Here's what this 4.6GHz number tells me. First, it tells me that either the presenter at ISSCC was ridiculously vague or that the journalist is an idiot. Call me cynical, but I doubt that even the mighty Sony has been able to clock a supposed POWER4 or POWER5 core to around double its previous frequency. So it could be a new design. If this is the case, the clockspeed tells me that the core is either extremely simple or has more pipeline stages than Prescott. We've been through this before, and that wouldn't be good. For the record, I kind of doubt that scenario is accurate.
Last I heard, the 4.6GHz number was actually the tested clock frequency of the SRAM on the chip. That's impressive and all, but it hardly begins to indicate performance of the SRAM itself. For all we know, it could run at 4.6GHz but only with 20 cycles of latency.
The short version? We need more details. Try not to get caught up in the hype machine for once.
I wish everyone would stop calculating dual core performance by just adding up speed ratings. The increase in performance from adding another core is ZERO for any single-threaded application. Likewise, the benefit for multi-threaded apps depends mostly on how much parallelism was extracted from the application.
If we're talking about an application that exhibits speedup that's near the absolute limit (2x speedup from adding an extra thread), then we can start talking about whether AMD or Intel is faster. The problem is, almost no desktop applications work that way. For example, you'll see about zero benefit for most games.
AMD's dual core solution may well be better than Intel's. The problem is that the parent post doesn't explain why. It hardly even presents any facts at all.
You'd kill all of them?! I admit, the "them" you refer to is ambiguous. But you also said that, by killing them all, you will free them from evil. Care to explain this further or should I assume that you're sociopathic? I'm pretty sure Jesus refrains from advocating "killing all of them" in the version of the Bible sitting on my shelf.
Cache memory is faster because it is closer to the processor and because it's smaller. Decoding hardware necessarily gets slower as the amount of data to decode/encode/address increases. Additionally, SRAM doesn't have to be refreshed periodically to maintain its data.
An SRAM bit also takes 6 (or 4) transistors compared to 1 for DRAM, but is more suited to fabrication on a process optimized for logic (as opposed to one optimized for DRAM or mixed-signal circuitry).
Why, praytell, does Intel need to either specialize their cores or move toward a massively parallel architecture like Cell is supposed to be? Do you have some foresight into the performance of as of yet nonexistent computing architectures?
Anyone even vaguely familiar with multithreaded programming and execution should know that it is difficult (sometimes impossible) to extract parallelism from the majority of the desktop computing applications. As such, throwing more cores at the problem doesn't necessarily help. I'm repeating myself, but it is very likely that next year's dual-core systems will perform worse than today's single core versions in most applications (which aren't multithreaded).
Forgive me if I'm somewhat skeptical of the Cell architecture as well. (We all know how hard the Sony hype machine worked for the Emotion engine.) It's not very likely that Cell will achieve some of the amazing performance numbers that have been hinted at. Perhaps Sony/IBM/Toshiba has a revolutionary SDK that allows easy multithreading of sequential applications. But nothing like that has even come up in rumormongering.
I'm not sure why this comment got modded down. Fact is, Sony hyped the "Emotion Engine" until it was blue in the face. That undeniably helped kill the Dreamcast, despite the PS2 having a far poorer library of games (and inferior) graphics for at least the first year of its existence.
I think it's reasonable to assume that there's some serious hype going on with Cell too. Short of some revolutionary SDK, I don't see how all these processing elements will be able to work together efficiently for gaming.
If you don't care about gaming, then buy a Radeon 7000 for $30 and be happy. 2-D hasn't been a concern for, gosh, at least six or so years now. DVD playback isn't a concern anymore either. The only new non-gaming feature I can even think of is the MPEG-2 HDTV encoder/decoder on the Geforce 6x00 series.
SOI only helps reduce one particular source of static power consumption. While static power is a big issue at 90nm, SOI doesn't magically solve it. Further, the big problem with Prescott is power dissipation and heat under load--dynamic power consumption. I'm not sure where you heard this rumor, but even if true it's ancillary to the current discussion.
I'm not sure that anyone will read this, as I'm late to the party and replying to something way, way deep in the comment tree, but I feel like I need to respond.
I'm a Christian (Presbyterian to be exact) and Bush is hardly a "great leader" for me. Christianity at its roots values love and tolerance, two ideals which I'd hardly associate with the current administration. The Christians whose values coincide best with Bush's are evangelical and fundamentalist. I and many other Christians I'm sure, think that those particular sects give the rest of us a bad name.
It really saddens me to see that "Christian" is almost becoming derogatory because of a vocal minority. I just wish people would remember that most of us also value things like science and the right of everyone to determine their own way of life.
I've played Doom 3. To put it bluntly, it bored me. Sure the graphics were insane, even on a lowly 9500 Pro. Sure the first few monsters popping out of nowhere were cool. And there were some freaky moments. But there's only so much running down dark corridors I can deal with.
It's all speculation now, but the breadth of experience that HL2 seems to offer should make it far more interesting to me. It's unfortunate that they didn't build in regular deathmatch though. I could do without Counterstrike (gg awp).
The real question is: why bother cooking up a DP version of 915 when they probably already have working prototypes of a similar Xeon motherboard? That would be an awful lot of effort just to be able to lie to the press.
Here is a link to a multi-format double-blind listening test of several codecs at 128kbps. Not to give away the ending, but LAME MP3 and iTunes AAC are tied in 2nd place, behind Ogg. Somehow I doubt that 192k LAME MP3 is inferior to 128k AAC.
The issue of power consumption is rapidly becoming much more significant than even the parent poster realizes.
The general dynamic power (operating power) equation for CMOS circuits has switching frequency as a squared term. Voltage and junction capacitance (think die size here) are also present, but are not squared.
If Intel were to take the P-IV architecture as far as it had planned, an extra few bucks for electricity would be the least of its worries. Without some unforseen advancement, power per unit area would become a (relatively) intractable problem. Even though voltage and die size would probably decrease, the increase in frequency coupled with the reduced area would likely provide a serious problem for cooling. I've read papers that have estimated that air cooling won't be able to dissipate much more heat than it's already required to. Taken far enough, the head produced could just vaporize the silicon (obviously that's not occurring in the near future).
Psst, don't tell anybody, but somebody already made a game about invading the USA. It was Westwood and the game was called Red Alert 2. Personally, I thought running through the streets of DC and garrisoning inside the Smithsonian was fun. But what do I know? My sense of being offended at fiction isn't very well developed.
You have an American V8 or V6 or I4 car that can out accelerate, out corner, and out perform a Subaru WRX, WRX STi or a Mitsubishi Evo IV, V, VI, VII or the rally edition Audi Quattro (for a small example) and I will call you a liar straight to your face.
Well, I think it's time to call me a liar straight to my face. The Corvette Z06 outperforms all the cars you listed, unless you're racing in the rain. In fact, I'd be surprised if an F-body wouldn't turn in faster lap times than all those cars (autocross excluded).
I think maybe you need to lay off the Gran Turismo.
I'm a computer engineer (at VT) graduating next May and I have a really high GPA. I do not "suck up to the teachers" or hang out in their offices. I don't spend every waking hour studying. And I definitely don't ask stupid, irrelevant questions in class to prove how smart I [think I] am.
Did it ever occur to you that there might be people who can comprehend the subject matter, get their work done, and get good grades for it?
It's been my experience that the suck-ups aren't necessarily the ones getting the good grades anyway.
I saw the yardstick/newspaper thing on Mr. Wizard (best show ever) when I was a kid. It doesn't involve any explosions, but it is pretty counterintuitive.
Re:Fibinacci Number Algorithm.... Very fun
on
Deep Algorithms?
·
· Score: 1
Ah, yes. The above implementation is the textbook example of situations in which one shouldn't use recursion.
Here's my story. The first drive I ordered, back in March, came DOA. I figured that wasn't a big deal, and RMA'ed the drive. I got another one that appeared to work fine.
Fast forward to about a month or so ago. My drive gave me no clicking noise indications of failure or anything, but I became unable to access anything on the disk one time I booted the machine.
I ran DFT, and found the good old 0x70 error code, and that bad sectors were found about half way through scanning the disk. Other than DFT, no tool that I have would read anything or scan the disk.
I've contacted IBM, specifically stating that I will not accept another of these disks, since they obviously have reliability problems. Monday, I talked with their support, who requested all the system information I could provide them (thanks to SiSoft Sandra that was easy). Also they requested the.BLZ file from the DFT disk, which I also provided.
The rep I talked to acknowledged that if I needed the disk replaced I wouldn't want the same model, but I haven't heard anything since Monday. IBM seemed anxious to prove that my problem was due to the shutdown bug, but realized that the evidence wasn't there. I run Windows 2000 (longer shutdown cycle), my processor is slow by today's standards (550MHz PIII Katmai) and I was only running ATA66 rather than 100.
I only hope that all these responses will compel IBM to actually admit to and do something about this problem.
Slashdot Mirror
Here's what this 4.6GHz number tells me. First, it tells me that either the presenter at ISSCC was ridiculously vague or that the journalist is an idiot. Call me cynical, but I doubt that even the mighty Sony has been able to clock a supposed POWER4 or POWER5 core to around double its previous frequency. So it could be a new design. If this is the case, the clockspeed tells me that the core is either extremely simple or has more pipeline stages than Prescott. We've been through this before, and that wouldn't be good. For the record, I kind of doubt that scenario is accurate.
Last I heard, the 4.6GHz number was actually the tested clock frequency of the SRAM on the chip. That's impressive and all, but it hardly begins to indicate performance of the SRAM itself. For all we know, it could run at 4.6GHz but only with 20 cycles of latency.
The short version? We need more details. Try not to get caught up in the hype machine for once.
I wish everyone would stop calculating dual core performance by just adding up speed ratings. The increase in performance from adding another core is ZERO for any single-threaded application. Likewise, the benefit for multi-threaded apps depends mostly on how much parallelism was extracted from the application.
If we're talking about an application that exhibits speedup that's near the absolute limit (2x speedup from adding an extra thread), then we can start talking about whether AMD or Intel is faster. The problem is, almost no desktop applications work that way. For example, you'll see about zero benefit for most games.
AMD's dual core solution may well be better than Intel's. The problem is that the parent post doesn't explain why. It hardly even presents any facts at all.
You'd kill all of them?! I admit, the "them" you refer to is ambiguous. But you also said that, by killing them all, you will free them from evil. Care to explain this further or should I assume that you're sociopathic? I'm pretty sure Jesus refrains from advocating "killing all of them" in the version of the Bible sitting on my shelf.
Touche. I meant the Cell in particular though.
An SRAM bit also takes 6 (or 4) transistors compared to 1 for DRAM, but is more suited to fabrication on a process optimized for logic (as opposed to one optimized for DRAM or mixed-signal circuitry).
Anyone even vaguely familiar with multithreaded programming and execution should know that it is difficult (sometimes impossible) to extract parallelism from the majority of the desktop computing applications. As such, throwing more cores at the problem doesn't necessarily help. I'm repeating myself, but it is very likely that next year's dual-core systems will perform worse than today's single core versions in most applications (which aren't multithreaded).
Forgive me if I'm somewhat skeptical of the Cell architecture as well. (We all know how hard the Sony hype machine worked for the Emotion engine.) It's not very likely that Cell will achieve some of the amazing performance numbers that have been hinted at. Perhaps Sony/IBM/Toshiba has a revolutionary SDK that allows easy multithreading of sequential applications. But nothing like that has even come up in rumormongering.
I think it's reasonable to assume that there's some serious hype going on with Cell too. Short of some revolutionary SDK, I don't see how all these processing elements will be able to work together efficiently for gaming.
I'm not really sure what the fascination is with card game simulations, but so it goes. Then again, maybe I shouldn't knock it until I play it.
If you don't care about gaming, then buy a Radeon 7000 for $30 and be happy. 2-D hasn't been a concern for, gosh, at least six or so years now. DVD playback isn't a concern anymore either. The only new non-gaming feature I can even think of is the MPEG-2 HDTV encoder/decoder on the Geforce 6x00 series.
SOI only helps reduce one particular source of static power consumption. While static power is a big issue at 90nm, SOI doesn't magically solve it. Further, the big problem with Prescott is power dissipation and heat under load--dynamic power consumption. I'm not sure where you heard this rumor, but even if true it's ancillary to the current discussion.
I'm not sure that anyone will read this, as I'm late to the party and replying to something way, way deep in the comment tree, but I feel like I need to respond.
I'm a Christian (Presbyterian to be exact) and Bush is hardly a "great leader" for me. Christianity at its roots values love and tolerance, two ideals which I'd hardly associate with the current administration. The Christians whose values coincide best with Bush's are evangelical and fundamentalist. I and many other Christians I'm sure, think that those particular sects give the rest of us a bad name.
It really saddens me to see that "Christian" is almost becoming derogatory because of a vocal minority. I just wish people would remember that most of us also value things like science and the right of everyone to determine their own way of life.
I've played Doom 3. To put it bluntly, it bored me. Sure the graphics were insane, even on a lowly 9500 Pro. Sure the first few monsters popping out of nowhere were cool. And there were some freaky moments. But there's only so much running down dark corridors I can deal with.
It's all speculation now, but the breadth of experience that HL2 seems to offer should make it far more interesting to me. It's unfortunate that they didn't build in regular deathmatch though. I could do without Counterstrike (gg awp).
The real question is: why bother cooking up a DP version of 915 when they probably already have working prototypes of a similar Xeon motherboard? That would be an awful lot of effort just to be able to lie to the press.
Looks like maybe I should read the other replies next time (or mod myself down as redundant).
Here is a link to a multi-format double-blind listening test of several codecs at 128kbps. Not to give away the ending, but LAME MP3 and iTunes AAC are tied in 2nd place, behind Ogg. Somehow I doubt that 192k LAME MP3 is inferior to 128k AAC.
I was trying to avoid getting into things like subthreshold leakage current (to explain why Vdd can't be reduced arbitrarily).
That's what I get for posting ten minutes after I wake up I guess.
The general dynamic power (operating power) equation for CMOS circuits has switching frequency as a squared term. Voltage and junction capacitance (think die size here) are also present, but are not squared.
If Intel were to take the P-IV architecture as far as it had planned, an extra few bucks for electricity would be the least of its worries. Without some unforseen advancement, power per unit area would become a (relatively) intractable problem. Even though voltage and die size would probably decrease, the increase in frequency coupled with the reduced area would likely provide a serious problem for cooling. I've read papers that have estimated that air cooling won't be able to dissipate much more heat than it's already required to. Taken far enough, the head produced could just vaporize the silicon (obviously that's not occurring in the near future).
In short: good move, Intel.
Psst, don't tell anybody, but somebody already made a game about invading the USA. It was Westwood and the game was called Red Alert 2. Personally, I thought running through the streets of DC and garrisoning inside the Smithsonian was fun. But what do I know? My sense of being offended at fiction isn't very well developed.
Well, I think it's time to call me a liar straight to my face. The Corvette Z06 outperforms all the cars you listed, unless you're racing in the rain. In fact, I'd be surprised if an F-body wouldn't turn in faster lap times than all those cars (autocross excluded).
I think maybe you need to lay off the Gran Turismo.
Oh give me a break.
I'm a computer engineer (at VT) graduating next May and I have a really high GPA. I do not "suck up to the teachers" or hang out in their offices. I don't spend every waking hour studying. And I definitely don't ask stupid, irrelevant questions in class to prove how smart I [think I] am.
Did it ever occur to you that there might be people who can comprehend the subject matter, get their work done, and get good grades for it?
It's been my experience that the suck-ups aren't necessarily the ones getting the good grades anyway.
I saw the yardstick/newspaper thing on Mr. Wizard (best show ever) when I was a kid. It doesn't involve any explosions, but it is pretty counterintuitive.
Ah, yes. The above implementation is the textbook example of situations in which one shouldn't use recursion.
Should I be disturbed that August 29th is my birthday?
May be a tad redundant, but I can't help add my favorite reaction:
"I sat through three hours of that and ain't nobody thrown no ring offa no mountain...I fell asleep like five times."
Here's my story. The first drive I ordered, back in March, came DOA. I figured that wasn't a big deal, and RMA'ed the drive. I got another one that appeared to work fine.
.BLZ file from the DFT disk, which I also provided.
Fast forward to about a month or so ago. My drive gave me no clicking noise indications of failure or anything, but I became unable to access anything on the disk one time I booted the machine.
I ran DFT, and found the good old 0x70 error code, and that bad sectors were found about half way through scanning the disk. Other than DFT, no tool that I have would read anything or scan the disk.
I've contacted IBM, specifically stating that I will not accept another of these disks, since they obviously have reliability problems. Monday, I talked with their support, who requested all the system information I could provide them (thanks to SiSoft Sandra that was easy). Also they requested the
The rep I talked to acknowledged that if I needed the disk replaced I wouldn't want the same model, but I haven't heard anything since Monday. IBM seemed anxious to prove that my problem was due to the shutdown bug, but realized that the evidence wasn't there. I run Windows 2000 (longer shutdown cycle), my processor is slow by today's standards (550MHz PIII Katmai) and I was only running ATA66 rather than 100.
I only hope that all these responses will compel IBM to actually admit to and do something about this problem.