), has lots of features, and SHOULD be the top performing part of the market.
And that's where AMD is hurting. It costs them a lot of money to build these CPUs, but they can't get enough performance out of it to give the CPU the pricetag that it needs for AMD to get a nice profit.
Intel controls AMD's prices, because Intel has the fastest CPUs.

The integrated GPU market is huge, and that's where ATi is doing fine.

Not nearly as well as Intel. Intel is the largest GPU manufacturer on the market and they don't even do anything outside integrated GPUs at all!
Other than that, it isn't a very profitable market. You can't save your ailing CPU company by selling cheap integrated GPUs with a low profit margin.

They've been busy with mergers and game consoles. Sorta like how nvidia was busy with the xbox in the FX days. Of course they saw it coming, they just can't do much about it until they've freed up some resources.

It shouldn't have been that difficult to bail on the quadcore design and make it a triple core (heck, they're going to sell them as triplecores anyway... the design is obviously flexible enough for it).

Way ahead of you. They announced tri-core cpus a couple months ago, where its pretty much a quad with a disabled core (so they can pick out the quads with 1 bad core and still sell them). Arriving march-ish? I don't remember.

No they aren't way ahead of me. These tricores are actually quadcores with a failing core. So they still have the same huge die size as the quadcores, which still means they quite expensive to produce. I was talking about native tricores, where they'd just remove the fourth core from the design altogether, and end up with a CPU that is more efficient to produce.
 
What are the limits of silicon? If intel proceeds with shrinking won't they hit a wall, and have to figure out what to use in order to shrink their chips even more?

They already 'hit the wall' with the shrink to 90 nm. I think the 'failure' here caused a dramatic change in how they approach dieshrinks now. They are now constantly looking for new materials and other tricks (eg stretched silicon) to improve the chip quality and performance. A dieshrink is not just a dieshrink anymore. It's a completely new process.
65 nm has been hugely successful for Intel, and it seems that 45 nm is again going to be a big success, just judging from the fact that the TDP has decreased dramatically for the Core2 series, while offering larger caches and about 10% more performance per cycle.

But we don't know much about 32 nm yet, or smaller... it will have to end somewhere. Question is: is Intel already working on alternatives?
 
This is what AMD do - their last two top-end cards, the 2900XT and 3870, have competed against the nVidia number 3 of their time, the 8800GTS640 and 8800GT. Nothing they have released on the GPU front has touched the 8800GTX or Ultra, now they appear to be taking the same approach on the CPU front. They are rapidly going to become a bit-player if this carries on.
 
This is what AMD do - their last two top-end cards, the 2900XT and 3870, have competed against the nVidia number 3 of their time, the 8800GTS640 and 8800GT. Nothing they have released on the GPU front has touched the 8800GTX or Ultra, now they appear to be taking the same approach on the CPU front. They are rapidly going to become a bit-player if this carries on.

Sad to say but I think you're right, if they even manage to stay in buisness with all that debt.
 
Why is everyone always comparing a $200 CPU to a $1200 CPU? :p We already know AMD is far behind. Give them a year or two and they might be back in the game. The current situation isn't any good for the consumer with little to no competition. We want these two companies battling neck and neck so we can take advantage of lower prices. Well, we currently have rock bottom prices with AMD, just not Intel because they are far ahead.
 
...

But we don't know much about 32 nm yet, or smaller... it will have to end somewhere. Question is: is Intel already working on alternatives?

FYI they have 600 PHDs working 24/7 to push the boundaries on process tech.That's far far more than any other semi company.
 
My question is when is AMD going to scrap silicon for the new material that has been found to work much better?
 
Not before Intel, IBM, or TSMC. They can't afford to be the early adopters and end up working out the bugs for everyone else.
 
AMD is finished. They're going to get chopped up, and sold off to little semiconductor companies.

Honestly, who's buying AMD right now? And if so... WHY? What must their 2007 revenue look like? What must their 2008 projected revenue look like? It's grim, and it's not good. The only money AMD is going to get are from people with not enough money to upgrade to an Intel platform. I have no hope for AMD this year. They can recover in 2009 if they:

  1. Go 45nm - Yield and low heat dissipation is the name of the game.
  2. Dump their 'true-quadcore' architecture. I don't care what you read, the architecture makes it impossible to financially compete against Intel. If one core goes bad, then what? You'll try to market it as a tricore? When the entire architecture was designed around having four cores? What if two cores go bad? Are we going to get a dual core? What if all but one fails? This is a horrible business model.
  3. Spin off ATI back into the graphics market. I'd like to meet the guy who made this deal happen. I bet he doesn't have the corner office anymore.
 
[*]Go 45nm - Yield and low heat dissipation is the name of the game.

It goes further than that. If Pentium 4 has taught us anything, it has got to be that the chip design far outweighs the importance of the production process. Pentium D and Core2 were built on the exact same 65 nm process, yet the Core2 has far higher performance, with far lower power consumption, and paired with a much smaller die-size.
A clean sweep.
Athlon64 was such a success because it had many of the same advantages that Core2 had: low power consumption, small die size and high performance. This made it possible for AMD to compete with the Pentium 4, despite being behind one generation of the production process.
What AMD needs is an architecture that focuses on getting maximum performance out of minimum die-size. A native quadcore with integrated memory controlled and 3 levels of cache is precisely the opposite.
I suppose Intel is the smart one here, because they chose to use pretty much the same strategy that made Athlon64 a success (focusing on a cheap and efficient design). Since Intel already had the edge in production technology, that leaves virtually no room for AMD to compete at all. Intel is beating AMD at its own game. In fact, if you would try to estimate how many transistors the different parts of the CPU take in, it seems like Core2 doesn't actually need more transistors for its more efficient cores, and even the cache has a relatively low 'transistors-per-byte'-rating.
They really seem to have optimized the design to minimize the number of transistors, and maximize the performance.

Quite a contrast to the Pentium 4, where their strategy seemed to want to take advantage of the fact that they could put more transistors on a die than they would actually need. In fact, the original Smithfield proved that Intel could put two Pentium 4 CPUs on a single die, and still have good enough yields to price them competitively.
So they seemed to just want to go crazy with their transistor budget, and just reach for the sky in terms of clockspeed and trickery like tracecache and Hyperthreading and all, and combine it with incredibly large caches for that time.

I suppose the recent 'hype' of multicore made Intel think about the idea of maximizing the number of cores on a CPU rather than trying to go overboard with a single core. So cores had to become small and efficient. And if we look at Itanium for a while, we see that this architecture delivers exactly that. Their cores are roughly 60% the size of a modern x86 core. I wonder where Intel is going with that now... they did say they were going to move Itanium onto the cutting edge production technology now, develop it at the same time as x86, instead of a side-project on last-generation technology.

[*]Dump their 'true-quadcore' architecture. I don't care what you read, the architecture makes impossible to financially compete against Intel. If one core goes bad, then what? You'll try to market it as a tricore? When the entire architecture was designed around having four cores? What if two cores go bad? Are we going to get a dual core? What if all but one fails? This is a horrible business model.

Not only that, but the cores are only part of the entire processor. There is also a lot of shared logic. What if the memory controller goes bad? The L3-cache? The HT-bus? etc.
You just have to keep your die-size small to keep the risk of faulty dies under control.
 
Sacli2,

Good points, but I think we can both agree that yield is the most important factor in staying profitable. It's all about fitting the most amount of chips on a wafer, and hoping most of them make it out alive when it's time to chop and package. It's far more difficult to manufacture than it is to design, and thus the formula for all successful fabs is simple: Keep you designs simple, keep them decoupled, and keep them small.
 
Good points, but I think we can both agree that yield is the most important factor in staying profitable. It's all about fitting the most amount of chips on a wafer, and hoping most of them make it out alive when it's time to chop and package. It's far more difficult to manufacture than it is to design, and thus the formula for all successful fabs is simple: Keep you designs simple, keep them decoupled, and keep them small.

Yea, I just tried to point out that the design of the chip has a lot to do with the resulting yields, and how AMD doesn't really have the CPU design they need right now.

It's really all or nothing in the chip world. If you have the fastest chip around, you can ask a premium price, and compensate for high production costs.
But if you don't have the fastest, you have to follow the pricing of your competitor, and to be profitable, you had better have lower production costs, else your profit margin will be squashed into oblivion.
I don't think any other industry is that extreme. Probably because there really is no other industry where the actual production costs are so far out of line with the retail price.
You mostly pay for the investments in R&D and new production facilities, the actual slice of etched silicon is of little value.
 
Not sure why people are disagreeing with Scali2, he's pointing out documented facts, nothing wrong in that, the people opposing him are obviously not facing facts and are being very passive when it comes down to this issue, we all know AMD and Intel need to be competitive to provide the end user the performance/price that we need.

To put it bluntly, Phenom's problem is still, a problem and until it's fixed, it'll always be known as the buggy processor.
 
To put it bluntly, Phenom's problem is still, a problem and until it's fixed, it'll always be known as the buggy processor.
Unfortunately, it's probably going to be true irrespective how prevalent the errata manifests in actual operation. Just like the automotive industry, stigmas have a tenacious hold. They are very difficult to shake off and recent rumors pointing to a possible problem with the B3 stepping (bug still there), isn't boding well for the future of this architecture. I hope history doesn't recall Barcelona as a trial run for AMD's native quadcore. I'm starting to wonder if even the 45nm process will help AMD's processor division in a substantial or significant fashion at all.
 
AMD already has a fix for the bug even for their B2 Phenom but the fix for the B2 has a performance penalty. I strongly believe that the rumours about the B3 still has the bug are totally false because the truth is AMD already has a fix for the bug, what they are trying to do with the B3 is finding a fix that doesn't give any performance penalty.

The worst that could possibly happen is that the B3 will not have the bug anymore but the performance is not up to the same level as the B2 without the fix.
 
Back
Top