Oh, I agree that it'd be nice for AMD to do it. They're just making so few high-end sales that I don't think it'd hurt them financially if Athlon64 got delayed even more.I was referring to the high-end though. It's obvious that the Athlon XP has reached it's limit with the current .13 micron tech. So to compete with Intels high-end, AMD better get the release of A64 right now in September.
Well, 'planning' and realizing are two different things. Hopefully in the end that stupid rating concept will get flushed down the toilet. There is still time. (Of course it wouldn't surprise me if AMD doesn't change a thing.)First off, the good but *extremely* AMD-biased site AMDzone.com has as it's first news that HP is planning a "3100+" new system, with Athlon 64 CPU. Guess AMD won't give up their rating system, (which I didn't think they would anyways). Bad news anyways I think.
No arguments here. I personally am not even convinced that the Athlon64 is good enough.I guess you're right. But it should be in any serious company's interest to stay competitive with the rest.
And that sure needs some work before AMD can do that.
**ROFL** AMD couldn't even support that claim when they released the AXP3200+. I doubt that anything will change. What little marketing AMD ever does is often highly suspect. Like those AMD ME ads. To paraphrase: "It does more work per clock, so it's a better CPU." Of course they never mention that it's clock is a lot lower. **ROFL** Yeah, real honest marketing from AMD.AMD's focus sure seems to have been pointed at making the best sales at the low-end to mid-end sectors for sure, so you're right about that failing to release A64 this year won't do AMD any good but probably wouldn't do much bad either. But for people to relate to all those "world's fastest processor" press releases lately, they'd better have products that can back up their claims
I'm fairly certain that they could do .13 micron SOI AXPs with little effort. If they're smart, they've already done most (if not all) of the R&D anyway and would just have to set up their FAB to manufacture. After all, it's quite possibly the easiest backup plan in the world to avoid losing face should Athlon64 come up with any quirks.I guess AMD could just make SOI on their .13 micron tech or even better, make a .09 tech, but wouldn't that take a few months of testing to make sure the product works good at it's new die size? Of course, they could've done this testing the last 4 months or so without anyone knowing I guess
Yeah. Intel has been doing a lot to raise the IPC. They tweak the number of cycles an operation takes here and there in core revisions. They tweak their prediction algorithms. They up the FSB and actually utilize that bandwidth. They implemented HyperThreading into the Pentium4. And with Prescott they'll also be raising the cache yet again. (Which will help HT tremendously.) Intel has been raising the IPC slowly but surely with each new Pentium4 core revision. It's <i>almost</i> to the point where a new P4 has similar IPC to an old P3. A few years ago people would have laughed at a statement like that, and I'd bet that AMD would have been right up there laughing too. I really don't think AMD was ever prepared for that eventuality. The AXP just doens't show the same effort at raising the IPC. So yeah, the AXP is looking pretty sad lately compared to the latest P4s.But one thing have puzzled me lately and that is how the latest P4's seem to get more and more performance in relation to how much the Mhz that was actually raised. Guess the 800 FSB helps to keeps the pipelines better filled, or maybe even raising the IPC, something that would make the best AXP's look worse, and quite dated. That's why I think the AXP is starting to look faded.
Yeah. I'm hoping it won't be delayed any further. Actually, I hope that AMD will surprise people by releasing it early, like early July. That'd be a great move for them, even if a 64-bit Windows isn't ready. The hype alone would be astounding.But since there hasn't been any mention of further delays I suppose the A64 will indeed be released in September. Guess we have to wait and see.
Heh, I'd be surprised if AMD DID do something about it.Well, 'planning' and realizing are two different things. Hopefully in the end that stupid rating concept will get flushed down the toilet. There is still time. (Of course it wouldn't surprise me if AMD doesn't change a thing.)
Athlon 64 by itself will not cut it for AMD unless they start doing something special about it. I've already seen a premature CPU being beat by the AXP in some Preview awhile ago. While that test was by no means a final indication, it still painted a bad picture. AMD really needs to do some things to improve A64, first off, 1.8 GHz doesn't cut it in my opinion. They must get a better SOI than what they've been able to get so far. Also, they must do something else to improve IPC, your examples are good ones. I actually heard somewhere for quite a while ago that IPC was gonna be slightly inferior to the AXP, that might be wrong though, and my memory might be wrong as well.No arguments here. I personally am not even convinced that the Athlon64 is good enough.
If AMD can quad-pump (or better yet, octa-pump) the FSB and actually start utilizing the full bandwidth of a dual-channel PC3200 memory system (or perhaps something even better...) then I might start to be impressed.
If AMD could even guarantee an increase in IOPS and FLOPS in 32-bit mode without utilizing the additional registers I'd be impressed.
Hell, if AMD could even just tweak the speed that some of the operations are performed at to take less cycles and could buff up their prediction algorithms significantly, I'd be impressed.
But as it is the A64 looks like it'll hardly have any more IPC than the AXP when running un-optimized 32-bit code. That'll certainly never do. Intel is definately gaining IPC and if AMD doens't then AMD will either have to find a way to or start to ramp their overall frequencies higher.
I just don't see the Athlon64 as a very competetive product for most home users. It might make a nice low-end server, but it's not going to do much else. Right now as far as I can see it's only real selling point to most home users is just hype. I hope that I'm wrong or that it is more scalable than the AXP.
Well Intel has always been a top-selling chipmaker and AMD has not been doing equally well, thus they get desperate and shell out all those marketing faults, like claiming they got the fastest CPU when they actually don't, or when they say IPC matters more than MHz. It certainly is not acceptable but it is understandable. Somewhat.**ROFL** AMD couldn't even support that claim when they released the AXP3200+. I doubt that anything will change. What little marketing AMD ever does is often highly suspect. Like those AMD ME ads. To paraphrase: "It does more work per clock, so it's a better CPU." Of course they never mention that it's clock is a lot lower. **ROFL** Yeah, real honest marketing from AMD.
So no, I don't think AMD is in any way concerned about their "world's fastest processor" claims.
Really it would hurt AMD in the long run (like over the next two years) to delay the Athlon64 longer, but I don't see it doing any real short-term damage other than yet more bad PR. (Which they already have so much of lately that more hardly matters.)
Heaven forbid quirks of the A64 That certainly would NOT be good for AMD.I'm fairly certain that they could do .13 micron SOI AXPs with little effort. If they're smart, they've already done most (if not all) of the R&D anyway and would just have to set up their FAB to manufacture. After all, it's quite possibly the easiest backup plan in the world to avoid losing face should Athlon64 come up with any quirks.
The only real down side to that plan would be the cost to produce the CPUs and the reduced production of non SOI AXPs and/or Athlon64s and/or Opterons. But then considering the price point that they'd put an AXP 3400+ at anyway, the cost to produce hardly seems likely to be a major problem. They could even just paper-launch it and provide a rare few to some OEMs just to claim to be the king of performance once more. It wouldn't matter if we couldn't actually buy one for another quarter.
As for the .09 micron process, I'm really not sure how AMD is doing there. Last I heard the lithography tools just weren't available to either Intel or AMD yet, so as long as AMD keeps on the ball as well as Intel does for that process switch, AMD has nothing to worry about. The P4 is about at the end of it's line for .13 micron as well, so as long as they're both stuck at .13 micron neither the P4 nor the AXP is going much further.
So as long as AMD has a backup plan then their only real worry in staying afloat is the switch to the .09 process. So long as they can jump that hurdle, they can put off Athlon64 for even another year without a problem.
But again, hopefully the Athlon64 won't get delayed any more. (And hopefully it will be more capable of handling the IPC jumps that Intel has been making lately than I'm expecting it to.)
Right on it. This might become a real problem. The Athlon 64 and Athlon XP CPU's are really similar. Thus it's not safe at all to bet that the A64 will vastly improve performance in 32-bit apps over the AXP.Worse though is how much the Athlon64 itself is based on the AXP in design. I'm seriously hoping that AMD has put more work into the A64 than first impressions give, or else the IPC of the A64 isn't going to be much better than the AXP itself, and that surely won't compete well unless AMD can really ramp the clockspeed.
I agree.Yeah. I'm hoping it won't be delayed any further. Actually, I hope that AMD will surprise people by releasing it early, like early July. That'd be a great move for them, even if a 64-bit Windows isn't ready. The hype alone would be astounding.
Sadly, you make a good point. I think I would be too.Heh, I'd be surprised if AMD DID do something about it.
I'm not surprised. When I think about it, I think I've been subconsciously avoiding looking at Athlon64 benchmarks and comparisons just so that I don't get depressed. Heh heh.I've already seen a premature CPU being beat by the AXP in some Preview awhile ago. While that test was by no means a final indication, it still painted a bad picture.
Well, theoretically the IPC will get slightly improved by the on-die memory controller. (For whatever that's worth.) And of course when running optimized software it will have all of those nice extra registers to play with. I'm not convinced that the SSE2 support will actually help the IPC any, but at least it shouldn't hurt it. Heh heh. So in theory the A64 will be better ... when running optimized software at the very least. I fear though that when not running optimized software (AKA the plethora of 32-bit software out now) it really will just be just a glorified overpriced AXP.AMD really needs to do some things to improve A64, first off, 1.8 GHz doesn't cut it in my opinion. They must get a better SOI than what they've been able to get so far. Also, they must do something else to improve IPC
It's debatable and really depends a lot on what final hardware AMD goes with too. I mean who would want an ondie single-channel DDR controller when they could have an offdie dual-channel DDR controller? And then there's the slightly longer pipeline. It shouldn't cause a problem if AMD can improve their prefetching and utilization of the memory bandwidth, but will they? So many questions, so few answers. We really won't know if it'll be better, worse, or exactly the same as an AXP until we have final silicon.I actually heard somewhere for quite a while ago that IPC was gonna be slightly inferior to the AXP, that might be wrong though, and my memory might be wrong as well.
The funny thing is that AMD doesn't even have to use such marketing tricks. Just doing advertisements on TV on how much less an AMD budget system costs to an equally-performing Intel budget system would do wonders. Heck, if AMD just advertised on TV so that people even recognize the AMD brand name would be a very welcome change. One of AMD's biggest problems is that they have the engineering skill to compete, but they also have really bad management. They've missed so many opportunities and done so many stupid things that I'm amazed they're even holding on as well as they are now.Well Intel has always been a top-selling chipmaker and AMD has not been doing equally well, thus they get desperate and shell out all those marketing faults, like claiming they got the fastest CPU when they actually don't, or when they say IPC matters more than MHz. It certainly is not acceptable but it is understandable. Somewhat.
But I'm definitely *not* approving such deceitful manners. I'm just saying I understand why it happens.
It sure wouldn't. I think if they're not run by complete idiots though, they'd survive long enough to fix those quirks.Heaven forbid quirks of the A64 That certainly would NOT be good for AMD.
The plans were scrapped, but I bet the R&D data is still there and ready to be used. It didn't make any economical sense at the time to have it cost that much more to produce the Barton CPUs, but should the A64 be delayed much more, it might start becoming feasable to use SOI so that AMD can make some even faster Bartons with good yields.There hasn't been any mention at all from what I've heard that indicated a SOI Barton. They talked about it a while ago, but then they said the plans were scrapped.
Exactly. Lately AMD has been selling their top-end CPUs for a lot more than they used to. (Which really was the first smart move by AMD in a while. Intel would have just lowered their prices to match AMD's top-end anyway, so AMD might as well make as much money as they can out of it instead of try to stupidly compete with the giant.) So the cost to produce SOI Bartons is probably finally economically feasable for AMD. I doubt that AMD will bother doing so unless they absolutely have to though.About price, it's like you say, since the current topdog AMD CPU's are expensive anyways, SOI wouldn't hurt the price too much, me thinks. Maybe the price of SOI wouldn't raise price much anyways. The Opteron 240 is cheap. Just a thought...
Supposedly.I don't know about .09 micron either. At least not from AMD. Heh, stupid question coming up: The first Prescott will come at .09 micron, right?
Yep. Intel just plain designed the P4 to scale to really high clock speeds. AMD designed the Athlon to compete with the Pentium 3. They haven't really tried to design a CPU from scratch to compete with the Pentium 4 yet. So really, it's impressive that AMD has held on for this long. Though I still have the strong impression that this is more by Intel's choice than by AMD's skill.Yeah, both Intel and AMD are reaching the limits of the current CPU's. Though Intel still seem to have a little more headroom, bigger pipelines to fill, more IPC to gain, I guess...
Yeah. That's pretty much how I see it. I think that AMD will just be relying on optimized software to improve the A64's performance though, at least for this version of the core.If AMD can find ways of improving the Athlon XP's IPC or so, then they might be able to keep up with Intel for a few more months, but if not, then A64 better come quick. But, as we talked about before, A64 alone as it is might not cut it either, AMD needs to find ways of improving the clock-for-clock performance.
Ugh...I hated to see those words, but I agree. I think that Intel basically has AMD "on the ropes" and has AMD right where Intel wants them. Intel can't put AMD out of business, for fear of monopolization, the resulting lawsuits and possible break-up of Intel.Though I still have the strong impression that this is more by Intel's choice than by AMD's skill.
Exactly. I get that impression too.Though I still have the strong impression that this is more by Intel's choice than by AMD's skill.
Yeah well, I _hope_ AMD does something about it though.Sadly, you make a good point. I think I would be too.
I'm not surprised. When I think about it, I think I've been subconsciously avoiding looking at Athlon64 benchmarks and comparisons just so that I don't get depressed. Heh heh.
Well, theoretically the IPC will get slightly improved by the on-die memory controller. (For whatever that's worth.) And of course when running optimized software it will have all of those nice extra registers to play with. I'm not convinced that the SSE2 support will actually help the IPC any, but at least it shouldn't hurt it. Heh heh. So in theory the A64 will be better ... when running optimized software at the very least. I fear though that when not running optimized software (AKA the plethora of 32-bit software out now) it really will just be just a glorified overpriced AXP.
It's debatable and really depends a lot on what final hardware AMD goes with too. I mean who would want an ondie single-channel DDR controller when they could have an offdie dual-channel DDR controller? And then there's the slightly longer pipeline. It shouldn't cause a problem if AMD can improve their prefetching and utilization of the memory bandwidth, but will they? So many questions, so few answers. We really won't know if it'll be better, worse, or exactly the same as an AXP until we have final silicon.
The funny thing is that AMD doesn't even have to use such marketing tricks. Just doing advertisements on TV on how much less an AMD budget system costs to an equally-performing Intel budget system would do wonders. Heck, if AMD just advertised on TV so that people even recognize the AMD brand name would be a very welcome change. One of AMD's biggest problems is that they have the engineering skill to compete, but they also have really bad management. They've missed so many opportunities and done so many stupid things that I'm amazed they're even holding on as well as they are now.
You could be right. There's several possible causes to what was the reason for the delay of the Athlon 64. Manufacturing problems, performance problems, or the more unlikely one of them all, that Microsoft hasn't yet released their x86-64 OS, which after all, when I think about it, might be a valid claim, because that OS would, when it comes out, bring better 64-bit capability with it, thus increasing the power and value for the Athlon 64. So in a way, AMD has a right to claim that. Of course, they can't claim it in a way to accuse Microsoft of being the cause of the delay. They could say they're delaying it because their CPU's lack performance without a 64-bit capable OS. I hope I'm making sense here for both myself and everyone else who might be reading thisIt sure wouldn't. I think if they're not run by complete idiots though, they'd survive long enough to fix those quirks.
At first I honestly though that AMD was having problems with manufacturing the A64. It had the same feeling as when they released the Thoroughbred As. Now though I'm thinking that the silicon itself is probably just fine and the manufacturing process is probably getting pretty good yields. I think this time the problem is that the performance of the A64 when running all of our favorite 32-bit software is simply exactly the same (or even slightly worse) than AXP systems. If AMD released their 'new' product line performing noticably worse than their current product line, people would mock.
So to ensure that this doesn't happen, I think AMD is just simply waiting for M$ to release an OS that takes full advantage of the A64. That way people won't be comparing apples to apples anymore, and AMD figures that most people will just never notice the pure 32-bit performance of the A64. The optimized software will run better and AMD can then easily claim that the A64 is a success. Without that optimized software though...
That sounds good, both as a thought and as a marketing strategy for AMDThe plans were scrapped, but I bet the R&D data is still there and ready to be used. It didn't make any economical sense at the time to have it cost that much more to produce the Barton CPUs, but should the A64 be delayed much more, it might start becoming feasable to use SOI so that AMD can make some even faster Bartons with good yields.
Couldn't resist quoting thatSupposedly.
Yeah. Since Intel is the market leader, AMD has until now mostly been quite forced to follow in Intels footsteps instead of inventing entirely new creations of their own. This might change now, hopefully so. Competition is always good.Yep. Intel just plain designed the P4 to scale to really high clock speeds. AMD designed the Athlon to compete with the Pentium 3. They haven't really tried to design a CPU from scratch to compete with the Pentium 4 yet. So really, it's impressive that AMD has held on for this long. Though I still have the strong impression that this is more by Intel's choice than by AMD's skill.
Right on it. All the P4 initially was for (it was a really really lousy performed considering how high the Mhz was and what price it came for. Also, SSE2 and Intels strong will and skill was what saved the P4 from becoming a total failure. And in the same way, AMD might very well have to concentrate on SSE2 for success, initially. And, I think it's a perfectly acceptable way of doing it. If SSE2 makes all of the Athlon 64's perform better than the XP3200+, then so be it. The new product is out, it's 64-bit compatible, performs good, has a working integrated thermal protection, and it comes for a good price. It's performing on par with the P4. If AMD can pull through with these things I mentioned, then A64 can become a success.Yeah. That's pretty much how I see it. I think that AMD will just be relying on optimized software to improve the A64's performance though, at least for this version of the core.
I suppose in a lot of ways it's really like when Intel first released the P4. Everyone said its FPU performance sucked. Intel just kept repeating "SSE2" back. At the time everyone laughed at a processor that needed optimized software just to do floating-point calculations decently. Yet now a darned lot of the software that we use does.
People may laugh at AMD for the bad pure 32-bit performance of their A64. I bet once we get used to software that is optimized to use the A64's full potential though, we'll get to the point where we won't really care either. All that will matter is that the software will run faster, I guess.
Well, here's to hoping for a successful Athlon64 launch sooner or later. Let's just hope that we have a version of Windows optimized for it.
Yeah. I heard that too. It's exactly the kind of thing that AMD needs to do if they want any good hype out of the A64.Hopefully it will change when it's time for release. Btw, I read somewhere that the A64 might be released in a limited quantity already in August. Good news in that case.
We'll really have to see some benchmarks comparing the AXP to the A64 on pure 32-bit performance. My guess is that an nForce2 running dual-channel with low-latency RAM (like Corsair 3200LL) is going to perform the same or better than the A64. However for optimized software the A64 should kick the AXP's butt.Exactly. I fear that too. Even though I'm now a proud owner of a spanking P4 2.8 with 800 FSB
Since the Athlon 64 is based on the K7 architecture, it's single-channel DDR and sometimes, SSE2 un-enabled performance might be just above the perf of the AXP, due to the big similiarities of the two CPU's.
I'd wager that in and of itself, the sc-ondie is slower than the dc-offdie. (And this will be especially and sickeningly true in any motherboards with an onboard graphics chip that shares system memory with the CPU.)You have a point there. I don't have a clue if the single-channel ondie controller is much faster than what offdie dual is. And, the longer pipeline will reduce the IPC by itself. So let's wait and see.
I'd swear that AMD is taking notes from Apple. They're both really making bad names for themselves lately and both by using the same tactics.I guess you're right. However AMD has really done much bad things lately, to market it's CPU's that they're the fastest available when they actually aren't, is a very bad thing. But you indeed are right. AMD desperately needs better management. Which, could lead to a well-needed burst in economy for the company.
Initially my wager was on manufacturing problems. Not that they couldn't produce the silicon, but just they were having problems getting the yields high enough to actually make money when they sold the A64 CPUs. (Since they'll cost less than the Opteron CPUs.)You could be right. There's several possible causes to what was the reason for the delay of the Athlon 64. Manufacturing problems, performance problems
Well those are my thoughts too, more or less. AMD is probably delaying the A64 because of a lack of an x86-64 version of Windows. I wouldn't be surprised if in fact an x86-64 verison of Windows had in it a strange scheme to run 32-bit apps on the A64 kind of like how Intel utilizes HT. It'd be some weird method where when multi-threading 32-bit apps Windows utilizes the extra registers in the A64 as an imaginary second processor. I'd be an interesting concept to try at least. I can't say such code will be in any version of Windows, but it'd make sense. It would allow the A64 to run 32-bit code really well compared to how it would run 32-bit code in a pure 32-bit mode. So maybe that's what AMD is holding out for.or the more unlikely one of them all, that Microsoft hasn't yet released their x86-64 OS, which after all, when I think about it, might be a valid claim, because that OS would, when it comes out, bring better 64-bit capability with it, thus increasing the power and value for the Athlon 64. So in a way, AMD has a right to claim that. Of course, they can't claim it in a way to accuse Microsoft of being the cause of the delay. They could say they're delaying it because their CPU's lack performance without a 64-bit capable OS. I hope I'm making sense here for both myself and everyone else who might be reading this
This all, was probably just what you meant as well, slvr, just writing down my own thoughts as well
We'll have to wait and see. My guess is that the 32-bit performance will be better, but only if running a special version of Windows. If you were to stick the A64 into a pure 32-bit mode with pure 32-bit software I bet the AXP would be slightly faster. With any luck though it won't matter.Anyways, AMD better have good 32-bit performance as well, it really must be a good piece better than the perf of the Athlon XP.
I'm hoping so. It'd be almost crazy for AMD to not have that kind of a backup plan. Heck, it'd even make sense to do it anyway.That sounds good, both as a thought and as a marketing strategy for AMD
That could very well be what they was planning actually.
I'm hoping that AMD will add SOI to Barton and keep the AXP line alive and kicking no matter if the A64 is a success and released on time (or early) or not. Intel isn't pushing 64-bit only home users yet, so likely a lot of software will remain 32-bit. Thus the AXP will still have an important role.But now when they A64 seems to really be released in August/September, I guess they won't have to add SOI to Barton.
The A64 will likely have a big number of sales in the beginning, if for no other reason than a lot of people are looking at throwing an x86-64 version of Linux on an A64 PC and run it as an ultra-cheap 64-bit server. It could very well become the low-cost small-business server of the future.Unless A64 is lacking that last edge to put Barton to rest. In that case, A64 could be the high-end, and Barton some mid-end to high-end alternative, with a only slighter lower price, so that users would prefer to buy the A64. Unless the A64 is MUCH faster than the AXP, a low price is what would initially need to be done to save A64 from a possible low sale.
I do too. I just say supposedly because in theory from what I've read the 0.09 micron tools just aren't ready yet. Yet Intel is making working Prescott engineering samples. So either Intel found a way to circumvent the tools that aren't ready yet (such as with prototypes of the tools perhaps) or they're making 0.13 micron Prescotts right now. I'm voting on the former, but you never know...I strongly believe the first Prescott is on .09.
I don't think AMD has been 'forced' to follow Intel's footsteps. I think that AMD has just lacked the funding / inginuity / manpower / drive (any or all of these things) to try making something bold and new. I really just don't think that the folks at AMD are trying very hard to compete anymore. They've had plenty of opportunities to do more than they have, and they've wasted each and every one of them.Yeah. Since Intel is the market leader, AMD has until now mostly been quite forced to follow in Intels footsteps instead of inventing entirely new creations of their own. This might change now, hopefully so.
I have to disagree. Software engineers in general have gotten to be incredibly lazy at optimizing their code because PCs come with faster and faster processors and more and more RAM. Hard drive space isn't even a worry at all anymore. While these advancements themselves are 'good' things it's made the software itself get worse and worse in quality. Most software just isn't using hardware to its full potential anymore.Competition is always good.
Well, the A64 will probably be a marginal success no matter what AMD does. Some people will love it for it's 64-bit capabilities. In the end most people won't see 64-bit as being worth it, at least not for another few years until software can really catch up. And software won't really catch up until software developers know that they will have enough customers desiring a 64-bit version of the software to make it worth spending the time and money to make. It'll likely be a slow process. There will be an initial rush of customers for the A64, but that will likely quickly taper off into the typical minority of people who buy high-end PCs.Right on it. All the P4 initially was for (it was a really really lousy performed considering how high the Mhz was and what price it came for. Also, SSE2 and Intels strong will and skill was what saved the P4 from becoming a total failure. And in the same way, AMD might very well have to concentrate on SSE2 for success, initially. And, I think it's a perfectly acceptable way of doing it. If SSE2 makes all of the Athlon 64's perform better than the XP3200+, then so be it. The new product is out, it's 64-bit compatible, performs good, has a working integrated thermal protection, and it comes for a good price. It's performing on par with the P4. If AMD can pull through with these things I mentioned, then A64 can become a success.
Yeah. I heard that too. It's exactly the kind of thing that AMD needs to do if they want any good hype out of the A64.
That sounds likely, unless AMD can ramp the Mhz up to something like 2.1 or 2.2 Ghz, whereas the A64 _should_ conquer over the AXP even in programs where there's no SSE2 support.We'll really have to see some benchmarks comparing the AXP to the A64 on pure 32-bit performance. My guess is that an nForce2 running dual-channel with low-latency RAM (like Corsair 3200LL) is going to perform the same or better than the A64. However for optimized software the A64 should kick the AXP's butt.
Is that so? That sounds bad indeed. Did AMD really need to include just a Single-channel memory interface on the A64 to make it perform enough much worse than how the Opteron performs, to be sure not to derail the success of the Opteron?I'd wager that in and of itself, the sc-ondie is slower than the dc-offdie. (And this will be especially and sickeningly true in any motherboards with an onboard graphics chip that shares system memory with the CPU.)
Shame on themI'd swear that AMD is taking notes from Apple. They're both really making bad names for themselves lately and both by using the same tactics.
Agreeing on that. Haven't AMD had lots of problems ramping up raw mhz lately? They have, if my memory serves me.Initially my wager was on manufacturing problems. Not that they couldn't produce the silicon, but just they were having problems getting the yields high enough to actually make money when they sold the A64 CPUs. (Since they'll cost less than the Opteron CPUs.)
Now though, I'm thinking that it's really just a matter of their 32-bit performance and the lack of an x86-64 version of Windows. If they can get the latter then they can hide the former by not doing pure 32-bit comparisons.
Yeah no doubt there's a risk that AXP and A64 would be too similar in 32-bit performance for AMD to be really happy.We'll have to wait and see. My guess is that the 32-bit performance will be better, but only if running a special version of Windows. If you were to stick the A64 into a pure 32-bit mode with pure 32-bit software I bet the AXP would be slightly faster. With any luck though it won't matter.
Here I'm not sure what we were talking aboutI'm hoping so. It'd be almost crazy for AMD to not have that kind of a backup plan. Heck, it'd even make sense to do it anyway.
Of course. Only business people would make good use of 64-bit systems as of today.I'm hoping that AMD will add SOI to Barton and keep the AXP line alive and kicking no matter if the A64 is a success and released on time (or early) or not. Intel isn't pushing 64-bit only home users yet, so likely a lot of software will remain 32-bit. Thus the AXP will still have an important role.
Supposedly, Intel is lowering prices on the IA-64 to make people use one on an Linux 64-bit OSThe A64 will likely have a big number of sales in the beginning, if for no other reason than a lot of people are looking at throwing an x86-64 version of Linux on an A64 PC and run it as an ultra-cheap 64-bit server. It could very well become the low-cost small-business server of the future.
AMD will tout the 64-bitness as uber-important and so even if the A64 doesn't perform so great and even if it costs more than the AXP people will be glad to pay that price. So you're right in that it'll probably be the new high-to-middle end for AMD and the AXP will become the middle-to-low end.
That is right I suppose.I do too. I just say supposedly because in theory from what I've read the 0.09 micron tools just aren't ready yet. Yet Intel is making working Prescott engineering samples. So either Intel found a way to circumvent the tools that aren't ready yet (such as with prototypes of the tools perhaps) or they're making 0.13 micron Prescotts right now. I'm voting on the former, but you never know...
That sounds reasonable. Especially lately when AMD seems to have been doing lotsa wrong, gearing up to 250 km/h when they introduced Barton, with it's low performance compared to it's steep price and rating.I don't think AMD has been 'forced' to follow Intel's footsteps. I think that AMD has just lacked the funding / inginuity / manpower / drive (any or all of these things) to try making something bold and new. I really just don't think that the folks at AMD are trying very hard to compete anymore. They've had plenty of opportunities to do more than they have, and they've wasted each and every one of them.
Oh I didn't know that...thanks for the info!Even now the A64 isn't exactly an original concept. A 64-bit version of x86 has been tossed around for years. The idea existed even before Intel's IA64 processors. It had just never been put into use because no one wanted to work with something that ugly anymore. Years later AMD picks up the concept, implements it in a half-arsed way (since as I understand it their x86-64 is missing a lot of registers that a 'true' 64-bit version of x86 should have) and modifies their hardly-changed Athlon core to utilize this hyrbid monstrosity. It's hardly any stunning feat of engineering. It's hardly even mediocre considering how much time they've spent working on it. The only reason that AMD beat everyone to it is because no one else wanted to even touch it in the first place.
Yeah, possible. Unless combined 32-bit and 64-bit usage becomes a real hit, and Intel gets problems with answering with an eventual Yamhill...If AMD wants to actually run as top dog again they're really going to have to come up with a completely new core. (Or at the very least spend more effort tweaking and improving their existing cores.)
All in all I've been pretty depressed with the lack of innovation into PCs in the last two years. Intel and AMD both have been uninspiring in my opinion.
You have a point here, definitely. Not sure if I'm agreeing or half-agreeing thoughI have to disagree. Software engineers in general have gotten to be incredibly lazy at optimizing their code because PCs come with faster and faster processors and more and more RAM. Hard drive space isn't even a worry at all anymore. While these advancements themselves are 'good' things it's made the software itself get worse and worse in quality. Most software just isn't using hardware to its full potential anymore.
A good period of time without any competition in hardware would see this corrected. It would also allow the PC that you buy today to be great at running software three years from now, instead of having to be constantly upgraded and/or replaced in half of that time just to keep up with the latest software.
Competition is also the reason for diverging standards. If there is no competition then only one group is making standards and everyone just has to follow that one standard. It's a much cleaner system.
Competition and stagnation both have their pros and cons, and really the 'best' situation is when you have a balance of both.
Well, the A64 will probably be a marginal success no matter what AMD does. Some people will love it for it's 64-bit capabilities. In the end most people won't see 64-bit as being worth it, at least not for another few years until software can really catch up. And software won't really catch up until software developers know that they will have enough customers desiring a 64-bit version of the software to make it worth spending the time and money to make. It'll likely be a slow process. There will be an initial rush of customers for the A64, but that will likely quickly taper off into the typical minority of people who buy high-end PCs.
I disagree with this. The Opteron was launched without 64-bit Windows support. When A64 launches in September, it will still be without 64-bit Windows support. And when the SP for server 2003 is released at the end of the year, which does support X86-64, it is the server version. I have not heard if MS plans to release a desktop version or not, but it seems clear to me that AMD is not waiting for 64-bit MS support. I suspect they are having much difficulty with yields at the 2 GHz speed, which has always been the speed they wanted to launch it at.Initially my wager was on manufacturing problems. Not that they couldn't produce the silicon, but just they were having problems getting the yields high enough to actually make money when they sold the A64 CPUs. (Since they'll cost less than the Opteron CPUs.)
Now though, I'm thinking that it's really just a matter of their 32-bit performance and the lack of an x86-64 version of Windows. If they can get the latter then they can hide the former by not doing pure 32-bit comparisons.
Of course they are still working out the kinks in the .09 process, but the tools are in fact working considerably well from what I've heard. They are still on track to launch Prescott on .09um in 2H, most likely sometime in Q4.I just say supposedly because in theory from what I've read the 0.09 micron tools just aren't ready yet. Yet Intel is making working Prescott engineering samples. So either Intel found a way to circumvent the tools that aren't ready yet (such as with prototypes of the tools perhaps) or they're making 0.13 micron Prescotts right now. I'm voting on the former, but you never know...