will amd be making a comeback?

[Elledan]

Flame bait.

Intel was hurting badly with the Netburst/P4 debacle and only their massive financial reserves and excellent marketing department (which got them into the mess in the first place) got them out of it. AMD's CPUs right now are hardly P4-level failures and will do just fine. Heck, even Kyle liked the X3 for a gaming box, which says a lot, as he isn't the person to sugar coat anything *g*

IMO AMD has never left the stage and is just doing what it does best: supply the mainstream market with mainstream chips, while not ignoring the budget market either. Intel has got little to contribute to those markets other than some rapidly ageing chips.

 

[Zoogle]

Just out of curiosity, I saw some people post that Intel would not want AMD to shut down operations... Why not? The best situation for a manufacturer is to have a monopoly. You set the market price and people have to cough up the $ if they want to use the only product available. I can only imagine how prohibitively MOAR expensive Macs would become if Intel were the only chip supplier.

 

[brucedeluxe169]

Just out of curiosity, I saw some people post that Intel would not want AMD to shut down operations... Why not? The best situation for a manufacturer is to have a monopoly. You set the market price and people have to cough up the $ if they want to use the only product available. I can only imagine how prohibitively MOAR expensive Macs would become if Intel were the only chip supplier.

yup... for some (weird) reason, people on these boards seem to think that somehow being a monopoly is illegal.... which it isnt, by any stretch....

and intel wouldn't be a true monopoly anyway, as there would still be many other microprocessor manufacturers left if AMD actually went under....

 

[MrWizard6600]

Just out of curiosity, I saw some people post that Intel would not want AMD to shut down operations... Why not? The best situation for a manufacturer is to have a monopoly. You set the market price and people have to cough up the $ if they want to use the only product available. I can only imagine how prohibitively MOAR expensive Macs would become if Intel were the only chip supplier.

EU + Microsoft is the above's antithesis. Once you hit that monopoly scale market penetration regulations come down on you like a swift kick groin. Running a monopoly is a messy business.

 

[Snoozer]

Just out of curiosity, I saw some people post that Intel would not want AMD to shut down operations... Why not? The best situation for a manufacturer is to have a monopoly. You set the market price and people have to cough up the $ if they want to use the only product available. I can only imagine how prohibitively MOAR expensive Macs would become if Intel were the only chip supplier.

AMD originally got a license to produce Intel designs because the US government didn't want to buy computers from a sole source. Since the 486 they have been playing on their own turf.

The Socket 3's are making a real run through the enthusiast circles and that is where the hoopla gets filtered down to John Q. Public. It will always be a see saw and the more it is the better for all of us.

 

[Elledan]

EU + Microsoft is the above's antithesis. Once you hit that monopoly scale market penetration regulations come down on you like a swift kick groin. Running a monopoly is a messy business.

Yup, especially the EU isn't happy with big companies doing things like price-fixing, or abusing a monopoly. The US meanwhile is still busy after 13 years with the MSFT monopoly case

 

[Scali2]

I know the roadmap for AMD shows it being steps behind in manufacturing process, but one thing to note is the sledgehammer, clawhammer, and newcastle a64s were 130nm and keeping up just fine with intel which had already jumped to 90nm with prescott in early 04. Good luck AMD!

AMD was only able to compete because Intel's microarchitecture at the time proved not to be a very successful one.
Currently Intel has not one, but two microarchitectures (Core2 and Core i7) that are both better than anything AMD has to offer, and that's not even taking the manufacturing process into account.

AMD will need a microarchitecture that is not only better, but good enough to bridge the manufacturing process. That is unlikely to happen anytime soon, since AMD can only release a new microarchitecture every few years, and their current one is only about a year old. This means that AMD is basically 'stuck' with this microarchitecture for a while, and not likely to get into a position where it can keep up with Intel.

Ofcourse in the meantime Intel isn't just going to rest on its laurels. By the time AMD releases its next architecture, Intel will have a successor to the Core i7 microarchitecture aswell. If its successor is again the leap forward that Core2 and Core i7 were, then it's almost impossible for AMD to jump from their current position all the way over three Intel microarchitectures and take the lead.
Even if Intel doesn't develop anything better than Core i7, AMD has its work cut out to beat it with a manufacturing process disadvantage.

Conclusion: Unless Intel makes another misstep like in the Pentium 4 days, I don't think AMD will ever get back in the race.
Since Intel changed its strategy from using a microarchitecture for 5 years or more, to only 2 years (tick-tock), it's basically reduced the opportunity for AMD to take advantage of one of Intel's missteps to nearly 0. AMD would have at most 2 years to come up with a product that beats Intel's 'misstep', or else Intel has its next microarchitecture ready again, correcting its misstep, just like Core2 corrected Pentium 4. 2 years just isn't enough for AMD to come up with a new architecture.
Aside from that, the chances of a misstep from Intel's side will be smaller, as they only spend 2 years on a new microarchitecture, which will have a more evolutionary character, rather than the revolutionary step that was Pentium 4.

 

[sirmonkey1985]

yup... for some (weird) reason, people on these boards seem to think that somehow being a monopoly is illegal.... which it isnt, by any stretch....

and intel wouldn't be a true monopoly anyway, as there would still be many other microprocessor manufacturers left if AMD actually went under....

ok name a microprocessor manufacture that can compete with intel on the desktop market?

via's a joke.. IBM doesnt make hardware anymore.. so name one that could compete with intel..

sure a company looks to control the market in hope that they have no competition.. but at the same time no competition also stops the growth in the market.. if theres nothing for intel to compete with.. whats the point in coming out with faster hardware.. if AMD fell.. they could stick with the i7 forever and not care about coming out with something newer, faster, using less power..

and if AMD was to fall.. sure they could hike up prices on hardware but then at the same time the companies making the transistors and material to create the cpu's could raise the cost.. so in the end intel would probably just break even with what they make per cpu now.. with out AMD there would be no competition through out the entire market.. and would take a lot of companies down with it..

a good example would be if the 3 major automakers in the US were to completely fail.. sure theres still companies in japan and europe.. but the 3 major automakers here in the US make up a huge part of the market(steel, aluminum, glass, rubber, copper, light bulbs(material used to make them)).. theres a lot more to it then people want to think about.. they just look at AMD and Intel and dont bother to count in all the companies they use and supply them to make their microprocessors..

 

[Scali2]

AMD/ATI is making a comeback, its just your perspective is skewed on how it's doing it. Its going for competitive performance/price ratios in specific price points which have higher demand.

Very much true. AMD should aim for price/performance.
However, yesterday's enthusiast high-end products are today's low- and mid-end products.
By the end of the year, Core i5 will bring much of the Core i7's technology and performance to the mid-end market. So in a way AMD will always have to compete with Intel's technology and performance. They will just have a few extra months to respond.

 

[Scali2]

i wouldn't think it would cost them $200 per chip to make. lolol

Probably more like $40 to $50 to produce the chip (you can look up info on chip costs from companies like TSMC, Intel and AMD should be in the same ballpark).
The rest of the price is to cover for the R&D investments (developing a CPU architecture costs billions of dollars, and that also takes into account the new 45 nm production facilities).
Core i7 is technically little or no more expensive to make than Core2 Quad. Die-size isn't too far apart, and Core i7 doesn't require two dies on a single package, so it's slightly easier/cheaper to validate and package the chip.
The difference in price is mainly because Core2 has been on the market for over 2 years now, and Intel already has its return on investment. So they turned down the price margin on Core2, and moved it over to Core i7.

 

[Scali2]

Right now, AMD is doing well in the server market imo.
...
Don't get me wrong, if there was a two socket Nehalem server I'd probably choose that depending on the price premium, but no such thing exists.

You pretty much signed AMD's fate there:
AMD is doing well because Nehalem isn't available in the server market yet.
This gives AMD the advantage of their HyperTransport links, and gives AMD competitive performance against Intel's Core2-based Xeons.
Nehalem will cancel AMD's advantage, and combine it with more bandwidth, HyperThreading, and overall higher IPC.

AMD is going to lose big time in the server market once Nehalem is released. And that is probably going to hit them very hard, because the server market is very profitable (look at the prices AMD demands for their 4-8 socket Opterons).

 

[Scali2]

AMD's CPUs right now are hardly P4-level failures and will do just fine.

Nope, but the first Phenoms were. Not only did they run stupid hot, and were released way under their projected clockspeed of 3 GHz... but they had a nasty TLB bug aswell.
AMD recovered with the 45 nm shrink... But to be fair, Intel's P4 didn't entirely fail either.
Willamette wasn't great, but Northwood was fine. Prescott wasn't good, but Presler was about as good a recovery as AMD made now, with Phenom II.

IMO AMD has never left the stage and is just doing what it does best: supply the mainstream market with mainstream chips, while not ignoring the budget market either. Intel has got little to contribute to those markets other than some rapidly ageing chips.

I think that's the problem right there.
"Old school" people like myself, have always seen AMD as a budget/mainstream supplier, not a top performer (as such, I've never owned an AMD for my primary system up to the Athlon era).
Because of the temporary success of the Athlon CPUs, many people new to the PC market have come to know AMD as a company that can compete with Intel for performance, while historically it never did. As such, they have this misguided expectation that AMD will return to top performance soon, because in their limited experience, it has always been that way.
AMD is back where it always was: value.
Intel was never very interested in the value market, other than dumping their previous generation chips (or CPUs with production flaws, such as the Celerons where some or all the cache was disabled), and it seems that this will largely continue to be the case.

Mind you, I personally see it as follows:
$0-$200: low-budget
$200-$500: midrange
$500+: high-end
$1000+: enthusiast.

I always shop in this midrange. As such I have never bought an AMD CPU prior to the Athlon (why would I buy a < $200 AMD CPU which may have good price/performance if I can afford the $500 to get a completely different level of performance, which may be a year or more ahead of what AMD offers?).. And as such I don't consider AMD to be competitive in anything but the low-budget market currently, as they historically did.

 

[sirmonkey1985]

You pretty much signed AMD's fate there:
AMD is doing well because Nehalem isn't available in the server market yet.
This gives AMD the advantage of their HyperTransport links, and gives AMD competitive performance against Intel's Core2-based Xeons.
Nehalem will cancel AMD's advantage, and combine it with more bandwidth, HyperThreading, and overall higher IPC.

AMD is going to lose big time in the server market once Nehalem is released. And that is probably going to hit them very hard, because the server market is very profitable (look at the prices AMD demands for their 4-8 socket Opterons).

correct.. performance with the nehalem will murder the opteron.. but how many people are willing to pay the premium? dual socket nahalem will come out with the new 32nm cores.. but then your probably going to be looking at 300-400 dollars a processor for the lowest one.. then a dual socket 1366 board.. probably in the 600-700 dollar range minimum.. at that point your well into the 1000's before you even buy memory. though i think it will hurt AMD.. i just dont think it will happen right away.. the hit probably wont happen until late 2010 ish.. especially since the current opteron systems are much cheaper and easier to upgrade.. but we will have to see.. time can only tell..

though in my opinion i think the biggest issue nahalem is going to have is heat.. its going to cost data centers a lot more money trying to keep racks of dual and quad socket nahalem systems cool.. so that might play a big role in what some data centers choose to run..

 

[sirmonkey1985]

Nope, but the first Phenoms were. Not only did they run stupid hot, and were released way under their projected clockspeed of 3 GHz... but they had a nasty TLB bug aswell.
AMD recovered with the 45 nm shrink... But to be fair, Intel's P4 didn't entirely fail either.
Willamette wasn't great, but Northwood was fine. Prescott wasn't good, but Presler was about as good a recovery as AMD made now, with Phenom II.



I think that's the problem right there.
"Old school" people like myself, have always seen AMD as a budget/mainstream supplier, not a top performer (as such, I've never owned an AMD for my primary system up to the Athlon era).
Because of the temporary success of the Athlon CPUs, many people new to the PC market have come to know AMD as a company that can compete with Intel for performance, while historically it never did. As such, they have this misguided expectation that AMD will return to top performance soon, because in their limited experience, it has always been that way.
AMD is back where it always was: value.
Intel was never very interested in the value market, other than dumping their previous generation chips (or CPUs with production flaws, such as the Celerons where some or all the cache was disabled), and it seems that this will largely continue to be the case.

Mind you, I personally see it as follows:
$0-$200: low-budget
$200-$500: midrange
$500+: high-end
$1000+: enthusiast.

I always shop in this midrange. As such I have never bought an AMD CPU prior to the Athlon (why would I buy a < $200 AMD CPU which may have good price/performance if I can afford the $500 to get a completely different level of performance, which may be a year or more ahead of what AMD offers?).. And as such I don't consider AMD to be competitive in anything but the low-budget market currently, as they historically did.

ok rich guy.. (lol had to say it before anyone else did so that they cant say it now..)

but the i7 completely killed your pricing comparisons for CPU's.. since its considered an enthusiast cpu yet sits in your mid range pricing.. guess its time to re-think that pricing chart, isnt it..

and yes i know you'll bring up the 965.. which honestly is not an enthusiast processor.. its just for people that think epenors rule the world and has no added benefit to anything..

 

[Scali2]

if theres nothing for intel to compete with.. whats the point in coming out with faster hardware.. if AMD fell.. they could stick with the i7 forever and not care about coming out with something newer, faster, using less power..

What's there to compete with? Itself!
Why else would I ever buy a new CPU once I have a Core i7? There won't be anything faster/better available, and my CPU won't wear out.
Currently I upgrade my system every 2-3 years, because there are huge improvements CPU performance, like doubling the number of cores, much higher IPC, lower power consumption etc.
Once Intel stops offering me this, I won't buy anything new.
After a few years, everyone will have upgraded to a Core i7, and Intel sales will drop to 0.

That's just the problem... Currently Intel is used to this upgrade-cycle of CPUs (the market has pretty much been saturated for years now, at least in the Western world... Most people already own a PC, and just buy new to replace their current PCs). They will need to keep this upgrade-cycle going if they want to continue operating at their current size.
They really can't afford to ask astronomical prices for their CPUs, nor can they afford to slack too much in developing faster CPUs.
I don't think Intel will change much, if at all, if AMD were to go out of business. I think AMD currently has very little effect on Intel's products anyway, Intel seems to mostly be doing their own thing.

 

[Scali2]

ok rich guy.. (lol had to say it before anyone else did so that they cant say it now..)

but the i7 completely killed your pricing comparisons for CPU's.. since its considered an enthusiast cpu yet sits in your mid range pricing.. guess its time to re-think that pricing chart, isnt it..

and yes i know you'll bring up the 965.. which honestly is not an enthusiast processor.. its just for people that think epenors rule the world and has no added benefit to anything..

Not at all. I think the different ranges depend solely on price, not on performance. They are price ranges, in other words.
As such, I don't consider the Core i7 920 a high-end CPU because of its price. Sure, it performs close to enthusiast levels, but its price doesn't reflect it.
Just think about it, when I buy a CPU, I look at the price first, right? If I'm shopping in the $200-$500 range, and I see that the Core i7 920 fits in there, there's no reason for me NOT to buy it, just because it delivers enthusiast performance levels, right? In fact, that's only more incentive to buy it.

These are infact exactly the CPUs that I tend to buy... The ones that are very close to the top performers, but with a price in the mid-range, giving them a very good price/performance ratio.
My current CPU is the E6600, which is exactly that. I bought it shortly after introduction, with only the 6700 and 6800 above it. They were both only marginally faster, but they were much more expensive. It's much like the current I7 920/940/965 scenario.
I'd buy the 920 because it has by far the best price/performance of the three, and it performs better than anything else on the market, and offers me all the latest technology aswell. The i7 920 is just the 'sweet-spot' in today's CPU market, just like the E6600 was the sweet-spot a few years ago.

 

[Elledan]

200 bucks for a CPU? For my last workstation I picked a 4450e Athlon X2 CPU which cost me exactly 65 Euro. I use the system mainly for 3D modelling/rendering and compiling. Compared to the X2 3800+ I used previously it's already a nice boost in performance. If I had wanted to go Intel, the cheapest (lower or comparable performance than the AMD chip) C2D Intel CPU would have cost me 129 Euro. To go budget in this case would have meant opting for a Sempron or Celeron/Pentium CPU.

AMD was struggling during the K5/6 days, sure, but with the K7 and then again with the K8 they showed that they could easily compete with Intel and even IBM (POWER), both on the desktop and in the server rooms. I would not put it past them to come up with another innovative architecture. Don't forget that Intel was ready to abandon x86 and focus on IA64 (Itanium), while AMD was the one pushing x86-64. After IA64 failed badly, Intel was forced to adopt AMD's architecture changes.

Even Intel sometimes stumbles over its feet.

 

[Scali2]

200 bucks for a CPU?

Why not? If I'm going to spend a lot of money on a new system anyway, like motherboard, memory, harddisks, videocard... why would I skimp on the CPU?
Sure, if you just look at the CPU, then $100 or $200 would make a big difference.
However, put it in perspective of a complete PC costing $900 or $1000, it suddenly isn't that big a deal anymore.
I'd rather pay the extra $100 because I can afford it, and I know it will give me a considerable performance advantage.
You have to realize that I bought my E6600 2.5 years ago, and even today it is by no means a slouch. The Athlon X2 you bought is actually still SLOWER than mine. So you'd have to upgrade your entire system once more before you'd even reach the performance level of the system that I've been enjoying for 2.5 years already, and continue to enjoy.
And in those 2.5 years, I've saved a lot of time on compiling and rendering, because it's just faster. So yea, you bet that those few extra bucks for the CPU were worth it.

I would not put it past them to come up with another innovative architecture.

K7 wasn't innovative at all. It was very similar to the Pentium Pro that had been on the market for years. AMD just copied the design and improved it a bit with technology bought from Digital Alpha. Sure, it was competitive with the Pentium 3, but only because the Pentium 3 was based on an old architecture that was about to be replaced anyway, by the Pentium 4.
It's just that Pentium 4 got off to a bad start, and it took until the Northwood revamp of the core to finally get it to perform adequately.

The K8 was the only time that AMD actually had a decent architecture with some smart advantages over Intel. The integrated controller and HyperLinks gave them a performance advantage in the server market.
Intel just let them have that advantage for a long time, but now AMD has lost this advantage. As Intel already said: You can only integrate the memory controller once. What is AMD going to do now?

Don't forget that Intel was ready to abandon x86 and focus on IA64 (Itanium), while AMD was the one pushing x86-64. After IA64 failed badly, Intel was forced to adopt AMD's architecture changes.

Yes, I still hate AMD for this. x86 should have died. IA64 may not be perfect, but it would have been a fresh start, and better suited to multicore and power-efficient CPUs, as the core logic for IA64 is very small and elegant.
I suppose Intel was a bit too arrogant there, thinking that they could force IA64 despite AMD's attempt at bringing 64-bit to the x86.
From a technological point of view, IA64 should have won.
But from a psychological point of view, Intel shouldn't have attempted this. Especially Intel should have known that people always choose the short-term benefit of compatibility and performance with legacy applications over superior technology. It's what x86 has always been about.

 

[Elledan]

Yes, I still hate AMD for this. x86 should have died.

Can't disagree too much with this.

IA64 may not be perfect, but it would have been a fresh start, and better suited to multicore and power-efficient CPUs, as the core logic for IA64 is very small and elegant.
I suppose Intel was a bit too arrogant there, thinking that they could force IA64 despite AMD's attempt at bringing 64-bit to the x86.
From a technological point of view, IA64 should have won.

IA64 was and still is a bloody joke. It hardly was the first VLIW design (Transmeta is an example of how to do it right) and by making it in-order and leaving any and all optimization steps to the compiler it did a very poor job of preparing for real-life tasks, as code stream optimizations in hardware are well understood while compilers do a very poor job of it in comparison. Heck, there's a reason why JIT VMs like the Java VM can run circles around some kinds of compiled code and it isn't due to the language used.

At this point IA64 is being slaughtered in the server and supercomputing markets by Opteron, POWER, Cell and GPU-based systems. There's the odd Intel system in there as well. IA64 has always been a super-niche idea simply by being such a badly designed and esoteric architecture.

But from a psychological point of view, Intel shouldn't have attempted this. Especially Intel should have known that people always choose the short-term benefit of compatibility and performance with legacy applications over superior technology. It's what x86 has always been about.

Oh, I don't know. After seeing entire markets fall over themselves in order to release new ARM-based devices, including netbooks and UMPCs, I'm not so sure that outside of the desktop market x86 has such a stranglehold on the consumer. Don't forget that ARM chips outsell x86 chips by at least 9:1. There's an ARM chip in nearly everything embedded you use, and even a number of very popular handheld consoles (GBA, DS(i), the PSP is the odd one out with MIPS).

 

[Scali2]

IA64 was and still is a bloody joke. It hardly was the first VLIW design (Transmeta is an example of how to do it right) and by making it in-order and leaving any and all optimization steps to the compiler it did a very poor job of preparing for real-life tasks, as code stream optimizations in hardware are well understood while compilers do a very poor job of it in comparison.

Correct me if I'm wrong, but I thought the Transmeta was in-order aswell.

Heck, there's a reason why JIT VMs like the Java VM can run circles around some kinds of compiled code and it isn't due to the language used.

That has little to do with the underlying architecture though.
The Java VM can do certain things like loop unrolling or function inlining with runtime information. I see no reason why this wouldn't work on IA64. In fact, it's probably easier to write a JIT compiler for IA64 than it is for x86, with all its archaic limitations and weird performance pitfalls.
Windows XP SP2 for IA64 actually came with a JIT compiler for x86 code, which gave very acceptable performance at the time.

At this point IA64 is being slaughtered in the server and supercomputing markets by Opteron, POWER, Cell and GPU-based systems. There's the odd Intel system in there as well. IA64 has always been a super-niche idea simply by being such a badly designed and esoteric architecture.

Not correct at all. Itanium has always had very strong floating point performance, and Itanium 2 was actually trading blows with POWER directly on various applications.
You should not forget that the Itanium hasn't seen any major updates in years, and the CPUs are in fact still produced on 90 nm technology(!).
No wonder that they aren't more competitive with CPUs that are more up-to-date (try comparing 90 nm x86 CPUs to the current crop, they will pale in comparison). That isn't a flaw of the architecture itself, but is just a result of Intel focusing on Conroe and Nehalem in the past few years.
However, Intel has announced that they will now bring Itanium back on par with x86, making it use the latest manufacturing process again, and in fact, they want to share the socket for server systems between Itanium and x86.
Itanium may well be making a comeback in a major way. I wouldn't write it off just yet.

Oh, I don't know. After seeing entire markets fall over themselves in order to release new ARM-based devices, including netbooks and UMPCs, I'm not so sure that outside of the desktop market x86 has such a stranglehold on the consumer.

But we weren't talking about those markets, because IA64 was supposed to supercede x86 on the desktop. It actually did quite well in the server market for a while, and obviously it was never aimed at netbooks or anything embedded.
Intel isn't planning to attack ARM with IA64 either, they're going after them with Atom. Funny enough Intel was one of the biggest suppliers of ARM chips only a few years ago.

 

[Kuja]

The thing with Intel and AMD is that Intel doesn't play fair. They never have. They give you special pricing so long as you don't offer their competitor products. Do you remember when HP/Compaq, Dell, Lenovo, etc etc didn't sell any machines with AMD bits? It wasn't very long ago when AMD had a stranglehold on no market.

For the period where K8 was beating the pants off of the P4, those OEMs decided to bite the bullet and offer AMD anyway, because the price/performance was way better.

Since the core architecture was introduced, Intel processors run cooler, faster, and on far less power. AMD has no advantages left, not even price/performance. The only difference between now and before K7 is that now they've captured the budget portion of the OE market. They're not going anywhere.

 

[crackbone]

Not at all. I think the different ranges depend solely on price, not on performance. They are price ranges, in other words.
As such, I don't consider the Core i7 920 a high-end CPU because of its price. Sure, it performs close to enthusiast levels, but its price doesn't reflect it.
Just think about it, when I buy a CPU, I look at the price first, right? If I'm shopping in the $200-$500 range, and I see that the Core i7 920 fits in there, there's no reason for me NOT to buy it, just because it delivers enthusiast performance levels, right? In fact, that's only more incentive to buy it.

These are infact exactly the CPUs that I tend to buy... The ones that are very close to the top performers, but with a price in the mid-range, giving them a very good price/performance ratio.
My current CPU is the E6600, which is exactly that. I bought it shortly after introduction, with only the 6700 and 6800 above it. They were both only marginally faster, but they were much more expensive. It's much like the current I7 920/940/965 scenario.
I'd buy the 920 because it has by far the best price/performance of the three, and it performs better than anything else on the market, and offers me all the latest technology aswell. The i7 920 is just the 'sweet-spot' in today's CPU market, just like the E6600 was the sweet-spot a few years ago.

I'm going to agree with about everything that Scali is saying here.

I'm 32 years old now (ugh) and I've been building PCs since I was 8 years old. My father and I started with a 286, and I've been an AMD guy for a long time. Honestly, I wouldn't recommend an AMD system to anyone today. It's not because they aren't solid PCs, its because their performance versus their price doesn't make them a worthwhile buy.

The problem for AMD is , that the i7 920 is the sweet spot. It's inexpensive based on the performance. Granted, its base price is HIGHER, but for the performance gains, its no contest. With Nehalem hitting the server market, I fear AMD is going to start losing major share on the enterprise level as well.

Intel's current tick tock development cycle is the killer for AMD. They hit paydirt with the Athlon64 design, and we are all better off for it, but I think it's the blind squirrel finding the the nut.

Anything can happen once, but I don't think AMD is going to strike lightning twice. I think they'll be forced into the budget segment,. which I believe is going to wind up being the end of them. Intel has all of the margin...

 

[MrMike]

You pretty much signed AMD's fate there:
AMD is doing well because Nehalem isn't available in the server market yet.
This gives AMD the advantage of their HyperTransport links, and gives AMD competitive performance against Intel's Core2-based Xeons.
Nehalem will cancel AMD's advantage, and combine it with more bandwidth, HyperThreading, and overall higher IPC.

AMD is going to lose big time in the server market once Nehalem is released. And that is probably going to hit them very hard, because the server market is very profitable (look at the prices AMD demands for their 4-8 socket Opterons).

Yep, the 4 and 8 socket systems are pretty pricey. If they dropped those to the same price as the 2 socket on Nehalem's release, then I think there would be a possibility of them remaining competitive based on price.

correct.. performance with the nehalem will murder the opteron.. but how many people are willing to pay the premium? dual socket nahalem will come out with the new 32nm cores.. but then your probably going to be looking at 300-400 dollars a processor for the lowest one.. then a dual socket 1366 board.. probably in the 600-700 dollar range minimum.. at that point your well into the 1000's before you even buy memory. though i think it will hurt AMD.. i just dont think it will happen right away.. the hit probably wont happen until late 2010 ish.. especially since the current opteron systems are much cheaper and easier to upgrade.. but we will have to see.. time can only tell..

though in my opinion i think the biggest issue nahalem is going to have is heat.. its going to cost data centers a lot more money trying to keep racks of dual and quad socket nahalem systems cool.. so that might play a big role in what some data centers choose to run..

Sirmonkey's concerns here echo my own. Although my preference would be twin Nehalem for a platform, the question is 'How much will it cost to get the same features?' Right now, a comparable 2-socket Intel virtualization platform is more expensive than an AMD one and lacks features that the AMD platform has.

To be honest, I expect the new Nehalem based servers to simply match AMD on features (128GB RAM, 4x NICs, etc) and just have superior IPC. Then it's a question of another thing sirmonkey brought up, the heat and power usage.

The thing that makes Nehalem not such a big deal is that it most likely will not raise consolidation ratio until Intel gets 8-core out the door, which I imagine will require a die-shrink. IMO that's what it will take for Intel to take market share back. That's why Dunnington looks so good regardless of the cost. 4U of rack space with 4x6=24 and results in a higher consolidation ratio than 4x4=16.

So yes, overall I do agree. When 8-core Nehalem hits the server market, AMD is in for some rough times unless they can match Intel's consolidation ratio.

 

[Scali2]

If you followed the blog on Anandtech... They are hinting at the fact that the dual socket Nehalem systems will perform at the same level as quad socket Opteron systems, thanks to Nehalem's HyperThreading, better IPC and triple channel memory. They actually demonstrated this with a few pre-release benchmarks from OEMs.
This makes the future for AMD look really grim in the server market. There's no way you can compete if you need 4 CPUs to get the performance of 2 of your competitor's CPUs. Quad socket systems are just much larger, use more power, and are overall far more expensive.

 

[facebook]

If you followed the blog on Anandtech... They are hinting at the fact that the dual socket Nehalem systems will perform at the same level as quad socket Opteron systems, thanks to Nehalem's HyperThreading, better IPC and triple channel memory.

You should try vmware with Phenom II compared to i7 I don't think you would like the result

 

[Scali2]

You should try vmware with Phenom II compared to i7 I don't think you would like the result

You mean like this?
http://www.zdnet.com.au/reviews/har...re-i7-Nehalem-/0,139023402,339293122-3,00.htm

 

[MrMike]

If you followed the blog on Anandtech... They are hinting at the fact that the dual socket Nehalem systems will perform at the same level as quad socket Opteron systems, thanks to Nehalem's HyperThreading, better IPC and triple channel memory. They actually demonstrated this with a few pre-release benchmarks from OEMs.
This makes the future for AMD look really grim in the server market. There's no way you can compete if you need 4 CPUs to get the performance of 2 of your competitor's CPUs. Quad socket systems are just much larger, use more power, and are overall far more expensive.

I don't really follow Anandtech's blog. Did Intel change their stance on HT and server loads? I thought the recommended configuration was always to disable it. That would be pretty impressive if a Nehalem core is literally worth two Shanghai cores for virtualization loads.

If that's true, then you can fit the equivalent of a 4U 4-socket system into a 1U 2-socket system, which is ridiculous from a density perspective. Heat and power mean nothing when you only need 1/4th the rack space.

 

[facebook]

You mean like this?
http://www.zdnet.com.au/reviews/har...re-i7-Nehalem-/0,139023402,339293122-3,00.htm

try it, you just seems to read about hardware?

I have tried both, the difference rather large
I don't think vmware users runs cinebench...

 

[rflcptr]

try it, you just seems to read about hardware?

I have tried both, the difference rather large
I don't think vmware users runs cinebench...

I'm sorry. Your incoherent rambling is ruining what is otherwise a nice discussion. Care to back any of this up? Or contribute in any meaningful way?

 

[Acer_Sheep]

Nope, but the first Phenoms were. Not only did they run stupid hot, and were released way under their projected clockspeed of 3 GHz... but they had a nasty TLB bug aswell.
AMD recovered with the 45 nm shrink... But to be fair, Intel's P4 didn't entirely fail either.
Willamette wasn't great, but Northwood was fine. Prescott wasn't good, but Presler was about as good a recovery as AMD made now, with Phenom II.



I think that's the problem right there.
"Old school" people like myself, have always seen AMD as a budget/mainstream supplier, not a top performer (as such, I've never owned an AMD for my primary system up to the Athlon era).
Because of the temporary success of the Athlon CPUs, many people new to the PC market have come to know AMD as a company that can compete with Intel for performance, while historically it never did. As such, they have this misguided expectation that AMD will return to top performance soon, because in their limited experience, it has always been that way.
AMD is back where it always was: value.
Intel was never very interested in the value market, other than dumping their previous generation chips (or CPUs with production flaws, such as the Celerons where some or all the cache was disabled), and it seems that this will largely continue to be the case.

Mind you, I personally see it as follows:
$0-$200: low-budget
$200-$500: midrange
$500+: high-end
$1000+: enthusiast.

I always shop in this midrange. As such I have never bought an AMD CPU prior to the Athlon (why would I buy a < $200 AMD CPU which may have good price/performance if I can afford the $500 to get a completely different level of performance, which may be a year or more ahead of what AMD offers?).. And as such I don't consider AMD to be competitive in anything but the low-budget market currently, as they historically did.

This means AMD sell more processors for lower price(the most personal computers on the world are in mainstream) and Intel less processors for higher price(not every computer is high-end).

 

[Flyboat]

I'm going to agree with about everything that Scali is saying here.

I'm 32 years old now (ugh) and I've been building PCs since I was 8 years old. My father and I started with a 286, and I've been an AMD guy for a long time. Honestly, I wouldn't recommend an AMD system to anyone today. It's not because they aren't solid PCs, its because their performance versus their price doesn't make them a worthwhile buy.

The problem for AMD is , that the i7 920 is the sweet spot. It's inexpensive based on the performance. Granted, its base price is HIGHER, but for the performance gains, its no contest. With Nehalem hitting the server market, I fear AMD is going to start losing major share on the enterprise level as well.

Intel's current tick tock development cycle is the killer for AMD. They hit paydirt with the Athlon64 design, and we are all better off for it, but I think it's the blind squirrel finding the the nut.

Anything can happen once, but I don't think AMD is going to strike lightning twice. I think they'll be forced into the budget segment,. which I believe is going to wind up being the end of them. Intel has all of the margin...

If that is true, I don't see any innovation for next 5 years. It will be back to days with Intel 133 mhz => 166 mhz => 200 mhz => 233 mhz.

 

[facebook]

I'm sorry. Your incoherent rambling is ruining what is otherwise a nice discussion. Care to back any of this up? Or contribute in any meaningful way?

http://www.vmware.com/pdf/RVI_performance.pdf

 

[crackbone]

If that is true, I don't see any innovation for next 5 years. It will be back to days with Intel 133 mhz => 166 mhz => 200 mhz => 233 mhz.

Potentially.

However, I see an opportunity for AMD to make gains in the GFX market. That might be the only thing that gives them an opportunity to hold on.

IMO, Nvidia is starting to look a helluva lot like Intel did during Netburst.

However, I don't see an Athlon 64 type opportunity in the near term. Intel is ahead in performance, they are ahead in fabrication, and they are ahead in concept (gfx on die, etc).

Look, @ heart, I love AMD. But you have to understand that the market reality is, they are making inferior products(CPU and chipset). As much as I like them, after the introduction of the Core2 architecture, there's been no reason to buy from AMD. Sure, it's been cheaper, and they've been pretty good about holding onto sockets, but other than that, price vs performance has been squarely with Intel.

 

[Scali2]

This means AMD sell more processors for lower price(the most personal computers on the world are in mainstream) and Intel less processors for higher price(not every computer is high-end).

Actually, it doesn't. Just because AMD isn't active in all segments doesn't mean that Intel isn't.
Intel outsells AMD in all segments, and Intel also has higher profit margins than AMD on every product.
If what you said were true, AMD would have the biggest marketshare.

 

[rflcptr]

http://www.vmware.com/pdf/RVI_performance.pdf

This is nice and all, but it doesn't look like it's comparing an AMD system to an Intel system, which is what you did.

 

[Scali2]

http://www.vmware.com/pdf/RVI_performance.pdf

It doesn't contain any data on Core i7, or any Intel CPU at all.
How do we compare?

 

[Silent.Sin]

You mean like this?
http://www.zdnet.com.au/reviews/har...re-i7-Nehalem-/0,139023402,339293122-3,00.htm

I would like to see some Nahelem VM benches with >4 VMs running at full capacity. That bench just showed 2 which is almost the exact same thing as running 2 instances of Cinebench on the host OS desktop -which we already know gets eaten for breakfast by the i7- and doesn't show the advantage or typical usage scenario of running multiple VMs on a server.

Some personal criticisms of that article: The system setup page for those benches leaves a lot to be desired as it doesn't even properly denote the total RAM for the i7 or AMD systems, just the number of DIMM slots populated. I'm assuming it's 3GB for i7 and 4GB for AMD, but there are actually more 3x2GB triple channel kits out there than 3x1 so its hard to know for sure. Not to mention they inexplicably use a different set of DDR3 modules for the i7 and the Core2 systems even though they are at the same speed. I'm just being a stickler for those specs because VMs are much more sensitive to memory capacity than a single OS system would be. On the same note, they also didn't list specifics for the VM setup concerning any affinity settings, RAM allocation, etc that are typically tweaked for systems running multiple VMs.

I'm not even entirely certain how VMware deals with the addition of HT. Is the host OS solely responsible for scheduling threads as it deems necessary in that situation? I know in the past with the P4 it was suggested that HT be turned off completely with VMware, but that could have changed since then. Anyway it would be nice to see the impact of dipping into those logical cores vs. physical with >4 VMs, can't seem to find any benches for such a setup. Got some links to that Scali?

The only reason I'm curious is the inevitable comparison between Nahalem EP and Istanbul which the Anandtech blogs went into. If the Nahalem EP 4-core systems can only run 4 VMs efficiently using the 4 physical cores (meaning that the addition of HT is moot for the purposes of server-level VM infrastructure) than the Istanbul (and Dunnington I guess) will still have the density advantage even if they don't have the per capita performance advantage...for a little while anyway.

Having some sort of competitive offering in the server realm is key for AMD's survival and I think they are pushing hard to get the Istanbul and Magny-Cours platforms out there ASAP. This isn't necessarily making a "comeback" as the poll asks about since AMD hasn't played second fiddle in servers for awhile, but it's certainly an area where they will have to be prepared for a fierce battle.

 

[Conker]

I think amd will surprise everyone. Amds current offerings aren't even that bad. As long as they keep their prices low then they should last along time. The doom and gloom theories always happen every generation lol.

 

[crackbone]

I think amd will surprise everyone. Amds current offerings aren't even that bad. As long as they keep their prices low then they should last along time. The doom and gloom theories always happen every generation lol.

True, but times are a bit different this go around.

The economy actually favors Intel. They have the dollars (7 billion to be exact) to push into additional fabs and R&D. They are considerably less leveraged than AMD, and have a new product line that directly threatens AMDs largest profit center(enterprise market). Not to mention, Intel doesn't have an ATI type of burden either.

Now, I understand that some of these could easily be looked @ by an optimist as AMD advantages, and granted, they might turn out to be. However, I'm going lean on recent history, and say that AMD is going to have a really tough time getting out of this hole.

 

[Scali2]

Got some links to that Scali?

That article was just the first link that came up when I googled. I wanted to know what facebook is on about.
I have 0 interest in servers and 0 interest in VMs to be perfectly honest with you.