First *real-world* benches of Barcelona?

[CTho9305]

Originally posted by: firewolfsm
It has to be wrong, it just has to be...

He says that a 2GHz Barcelona scores 253...

well a 2GHz Opty Single core scores 256

http://www.nvnews.net/vbulletin/showthread.php?t=72826

God I hope it's wrong.

I doubt CPUmark99 benefits from many of the enhancements in Barcelona - it probably doesn't use SSE, it probably uses small (byte-length) instructions that don't benefit from the fetch bandwidth, it probably fits in the L1/L2 and doesn't benefit from the L3, it's clearly not threaded etc.

It doesn't make sense for CPU vendors to optimize their products for benchmarks that aren't used any more - SPEC2006 is much more important than SPEC92 or SPEC95, so given the choice between a 20% improvement in the latter two or a 5% improvement in the former, I'd expect the 5% to be the one that's implemented.

 

[classy]

Originally posted by: dmens

Originally posted by: classy

Originally posted by: dmens
Except a CPU has static resources, whereas a car engine can burn more fuel because a supercharger fed it compressed air. CPU core frequency increases cannot never yield greater than a linear performance gain.

That is certainly not true today, when cpus have so many other factors that conincide with cpu frequency. Such as multiple cores, HT, integrated memory controllers, enhanced buffers, enhanced branch chain prediction and the etc. So cpu frequency in general in todays cpus is not the dominant factor that it once was.

What I said still applies and always will.

You have yet to give any reason beyond the fact your still stuck in 1990. If clock frequency was the end all be all, how come Willamette sucked? If what you said was true why does Penryn yield greater performance than conroe at the same clock speed? Cpu architectures today are so complex I can easily see how a cpu could yield a greater performance increase in contrast to clock speed. Even right now we can't fully see the power of the cpus we have because of so many other limitations. No offense, you got a few clues, but clearly your missing some others.

 

[Accord99]

Originally posted by: classy
You have yet to give any reason beyond the fact your still stuck in 1990. If clock frequency was the end all be all, how come Willamette sucked? If what you said was true why does Penryn yield greater performance than conroe at the same clock speed? Cpu architectures today are so complex I can easily see how a cpu could yield a greater performance increase in contrast to clock speed. Even right now we can't fully see the power of the cpus we have because of so many other limitations. No offense, you got a few clues, but clearly your missing some others.

Because you are comparing two different families of CPU. But for one CPU, a 33% increase in clockspeed at best will only result in a 33% increase in performance. And since most applications require memory access, HD access or video card, then that 33% increase in clockspeed starts to diminish in importance since the application is not spending all of its time in the CPU.

 

[CTho9305]

Originally posted by: classy

Originally posted by: dmens

Originally posted by: classy

Originally posted by: dmens
Except a CPU has static resources, whereas a car engine can burn more fuel because a supercharger fed it compressed air. CPU core frequency increases cannot never yield greater than a linear performance gain.

That is certainly not true today, when cpus have so many other factors that conincide with cpu frequency. Such as multiple cores, HT, integrated memory controllers, enhanced buffers, enhanced branch chain prediction and the etc. So cpu frequency in general in todays cpus is not the dominant factor that it once was.

What I said still applies and always will.

You have yet to give any reason beyond the fact your still stuck in 1990. If clock frequency was the end all be all, how come Willamette sucked? If what you said was true why does Penryn yield greater performance than conroe at the same clock speed? Cpu architectures today are so complex I can easily see how a cpu could yield a greater performance increase in contrast to clock speed. Even right now we can't fully see the power of the cpus we have because of so many other limitations. No offense, you got a few clues, but clearly your missing some others.

I'd suggest you give up now. I rarely agree with dmens, but in this case he's right. dmens is saying that an x% frequency increase will produce at most an x% performance improvement on a given microarchitecture. All those features you mentioned (cores, branch prediction, etc) running at x% higher clock won't increase performance more than x%. Of course changing architectures means you can no longer compare based on frequency, but I think dmens is taking issue with Gary claiming that x% freq increase => >x% performance increase on the same microarchitecture.

 

[Keysplayr]

Originally posted by: dmens

Originally posted by: keysplayr2003
If the L2/L3 cache runs at twice the speed of the core, then I can see this happening. At 2GHz, a Barcelona's cache runs at 4GHz (hypothetically). Pump up the core speed to 2.6 GHz and the cache will be at 5.2GHz. So with only a 600MHz core increase, the cache speed increased 1200MHz.

The Pentium 4 did this, but after cranking up the clocks so high, diminishing returns were experienced. Wall was hit.

I'm not saying Barcelona will be this way, but it could explain the "power band" of the processor above 2.4GHz.

P4 ran parts of the ALU at double the core frequency, not the cache. Even with that kind of pipeline, frequency increases cannot not yield greater than a linear increase performance.

Apologies, I thought it was the cache.

 

[Hulk]

Originally posted by: CTho9305

Originally posted by: firewolfsm
It has to be wrong, it just has to be...

He says that a 2GHz Barcelona scores 253...

well a 2GHz Opty Single core scores 256

http://www.nvnews.net/vbulletin/showthread.php?t=72826

God I hope it's wrong.

I doubt CPUmark99 benefits from many of the enhancements in Barcelona - it probably doesn't use SSE, it probably uses small (byte-length) instructions that don't benefit from the fetch bandwidth, it probably fits in the L1/L2 and doesn't benefit from the L3, it's clearly not threaded etc.

It doesn't make sense for CPU vendors to optimize their products for benchmarks that aren't used any more - SPEC2006 is much more important than SPEC92 or SPEC95, so given the choice between a 20% improvement in the latter two or a 5% improvement in the former, I'd expect the 5% to be the one that's implemented.

That all sounds great until you consider the fact that C2D kicks ass on the CPUmark99 benchmark and it also kicks ass on real world applications. It would be strange and unlikely for Barcelona to not do well on that bench yet kick C2D's ass in real world applications.

 

[Arkaign]

Originally posted by: classy

Originally posted by: dmens

Originally posted by: classy

Originally posted by: dmens
Except a CPU has static resources, whereas a car engine can burn more fuel because a supercharger fed it compressed air. CPU core frequency increases cannot never yield greater than a linear performance gain.

That is certainly not true today, when cpus have so many other factors that conincide with cpu frequency. Such as multiple cores, HT, integrated memory controllers, enhanced buffers, enhanced branch chain prediction and the etc. So cpu frequency in general in todays cpus is not the dominant factor that it once was.

What I said still applies and always will.

You have yet to give any reason beyond the fact your still stuck in 1990. If clock frequency was the end all be all, how come Willamette sucked? If what you said was true why does Penryn yield greater performance than conroe at the same clock speed? Cpu architectures today are so complex I can easily see how a cpu could yield a greater performance increase in contrast to clock speed. Even right now we can't fully see the power of the cpus we have because of so many other limitations. No offense, you got a few clues, but clearly your missing some others.

Bleh, I hate to join in this group pounding you're receiving, but the sheer ridiculousness of your post sort of invites it.

(1)- No one that I see in this thread has said anything close to 'clock frequency is the end all be all.'

(2)- Nobody here is stuck in 1990, except perhaps yourself.

(3)- Willamette was an okay performer for the time, as it split the benchmarks at the 1.5Ghz launch-day top end speed with the Athlons available at the time. Willamette's real problems were the ridiculous cost of the processor and RDRAM to go with it. The FACT remains, that when increasing clock speed while all other variables remain static (core revision, memory speed and latency, etc, etc) you will at BEST garner an equal % of performance along with the raised clock speed.

(4)- Following the undeniable logic of (3), you can see the significant problem facing K10 in today's market. If the IPC performance of K10 is 1:1 vs. Conroe, then it needs to be clocked at least as high as C2D to equal it's performance. Launching at such pathetic speeds bodes ill for AMD at this juncture. There is one caveat, also noted in this thread : If K10's scaling penalty (the unavoidable diminishing returns as clock speeds increase, eq : raising clock speed 10% gives 9% real-world performance increase, increasing clock speed 20% gives 17% real-world performance increase, and so on) is less severe than for Conroe or Penryn, then a 3Ghz K10 may be a few % faster than C2D/C2Q.

(5)- The final point here, is that nobody knows for sure what real-world performance of K10, not to mention overclocking potential, heat/power profile, and so on. About the ONLY thing that is for sure is that you cannot exceed a 1:1 increase in performance with a given clock speed increase, and that's the one thing you so foolishly mouthed off about in sheer ignorance (or perhaps a bone-headed moment that lacked clarity, I have those myself too!).

P.s. - It's good for everyone (even Intel, I think) if K10 is a real contender. With what little we know at the moment, it looks very iffy at best. The most potent problem facing AMD is the fact that C2D/C2Q cores are already so stable at even the mid-3ghz range, showing the potential to just roll competing processors right off the line, WITHOUT NEEDING A RE-SPIN. Literally just a different multiplier, and perhaps a better boxed cooler. That is a hard nut to crack if I've ever seen one. I think AMD needs to drop something that will top Intel's best current and also a theoretical retail 3.6ghz C2D/C2Q, which would be childishly easy to produce at the drop of a hat. This alone will have the potential value to increase AMD's ASP so that they stop hemmoraghing cash, which if they do not staunch, will destroy AMD before Q3 '08, perhaps sooner.

 

[classy]

Here is some proof, while small proves my point about clock speed.

This a review by techreport HERE.

In the review I looked at the scores of the 6000+ and 5600+. Exact cpu and tested with the exact same hardware. The 6000 has 7.14% advantage in clock speed over the 5600. But in the review you can see that in POV-Ray rendering while running one thread, the 6000 is a full 8% faster despite only having a 7.14% advantage in clock speed. There are several other benchmarks you see can where the 6000 performance exceeded the 7.14% advantage in clock speed, albeit ever so slightly. So while the linear statement does usually 95% of the time hold water, we are clearly starting to see that the development of not only the cpu but the platforms they run on, that the linear statement is not always the case. There are other instances in other reviews across the net as well as a testament of this. While I thought the second I seen the barcy benches they were bogus, I can completely see that there is more then a possible chance that the cpu could yield greater performance in comparison to clock speed. Hell HT alone can change the clock speed theory, let alone maybe half a dozen of other things.

 

[classy]

Originally posted by: Arkaign
This alone will have the potential value to increase AMD's ASP so that they stop hemmoraghing cash, which if they do not staunch, will destroy AMD before Q3 '08, perhaps sooner.

Hey look I am not a fanboy so I could care less who gets destroyed ok, so calm down. sheesh...... I am just simply stating the statement dmens made about clock speed is not entirely true. Simply put, the platform alone can do more with the same cpu cycles. So in the case of AMD with barcy, it is quite possible that the platform can bring about better performance percentage wise in comparison to clock speed increase, Thats all I am saying.

 

[Arkaign]

Originally posted by: classy
Here is some proof, while small proves my point about clock speed.

This a review by techreport HERE.

In the review I looked at the scores of the 6000+ and 5600+. Exact cpu and tested with the exact same hardware. The 6000 has 7.14% advantage in clock speed over the 5600. But in the review you can see that in POV-Ray rendering while running one thread, the 6000 is a full 8% faster despite only having a 7.14% advantage in clock speed. There are several other benchmarks you see can where the 6000 performance exceeded the 7.14% advantage in clock speed, albeit ever so slightly. So while the linear statement does usually 95% of the time hold water, we are clearly starting to see that the development of not only the cpu but the platforms they run on, that the linear statement is not always the case. There are other instances in other reviews across the net as well as a testament of this. While I thought the second I seen the barcy benches they were bogus, I can completely see that there is more then a possible chance that the cpu could yield greater performance in comparison to clock speed. Hell HT alone can change the clock speed theory, let alone maybe half a dozen of other things.

You're really reaching. Look at the rest of the benchmarks, and tell me what you see.

Also, the .86% differential is WELL within the margin for error/variance for testing purposes. If you run the same benchmark 100 times on the same setup, you'll get 100 SLIGHTLY difference results, depending on the granularity of the test output.

Also note that the 6000+ may have extremely minor, but real, differences in production. Even something with the same latency, cache, and bus speeds can have slight performance variance depending on the batch.

It remains a fundamental truth : increasing clock speed alone, WITH NO OTHER CHANGES, cannot account for an excess of 1:1 performance increase.

the techreport benchmark reveals a number of scores with underscore the truth of the unreliability of going by one score. Take the Valve Source particle test. The 6000+ scores LOWER than the 5600+ in that benchmark. Are we to extrapolate that data to judge that increasing clock speed on the X2 architecture (no other changes) LOWERS performance in that benchmark? No, and to think so is ludicrous. The only thing it proves is that you need a higher number of runs on the identical setup, and draw an average from them. At least 10x should be mandatory when looking for a more scientific methodology.

 

[classy]

Originally posted by: CTho9305

I'd suggest you give up now. I rarely agree with dmens, but in this case he's right. dmens is saying that an x% frequency increase will produce at most an x% performance improvement on a given microarchitecture. All those features you mentioned (cores, branch prediction, etc) running at x% higher clock won't increase performance more than x%. Of course changing architectures means you can no longer compare based on frequency, but I think dmens is taking issue with Gary claiming that x% freq increase => >x% performance increase on the same microarchitecture.

Thats problem with his statement. The cpu is only part of the microarchitecture. There are so many other factors that come into play its not even funny. In the case of HT alone, I don't know what effect that will have on performance numbers once it goes a certain speed. His whole point is that clock speed will determine the level of performance and that just isn't true

 

[CTho9305]

Originally posted by: classy
Here is some proof, while small proves my point about clock speed.

This a review by techreport HERE.

In the review I looked at the scores of the 6000+ and 5600+. Exact cpu and tested with the exact same hardware. The 6000 has 7.14% advantage in clock speed over the 5600. But in the review you can see that in POV-Ray rendering while running one thread, the 6000 is a full 8% faster despite only having a 7.14% advantage in clock speed. There are several other benchmarks you see can where the 6000 performance exceeded the 7.14% advantage in clock speed, albeit ever so slightly

<1% is in the noise. They don't give standard deviations or run counts ("tests run at least 3 times", no specifics), so there's no confidence interval. Maybe the 6000+ was running a few MHz faster and the 5600+ a few MHz slower. Maybe the real scores were 498.5 / 462.49 rather than 499 / 462.

 

[Arkaign]

Originally posted by: classy

Originally posted by: Arkaign
This alone will have the potential value to increase AMD's ASP so that they stop hemmoraghing cash, which if they do not staunch, will destroy AMD before Q3 '08, perhaps sooner.

Hey look I am not a fanboy so I could care less who gets destroyed ok, so calm down. sheesh...... I am just simply stating the statement dmens made about clock speed is not entirely true. Simply put, the platform alone can do more with the same cpu cycles. So in the case of AMD with barcy, it is quite possible that the platform can bring about better performance percentage wise in comparison to clock speed increase, Thats all I am saying.

I don't need to calm down, I wasn't shouting at anyone. I want AMD to survive, and they need to take drastic action soon to make that happen. They've lost nearly 2 BILLION dollars in 3 quarters. This is unsustainable, and to recover they need to increase their ASP. If their product remains inferior to the competition, it is impossible to raise the ASP, as next to noone will pay a premium for a weaker product, without brand/market control.

One more time : all other factors remaining the same, cpu performance cannot exceed a 1:1 ratio. It's never been done, and doesn't make any kind of logical sense whatsoever. This truth is impossible to refute with any credibility. A variance of less than 1% on one benchmark when all others show less than a 1:1 increase is proof of only the variance that you inevitably see in benchmark results. It's why averaging many benchmark runs is so important.

 

[Keysplayr]

Originally posted by: classy

Originally posted by: Arkaign
This alone will have the potential value to increase AMD's ASP so that they stop hemmoraghing cash, which if they do not staunch, will destroy AMD before Q3 '08, perhaps sooner.

Hey look I am not a fanboy so I could care less who gets destroyed ok, so calm down. sheesh...... I am just simply stating the statement dmens made about clock speed is not entirely true. Simply put, the platform alone can do more with the same cpu cycles. So in the case of AMD with barcy, it is quite possible that the platform can bring about better performance percentage wise in comparison to clock speed increase, Thats all I am saying.

Classy, I know you want K10 to scale better with clock speed. Hell, we all do for Pete's sake. We also know that clock speed isn't everything when comparing one architecture to another. But you also have to accept the possibility of the K10 getting diminishing returns as clock speeds go up. You can't have it one way. Both possibilities are present. So why not stop talking about K10 getting geometrically faster as clock speeds go up as if it were already a fact? AMD is being very "annoyingly" quiet. Intel, for the past year, has given us more than a fair share of glimpses of what is to come. That is something I appreciate. I don't appreciate being left in the dark to the last bloody second. This annoys me to the point where I would almost say, to hell with AMD. And if we don't get some preliminary benchmarks soon, so should you. Fair is fair.

 

[classy]

Originally posted by: keysplayr2003
Classy, I know you want K10 to scale better with clock speed. Hell, we all do for Pete's sake. We also know that clock speed isn't everything when comparing one architecture to another. But you also have to accept the possibility of the K10 getting diminishing returns as clock speeds go up. You can't have it one way. Both possibilities are present. So why not stop talking about K10 getting geometrically faster as clock speeds go up as if it were already a fact? AMD is being very "annoyingly" quiet. Intel, for the past year, has given us more than a fair share of glimpses of what is to come. That is something I appreciate. I don't appreciate being left in the dark to the last bloody second. This annoys me to the point where I would almost say, to hell with AMD. And if we don't get some preliminary benchmarks soon, so should you. Fair is fair.

hehehehehe You do know I am an Intel guy. Gotta Gallatin P4 extreme I am typing on right now and a Core2 sitting right next to it. To be honest I haven't used AMD in any of my main boxes for years.

But there are many instances where performance has exceeded the 1:1 ratio.

 

[justly]

Originally posted by: Arkaign

Originally posted by: classy
Here is some proof, while small proves my point about clock speed.

This a review by techreport HERE.

In the review I looked at the scores of the 6000+ and 5600+. Exact cpu and tested with the exact same hardware. The 6000 has 7.14% advantage in clock speed over the 5600. But in the review you can see that in POV-Ray rendering while running one thread, the 6000 is a full 8% faster despite only having a 7.14% advantage in clock speed. There are several other benchmarks you see can where the 6000 performance exceeded the 7.14% advantage in clock speed, albeit ever so slightly. So while the linear statement does usually 95% of the time hold water, we are clearly starting to see that the development of not only the cpu but the platforms they run on, that the linear statement is not always the case. There are other instances in other reviews across the net as well as a testament of this. While I thought the second I seen the barcy benches they were bogus, I can completely see that there is more then a possible chance that the cpu could yield greater performance in comparison to clock speed. Hell HT alone can change the clock speed theory, let alone maybe half a dozen of other things.

You're really reaching. Look at the rest of the benchmarks, and tell me what you see.

Also, the .86% differential is WELL within the margin for error/variance for testing purposes. If you run the same benchmark 100 times on the same setup, you'll get 100 SLIGHTLY difference results, depending on the granularity of the test output.

Also note that the 6000+ may have extremely minor, but real, differences in production. Even something with the same latency, cache, and bus speeds can have slight performance variance depending on the batch.

It remains a fundamental truth : increasing clock speed alone, WITH NO OTHER CHANGES, cannot account for an excess of 1:1 performance increase.

the techreport benchmark reveals a number of scores with underscore the truth of the unreliability of going by one score. Take the Valve Source particle test. The 6000+ scores LOWER than the 5600+ in that benchmark. Are we to extrapolate that data to judge that increasing clock speed on the X2 architecture (no other changes) LOWERS performance in that benchmark? No, and to think so is ludicrous. The only thing it proves is that you need a higher number of runs on the identical setup, and draw an average from them. At least 10x should be mandatory when looking for a more scientific methodology.

I have to agree with Arkaign. I was about to make a similar response but I think he said it well enough.

 

[dmens]

Originally posted by: classy
Thats problem with his statement. The cpu is only part of the microarchitecture. There are so many other factors that come into play its not even funny. In the case of HT alone, I don't know what effect that will have on performance numbers once it goes a certain speed. His whole point is that clock speed will determine the level of performance and that just isn't true

Why are you talking about external factors when I'm only making a statement on performance when core frequency is the only metric being changed? On a given processor family, once the overall uarch is set in stone, the only thing left that can improve performance is clock speed. Frequency is still extremely important, don't let the PR fool you.

Even if you include the HT bus in the equation, assuming the bus multiplier is held constant, increasing the bus and core clocks in lock step will still yield a maximum of linear performance increase. And I doubt even doubling the bandwidth will yield much more than 10-15% improvement (my guesstimate). The CPU is the entire computational machine, it has almost all the bottlenecks and resource limitations. K10 is hardly starving for data from the bus.

 

[Keysplayr]

Originally posted by: classy

Originally posted by: keysplayr2003
Classy, I know you want K10 to scale better with clock speed. Hell, we all do for Pete's sake. We also know that clock speed isn't everything when comparing one architecture to another. But you also have to accept the possibility of the K10 getting diminishing returns as clock speeds go up. You can't have it one way. Both possibilities are present. So why not stop talking about K10 getting geometrically faster as clock speeds go up as if it were already a fact? AMD is being very "annoyingly" quiet. Intel, for the past year, has given us more than a fair share of glimpses of what is to come. That is something I appreciate. I don't appreciate being left in the dark to the last bloody second. This annoys me to the point where I would almost say, to hell with AMD. And if we don't get some preliminary benchmarks soon, so should you. Fair is fair.

hehehehehe You do know I am an Intel guy. Gotta Gallatin P4 extreme I am typing on right now and a Core2 sitting right next to it. To be honest I haven't used AMD in any of my main boxes for years.

But there are many instances where performance has exceeded the 1:1 ratio.

You've only shown us one that is most likely a fluke. And it does it matter you are an Intel guy? I gave no inclination that that mattered. I'm an Intel guy as well. But I still want AMD to kick some booty. There comes time where one has to force him/herself to be neutral.

 

[SerpentRoyal]

Again, we don't have enough data at this time to form an educated guess. Bottomline is that AMD will still be trailing Intel going into the Xmas shopping season. Blue Boys still hold a commanding lead in core speed, and they could probably crank up the MHz another 20 to 40% if needed.

 

[bryanW1995]

Originally posted by: CTho9305

Originally posted by: classy

Originally posted by: dmens

Originally posted by: classy

Originally posted by: dmens
Except a CPU has static resources, whereas a car engine can burn more fuel because a supercharger fed it compressed air. CPU core frequency increases cannot never yield greater than a linear performance gain.

That is certainly not true today, when cpus have so many other factors that conincide with cpu frequency. Such as multiple cores, HT, integrated memory controllers, enhanced buffers, enhanced branch chain prediction and the etc. So cpu frequency in general in todays cpus is not the dominant factor that it once was.

What I said still applies and always will.

You have yet to give any reason beyond the fact your still stuck in 1990. If clock frequency was the end all be all, how come Willamette sucked? If what you said was true why does Penryn yield greater performance than conroe at the same clock speed? Cpu architectures today are so complex I can easily see how a cpu could yield a greater performance increase in contrast to clock speed. Even right now we can't fully see the power of the cpus we have because of so many other limitations. No offense, you got a few clues, but clearly your missing some others.

I'd suggest you give up now. I rarely agree with dmens, but in this case he's right. dmens is saying that an x% frequency increase will produce at most an x% performance improvement on a given microarchitecture. All those features you mentioned (cores, branch prediction, etc) running at x% higher clock won't increase performance more than x%. Of course changing architectures means you can no longer compare based on frequency, but I think dmens is taking issue with Gary claiming that x% freq increase => >x% performance increase on the same microarchitecture.

you guys are missing the point. Say that cache increases at double the rate of the clock or amd has some other great new idea in barcelona. It doesn't have to increase at greater than a 1:1 ratio with clock speed, it doesn't even have to be close to that. It just has to increase true performance more rapidly with higher clocks than penryn or conroe.

 

[SickBeast]

My guess with respect to this 'non-linear scaling' issue is simply that the memory is too fast for the memory controller until the CPU clockspeed reaches a certain point.

In the A64, the speed of the memory controller is directly related to the CPU clockspeed. Perhaps the chip is memory bottlenecked until it is clocked high enough to take advantage of high-speed DDR2 memory.

 

[bryanW1995]

Originally posted by: Arkaign

Originally posted by: classy

Originally posted by: Arkaign
This alone will have the potential value to increase AMD's ASP so that they stop hemmoraghing cash, which if they do not staunch, will destroy AMD before Q3 '08, perhaps sooner.

Hey look I am not a fanboy so I could care less who gets destroyed ok, so calm down. sheesh...... I am just simply stating the statement dmens made about clock speed is not entirely true. Simply put, the platform alone can do more with the same cpu cycles. So in the case of AMD with barcy, it is quite possible that the platform can bring about better performance percentage wise in comparison to clock speed increase, Thats all I am saying.

I don't need to calm down, I wasn't shouting at anyone. I want AMD to survive, and they need to take drastic action soon to make that happen. They've lost nearly 2 BILLION dollars in 3 quarters. This is unsustainable, and to recover they need to increase their ASP. If their product remains inferior to the competition, it is impossible to raise the ASP, as next to noone will pay a premium for a weaker product, without brand/market control.

One more time : all other factors remaining the same, cpu performance cannot exceed a 1:1 ratio. It's never been done, and doesn't make any kind of logical sense whatsoever. This truth is impossible to refute with any credibility. A variance of less than 1% on one benchmark when all others show less than a 1:1 increase is proof of only the variance that you inevitably see in benchmark results. It's why averaging many benchmark runs is so important.

2 things:
1. See GM or Ford info if you think that losing 2billion in 3 quarters is unsustainable.
2. observer-created reality makes a lot less sense than a 1:1 or greater scaling ratio of cpus.

 

[Arkaign]

Originally posted by: bryanW1995

Originally posted by: Arkaign

Originally posted by: classy

Originally posted by: Arkaign
This alone will have the potential value to increase AMD's ASP so that they stop hemmoraghing cash, which if they do not staunch, will destroy AMD before Q3 '08, perhaps sooner.

Hey look I am not a fanboy so I could care less who gets destroyed ok, so calm down. sheesh...... I am just simply stating the statement dmens made about clock speed is not entirely true. Simply put, the platform alone can do more with the same cpu cycles. So in the case of AMD with barcy, it is quite possible that the platform can bring about better performance percentage wise in comparison to clock speed increase, Thats all I am saying.

I don't need to calm down, I wasn't shouting at anyone. I want AMD to survive, and they need to take drastic action soon to make that happen. They've lost nearly 2 BILLION dollars in 3 quarters. This is unsustainable, and to recover they need to increase their ASP. If their product remains inferior to the competition, it is impossible to raise the ASP, as next to noone will pay a premium for a weaker product, without brand/market control.

One more time : all other factors remaining the same, cpu performance cannot exceed a 1:1 ratio. It's never been done, and doesn't make any kind of logical sense whatsoever. This truth is impossible to refute with any credibility. A variance of less than 1% on one benchmark when all others show less than a 1:1 increase is proof of only the variance that you inevitably see in benchmark results. It's why averaging many benchmark runs is so important.

2 things:
1. See GM or Ford info if you think that losing 2billion in 3 quarters is unsustainable.
2. observer-created reality makes a lot less sense than a 1:1 or greater scaling ratio of cpus.

GM and Ford are many many many many times larger than AMD, and the auto market is far more stable than electronics.

 

[SerpentRoyal]

GM & F loss $ because of the massive benefits (past, present, and future) given to the union.

 

[swtethan]

Originally posted by: Arkaign

Originally posted by: bryanW1995

Originally posted by: Arkaign

Originally posted by: classy

Originally posted by: Arkaign
This alone will have the potential value to increase AMD's ASP so that they stop hemmoraghing cash, which if they do not staunch, will destroy AMD before Q3 '08, perhaps sooner.

Hey look I am not a fanboy so I could care less who gets destroyed ok, so calm down. sheesh...... I am just simply stating the statement dmens made about clock speed is not entirely true. Simply put, the platform alone can do more with the same cpu cycles. So in the case of AMD with barcy, it is quite possible that the platform can bring about better performance percentage wise in comparison to clock speed increase, Thats all I am saying.

I don't need to calm down, I wasn't shouting at anyone. I want AMD to survive, and they need to take drastic action soon to make that happen. They've lost nearly 2 BILLION dollars in 3 quarters. This is unsustainable, and to recover they need to increase their ASP. If their product remains inferior to the competition, it is impossible to raise the ASP, as next to noone will pay a premium for a weaker product, without brand/market control.

One more time : all other factors remaining the same, cpu performance cannot exceed a 1:1 ratio. It's never been done, and doesn't make any kind of logical sense whatsoever. This truth is impossible to refute with any credibility. A variance of less than 1% on one benchmark when all others show less than a 1:1 increase is proof of only the variance that you inevitably see in benchmark results. It's why averaging many benchmark runs is so important.

2 things:
1. See GM or Ford info if you think that losing 2billion in 3 quarters is unsustainable.
2. observer-created reality makes a lot less sense than a 1:1 or greater scaling ratio of cpus.

GM and Ford are many many many many times larger than AMD, and the auto market is far more stable than electronics.

would GM be 100x larger?