why doesnt Intel release a 4Ghz?

[chrisf6969]

It seem like 4Ghz is pretty easy on the N0 stepping chips. IE: 600 series... so I wonder why Intel didn't release a 4Ghz chip.
Its not like back in the 1Ghz+ chip days when NOONE could get a the P3's to much over 1Ghz with out problems. It seems like 4Ghz is easy with the latest stepping.

a 4+ Ghz eXtreme Edition on a 1066+ FSB would probably sell well in Dell's XPS's desktops.

Maybe they jumped the gun on cancelling it, when they were still having problems with their 570's barely making it.

I would think that there would be a market for a 4Ghz+ 1Ghz+FSB EE chip. Even though a decent FX system would beat it. There are many Intel loyalists, and OEM's with only Intel chips (Dell).

Which could sell the 4+ Ghz cherry picked chips on a 1066FSB for the big pocket gamer.

NOT THAT I WOULD BUY!! But I think there's a market for it.

 

[pigpen]

I think you hit it on the head. Plans to ramp up the clock any further have already been scrapped. And very publicly too. I think it's a case of them needing to follow through with their current plans and not do any backtracking. Investors need to see some semblance of confidence.

But yes, i'm sure they could do it quite successfully if they wanted.

 

[Dillusion]

Because it's useless. The xtra 200mhz over a 570J (Which i own, and is awesome as it is) still doesnt put intel anywhere special, its merely a speed bump and a new name, maybe some extra cache- AMD is still edging over them. intel HAD plans for a 4GHZ chip but they scraped it in favor of their Dual Core line-up being released earlier than planned (Which they did), which was a smart move for intel releasing affordable dual core chips to the masses before AMD did.

Intel has some big plans, but 4ghz isnt one of them...

 

[Deam]

I agree it could easily be done. I have my 530J at 4 ghz with minimal effort.

I just do not understand why Intel doesn't drop the Pentium M bomb. As we have seen, a low-clocked Pentium M can surpass even the best chips out by both companies. Why not release a chip on that platform. It's like a secret weapon that isn't that secret, yet they won't deploy it.

The Pentium M made for desktop us and desktop features would put an end to AMD's performance superiority, which admitedly they do have.

 

[Stugots]

4ghz, dual core, hyperthreading, 4mb cache, 1066FSB...

mmmmm...

But I'm sure it'll never happen.

 

[Kickus_assius]

Yeah the whole not using the pentium M for desktops is ridiculous. Only a few companies are making mobos and slot converters to utilize it. If intel officially supported it, they would have the upper hand on AMD by far. I have seen benchmarks of a Pentium M 1.8 GHz Dothan, that they put in one of the special Pentium M desktop boards. They put a nice fan on it and got it up to 2.55GHz no problem. It absolutely crushed the FX55 in every benchmark they threw at it and in almost every single game as well.

 

[Dillusion]

4ghz, dual core, hyperthreading, 4mb cache, 1066FSB...

mmmmm...

But I'm sure it'll never happen.

Not a chance....

Pentium M is never going to be released into a full desktop format with its current architecture, its a pentium 3 for gods sake with extra cache and new instructions. intel has plans for the new revamped pentium M or whatever their gonna call it in 2006, and it will be able to fit in the desktop boards assigned to it, but there will be full ATX boards for it.

 

[Kickus_assius]

Not a chance....

Pentium M is never going to be released into a full desktop format with its current architecture, its a pentium 3 for gods sake with extra cache and new instructions. intel has plans for the new revamped pentium M or whatever their gonna call it in 2006, and it will be able to fit in the desktop boards assigned to it, but there will be full ATX boards for it.

I was under the impression that Pentium M was built from the ground up on new architechture. Is it based on Netburst architecture?

And to make it plausible now, all intel has to do is release a desktop chipset for it that has Dual Channel Memory and all of the features of their desktop P4 boards. The only thing holding Pentium M off of the desktops are the crappy outdated chipsets that the special Socket 479 boards and laptops have.

 

[Stugots]

The Pentium M is loosely based on the Tualatin.

 

[chrisf6969]

I know a Pentium M variant is coming... but thats still a ways off.

And its just a marketing # (milestone) that people would be like....."OOOOHHHH I want a 4Ghz processor!!"

I just think its a (financial / sales) mistake not to release one.

I know its only a 5% increase in clock over a 3.8Ghz so it wouldn't be much of a performance improvement. But it would be a nice profitable chip I bet.

The could do a 4.0Ghz, 1066Mhz FSB, Extreme Edition so easy

(Personally, I would just OC a 630 for that. 266.6 x 15 multi = 4Ghz)

Or they could go balls to the wall with a 4.2+ Ghz, 1200Mhz FSB, Extreme Edition

They could taut it at their next stock meeting as a HUGE success for their production / process facility to squeeze an extra speed grade or 2 more than they had originally expected.

 

[Donnie27]

Not a chance....

Pentium M is never going to be released into a full desktop format with its current architecture, its a pentium 3 for gods sake with extra cache and new instructions. intel has plans for the new revamped pentium M or whatever their gonna call it in 2006, and it will be able to fit in the desktop boards assigned to it, but there will be full ATX boards for it.

The Pentium M farther removed from the P3 than the Athlon64 is from the first K7.

Intel is smart to not ship Penitum M in its current form as a Desktop Processor. It get's its ass handed to it once you're through Gaming. Yes it uses less power, runs cooler and owns for games. Yet, any streaming apps shows its not the kill hardware as most folks tag it.

 

[chrisf6969]

The Pentium M farther removed from the P3 than the Athlon64 is from the first K7.

Intel is smart to not ship Penitum M in its current form as a Desktop Processor. It get's its ass handed to it once you're through Gaming. Yes it uses less power, runs cooler and owns for games. Yet, any streaming apps shows its not the kill hardware as most folks tag it.

pump up the FSB from weak 400/533 to 800 or 1066, add dual channel ram support, and add SSE3 and that will offset MOST of the problem with thos streaming apps you're talking about.

 

[Deam]

Exactly. I wasn't suggesting doing a direct port of the P-M to the desktop. But give it the features that are available today with its underlieing architecture, and you have a winner.

 

[SacLANd]

i believe conroe is slated for release in 2006.
its basically the P-M on steroids, for the desktop.

Intel is moving to low speed, low power, high IPC. you lose some performance in some applications, but make up for it by again increasing the clock speed.
but for those who like high ghz, dempsey and presler are for you.

 

[Oldie]

4ghz, dual core, hyperthreading, 4mb cache, 1066FSB...

mmmmm...

But I'm sure it'll never happen.

Uh, I'm pretty sure we'll eventually have 10Ghz chips with 100mb cache on a 4 Ghz bus. Granted not anytime soon, but I wouldn't say never. Intel having a 4Ghz offering within the next year and a half is actually pretty likely IMHO (which probably isn't worth much).

 

[xonik]

If Intel were to progress to 4 GHz, it would be a battle won, but a battle for a cause they have already abandoned. A 4 GHz chip would obviously be on the headlines of every technology publication, and this is exactly the kind of publicity Intel is trying to avoid right now. Why? Intel has clearly moved beyond the MHz marketing strategy, towards a more multi-tasking, application-oriented performance profile. A 4 GHz chip would be seen as detractive to this strategy. Put another way, it would raise questions about Intel's commitment to the new design and marketing strategy the company has developed over the past two years or so. In short, it would be a marketing disaster, even if the announcement gave Intel some cheap publicity.

 

[Scali]

Conroe is not a P-M, or anything close.
It's a new 64-bit architecture, optimized for desktops.
Which is also the reason why there is no P-M for desktops now, and there won't be one.
P-M has some weak points, like the limited clockspeed and poor SSE/SSE2 and multitasking performance. It can be better than a P4 or an Athlon64 in some cases, if you overclock it, but it's still far behind in others. So the P-M cannot replace the P4... and Intel is probably not interested in having yet another desktop platform to support. Demand would probably not be that high anyway, since the biggest group of customers doesn't know what a P-M is. So Intel would have to market it. Why bother?

 

[xonik]

I can see the Pentium M marketed in the future for set-top boxes, or part of the "Media Center" spinoff going on. Certainly the Pentium M in its current state isn't ready for a desktop takeover, for the reasons you already stated.

 

[ScHpAnKy]

Conroe is not a P-M, or anything close.

Architectually, Conroe is very close to the P-M.

It's a new 64-bit architecture, optimized for desktops.

Of those optimizations is hyperthreading

Which is also the reason why there is no P-M for desktops now, and there won't be one.
P-M has some weak points, like the limited clockspeed and poor SSE/SSE2 and multitasking performance. It can be better than a P4 or an Athlon64 in some cases, if you overclock it, but it's still far behind in others.

So the P-M cannot replace the P4... and Intel is probably not interested in having yet another desktop platform to support.

Conroe is set to run LGA775 last time I checked.

Demand would probably not be that high anyway, since the biggest group of customers doesn't know what a P-M is. So Intel would have to market it. Why bother?

They grab the enthusiasts back, and the uninformed masses will already congregate to the "bigger name."

 

[DougLite]

A) Don't forget that Intel will also be designing a blade server platform around the new P-M
B) Also, remember that Intel has to keep the potential 4GHz chip within the TDP and ability of the retail cooler.
C) As others have said, a 4GHz chip is not the answer going forward, dual core is.

When the 65nm chips come out, you will probably see a modest clock speed bump. However, it will be overshadowed by an avalanche of new dual core offerings.

 

[Donnie27]

pump up the FSB from weak 400/533 to 800 or 1066, add dual channel ram support, and add SSE3 and that will offset MOST of the problem with thos streaming apps you're talking about.

Funny you should say that. Most of the SSE3 instructions have more to do with Hyperthreading and multithreading, so it's almost useless on current Pentium M. Intel needs to add a second core or Hyperthreading for SSE3 to be of anyvalue. Sorry, it's not a weak 400/533 LOL! The FSB was set at that speed to save power more than anything else. Streaming apps do better when optimized (that includes AMD and Intel btw) software and longer pipes or stages. Close knit Dual Core acts as more stages as well.

Make separate Pentium M ala Desktop, Mobile, Low Voltage and etc...

Donnie27

 

[Scali]

Architectually, Conroe is very close to the P-M.

Really? In what way?
Where do you get your information from?

Of those optimizations is hyperthreading

Funny, most people claim that HT is just a fix for the long pipeline of Netburst, and it has no use for short pipelines like the P-M or the Athlon64...

Conroe is set to run LGA775 last time I checked.

That's a socket. A platform consists of more than that. Chipsets, drivers, BIOS etc.
Also, I don't consider P-M and Conroe the same, remember?
Conroe will replace P4, so Intel still only has to support one desktop platform (well, two if you count Celeron separately).

 

[Scali]

Close knit Dual Core acts as more stages as well.

What do you mean?

 

[SacLANd]

A) Don't forget that Intel will also be designing a blade server platform around the new P-M
B) Also, remember that Intel has to keep the potential 4GHz chip within the TDP and ability of the retail cooler.
C) As others have said, a 4GHz chip is not the answer going forward, dual core is.

When the 65nm chips come out, you will probably see a modest clock speed bump. However, it will be overshadowed by an avalanche of new dual core offerings.

A: thats sossaman. insance CPU based off of Yonah. power usage is really low, but its DCDP. Its not quite as nice as what you will see the Dempsey platforms on, but it is a VERY powerful machine that could easily fit in a 1U server with little to no heat concerns.

Conroe is based off the P-M architecture. but like I said, its on steroids. Merom is its mobile counterpart. (actually, thats a little backwards, since conroe is based off of merom) they have a couple others too but its a secret.

 

[USMC2Hard4U]

I am not really intrested in Netburst anymore. So to me, 4Ghz P4 would me Meh.

What I am intrested in is to see a Dual Core 2.5Ghz Per Core "Pentium M" or "Yonah" etc..

That would make me happy.

 

[Donnie27]

What do you mean?

Like IBM's power processors where two cores share one cache and work as if they're one processor.

Donnie27

 

[Scali]

Like IBM's power processors where two cores share one cache and work as if they're one processor.

No, they work as if they're two processors, sharing a cache
Working as one processor would mean that they can only run one thread at a time. But they can run two threads at a time (or more, with SMT). It's expansion in the width, not in the length of the processor. More instructions at the same time, but the instructions don't go any faster.

 

[HOCP4ME]

4ghz, dual core, hyperthreading, 4mb cache, 1066FSB...

mmmmm...

But I'm sure it'll never happen.

How about all of that on a Venice? Man, that would crush anything Intel could ever make for the next 5 years at least!

Anyway, I think that Intel should have gone ahead with the 4Ghz chip not because it would be noticably faster, but because they could use "first to 4Ghz" as a good marketing slogan. Alienware sells their watercooled ALX system with Intel chips overclocked to 4Ghz. With the watercooling they could probably go much farther, but since 4Ghz is like a milestone they'll know they'll only sell more if they got it to 5Ghz, which would be difficult.

 

[chrisf6969]

Anyway, I think that Intel should have gone ahead with the 4Ghz chip not because it would be noticably faster, but because they could use "first to 4Ghz" as a good marketing slogan.

Yep, my point.

 

[Donnie27]

No, they work as if they're two processors, sharing a cache
Working as one processor would mean that they can only run one thread at a time. But they can run two threads at a time (or more, with SMT). It's expansion in the width, not in the length of the processor. More instructions at the same time, but the instructions don't go any faster.

Not really! Hyperthreading doesn't need more than one core. They work as one core but on separate parts of the same thread. That's two working like one IMHO. The end result is having twice the CPU parts, ALU's, FPU's Pipes/Stages, and etc.... From reading 2CPU a while back, it's similar to SLI.

Donnie27

 

[Scali]

How about all of that on a Venice? Man, that would crush anything Intel could ever make for the next 5 years at least!

Because that's technically impossible. The P4 design can reach 4 GHz and more. Venice will be lucky to reach 3 GHz.

 

[Scali]

Not really! Hyperthreading doesn't need more than one core.

What are you talking about? I never said it did!

They work as one core but on separate parts of the same thread. That's two working like one IMHO. The end result is having twice the CPU parts, ALU's, FPU's Pipes/Stages, and etc....

Are you sure? That doesn't seem to make any sense. A single out-of-order processor already works on separate parts of the same thread. You don't need two cores for that.
And if you have two cores, you can already split the workload into two threads, so you decide what the cores will process, which will be more efficient.

 

[Donnie27]

What are you talking about? I never said it did!

I think you took that one out of context.

Are you sure? That doesn't seem to make any sense. A single out-of-order processor already works on separate parts of the same thread. You don't need two cores for that. And if you have two cores, you can already split the workload into two threads, so you decide what the cores will process, which will be more efficient.

Now back to what I said about Hyperthreading. How any processor works on one thread should depend on the app/s, not any modern processor. That's why teh P4 gets whacked in games and kicks ass running optimized Streaming app/s and *most multitasking jobs.

No, I'm not sure. It was about paralleled computing and making something already fast, faster. Instead of two parts, the workload threads could be broken down into 4 or even 8 parts out of order and without dropping cycles. We don’t need SLI or Dual Cores for most everyday uses either, right? Maybe I misunderstood it?

Donnie 27

 

[Scali]

I think you took that one out of context.

Well, what did you mean by it then?

Now back to what I said about Hyperthreading. How any processor works on one thread should depend on the app/s, not any modern processor. That's why teh P4 gets whacked in games and kicks ass running optimized Streaming app/s and *most multitasking jobs.

What are you saying... that apps should determine the performance, not the hardware?
The P4 is slow in games for 2 reasons:
1) It has deep pipelines and a high-latency memory subsystem, which make random memory accesses slow.
2) Games are written so poorly that they don't actually try to make their memory accesses less random, which means that P4 takes the full hit, while in more optimized applications, the latency is hidden.

That has nothing to do with HyperThreading, since in Windows XP, turning HT on or off has little or no effect for performance. It's the design of the CPU. And unless all CPUs are designed the same, you will always have different performance characteristics.

 

[Donnie27]

Well, what did you mean by it then?

What were we talking about? Streaming apps seem to like more stages=P Dual Core has the similar affect on streaming as a single processor with more stages. That doesn't matter if it is X2, 840 or Power. That's also why the Dothan gets its ass handed to it by even a 2.4C at stock speed with almost ANY Audio or Video app BTW. Intel marketed the P4 as the "Center Of Your Digital World". For once, they didn't lie. Now this whole "It Isn't Fair" crap is total BS as they talk about how great games run.

That are you saying... that apps should determine the performance, not the hardware? The P4 is slow in games for 2 reasons:
1) It has deep pipelines and a high-latency memory subsystem, which make random memory accesses slow.
2) Games are written so poorly that they don't actually try to make their memory accesses less random, which means that P4 takes the full hit, while in more optimized applications, the latency is hidden..

#1 Not really at all. The P4 Northwood has lower latency than the Hammer and still gets whacked. NONE of the currently tested games makes use of SSE2 and its 144 instructions. Many of these have more to do than just streaming as well. There's way too much BS being past off as Fact on this one. Games can make use of SSE2 and Incoming Forces Proved that almost 4 years ago.

#2, Yes Games are written poorly (when we're talking P4) but its much more than just memory access as Northwood running Linpack shows. Poorly Optimized Software causes the P4 to take hit on more than just Games. L.A.M.E. Hexus used the Intel patch and the results does a 180 and the P4 hand AMD it ass when they loose when not used. Even the WillyDog P4 ran Incoming Forces faster clock for clock than the P3 and the Athlon TBird and Palomino.

Also short data bursts works with many starts and stops are better suited for shorter pipeline processors. Sort of like the Sprinter vs. the Miler.

So you're damned right I say it depends on the software more than the hardware. SIMDs are more than just for Audio and Video encodes as many try to say here and other places.

That has nothing to do with HyperThreading, since in Windows XP, turning HT on or off has little or no effect for performance. It's the design of the CPU. And unless all CPUs are designed the same, you will always have different performance characteristics.

Sorry, but the way I use MY COMPUTER I can easily tell when HT is on or off. Even when HT doesn't help games, other things running don't slow the game down. The end result is I can get more done. I intend to run both an AMD and Intel processor and run Multiplicity to use the best of both worlds. This would still better than an expensive P4EE or X2.

Then there's the whole thing about a *Clean or *Dirty computers that P4s have done better with WInXP but almost no one talks about it. Clean Computers equals what some reviewer who just did a fresh install and defrag on. These computers are NOT on a network, NO printers, NO USB devices, very little or NO burning or other everyday use apps that may have TSRs and or other services. Dirty computers have all of those and more things to deal with. Even with WinXP, they might be using 130MB when just sitting there. Sorry but this is getting old.

Meanwhile: "Originally Posted by Scali
Demand would probably not be that high anyway, since the biggest group of customers doesn't know what a P-M is. So Intel would have to market it. Why bother?

They grab the enthusiasts back, and the uninformed masses will already congregate to the "bigger name."

They don't need to know what Pentium M is. But they do know what teh hell Centrino is and its performance and sales show just that. It's a killer app, kicks the crap out of Turion and folks are going back for seconds. Folks thinking Turion is great are the ones uninformed LOL!

Maybe it is the AMD folks who are uninformed? When asked by Newegg what many planned to do with their Dual Core AMD processors, the answer given by 80% was, "Game". How in the hell does that make them smart or informed?

Donnie27

 

[DejaWiz]

When the 65nm chips come out, you will probably see a modest clock speed bump. However, it will be overshadowed by an avalanche of new dual core offerings.

Well said Doug.

I think 4GHz is a moot point right now. With dual cores out now and quad cores on the horizon, clock speed will become less pertinent than it once was. The biggest boost in performance with come along as more and more applications and games are written with multi-core optimizations.

There will be a day when processors hit 4GHz, but it might have to wait with 65nm or smaller fabrication.

 

[Scali]

What were we talking about? Streaming apps seem to like more stages=P

Not really. It's more like more stages like streaming apps best, because they have linear memory access and little or no branches. So no 'hazards'. Which brings out the strong points of long pipelines: high processing speeds.
We cannot really compare 1:1 because there is no proper match for the P4... but back when we had P3 and P4 at comparable performance levels, SIMD performance on the P3 wasn't really affected by the shorter pipeline or lower clockspeed.

Dual Core has the similar affect on streaming as a single processor with more stages.

Not at all, dual core requires a completely different approach. You need to rewrite the algorithm. In some cases you can get a good win, in others you can't.
In general the 'ideal' application for a dualcore system is not the same as the 'ideal' application for a deeply pipelined processor.

That doesn't matter if it is X2, 840 or Power. That's also why the Dothan gets its ass handed to it by even a 2.4C at stock speed with almost ANY Audio or Video app BTW.

No, Dothan is slow in streaming apps because they cut down on multimedia resources in the architecture, in order to keep the power usage down. In theory you can get much more multimedia performance out of a processor based on the Dothan.

#1 Not really at all. The P4 Northwood has lower latency than the Hammer and still gets whacked.

Are you kidding me? Both the P4's longer pipeline and external memory controller make memory accesses far more expensive than on the Hammer.
The P4 has less latency with its L1-cache, but that is of no use in the situations I was discussing.

#2, Yes Games are written poorly (when we're talking P4) but its much more than just memory access as Northwood running Linpack shows. Poorly Optimized Software causes the P4 to take hit on more than just Games.

Of course. It's about the way the system is used by an application, not the fact that the application happens to be a game or not. Even among games there is a lot of variation between CPU load and performance.

[quote[So you're damned right I say it depends on the software more than the hardware. SIMDs are more than just for Audio and Video encodes as many try to say here and other places.[/quote]

Depends on how you look at it. From the viewpoint of the customer, SIMD is indeed used almost exclusively in audio/video software. From the viewpoint of the developer, SIMD can be applied in nearly any situation with intensive arithmetic (and will be, once Windows x64 applications arrive: there is no alternative).

Sorry, but the way I use MY COMPUTER I can easily tell when HT is on or off. Even when HT doesn't help games, other things running don't slow the game down. The end result is I can get more done.

Ofcourse, that's what HT is for. But the point is that HT does not affect the performance of the GAME you run. You seemed to associate HT and game performance, there is no relation.

 

[Donnie27]

Not really. It's more like more stages like streaming apps best, because they have linear memory access and little or no branches. So no 'hazards'. Which brings out the strong points of long pipelines: high processing speeds.
We cannot really compare 1:1 because there is no proper match for the P4... but back when we had P3 and P4 at comparable performance levels, SIMD performance on the P3 wasn't really affected by the shorter pipeline or lower clockspeed.

Streaming apps are streaming apps and don't change depending on the processor. They stay the same. They access the same bandwidth and even with higher latency to RAM, they still run faster on longer pipe line processors. Yes P4 pay a high price for misses and I not crazy enough to argue that point LOL! But these misses are way over rated when lower pricision instead of higher precision could have been used in the first place. Incoming Forces proves that point quite well.

Not at all, dual core requires a completely different approach. You need to rewrite the algorithm. In some cases you can get a good win, in others you can't.
In general the 'ideal' application for a dualcore system is not the same as the 'ideal' application for a deeply pipelined processor.

See MaxPC Premiere Pro 7 that has SSE2/3, Hyperthread and is miltithreaded?

No, Dothan is slow in streaming apps because they cut down on multimedia resources in the architecture, in order to keep the power usage down. In theory you can get much more multimedia performance out of a processor based on the Dothan.

That's one I'd have to see to believe, hehehehe! Same goes for the P3 before it. The same "Cutting Pony Legs" that make the processor stop a dime, it what kills it when it tries to run long distances. Get two pairs of short legs and it will do a lot better LOL! Draw your own analogy.

Are you kidding me? Both the P4's longer pipeline and external memory controller make memory accesses far more expensive than on the Hammer.
The P4 has less latency with its L1-cache, but that is of no use in the situations I was discussing.

But weren't talking about the L1 and double the core or do I have to explain everything? Of course I'm talking about within the core and not to main memory.

[/QUOTE]Of course. It's about the way the system is used by an application, not the fact that the application happens to be a game or not. Even among games there is a lot of variation between CPU load and performance.[/QUOTE]

True! But yet Games down't stream and streaming apps need very little interaction like games need.

[quote[So you're damned right I say it depends on the software more than the hardware. SIMDs are more than just for Audio and Video encodes as many try to say here and other places.[/quote]

Depends on how you look at it. From the viewpoint of the customer, SIMD is indeed used almost exclusively in audio/video software. From the viewpoint of the developer, SIMD can be applied in nearly any situation with intensive arithmetic (and will be, once Windows x64 applications arrive: there is no alternative).

I don't buy that for a minute. I don't know why others didn't bother to run Intel Launcher software when the P4's lauched, hell, ask Kyle? That software launched with Utilites, Games, and streaming apps all optimized for SS2. The software that made up the package was ON-THE-SHELF at the time. It's only as exclusive as the software makers want it to be. Same can be said for Webmasters who avoided this stuff like it was a Computer virus or something.

Ofcourse, that's what HT is for. But the point is that HT does not affect the performance of the GAME you run. You seemed to associate HT and game performance, there is no relation.

With WinXP, it does affect gaming but how can you tell on a clean system? I'm trying to a associate HT with SYSTEM PERFORMANCE not game perform. (Capps here denote importance not shouting). I said Power Users don't run games exclusively. That's something the AMD guys came up with to fight off how useful HT was and still is. When just running a Game the P4 is more than like running the game slower that Hammers. If playing game while you wait for video or audio job to finish, then yes, HT does affect performance. I never implied that HT alone affected games. As I pointed out, HT can help and hender.

http://www.gamepc.com/labs/view_content.asp?id=pee840&page=1

The only difference is that the Pentium-D will not have Hyper-Threading enabled. In our tests, we only found a scant few applications which benefited from Hyper-Threading, some even performed slower. In the cases where Hyper-Threading does help, it’s typically between a 5-10% performance gain. Is this feature enough to justify the massive price hike over the standard Pentium-D? We certainly don’t think so. We think Intel has made a very large mistake in relying on Hyper-Threading as their sold differentiator between their dual core product line. It’s simply not an important enough feature to warrant such a large price gap.

Of course the other feature is Intel left the 840EE multiplier unlocked. Still a dumb move

Donnie27

 

[USMC2Hard4U]

Well I think a 4Ghz Per Core - Dual Core Yonah would pretty much rox my sox off.

Time For a Vapo

 

[Scali]

They access the same bandwidth and even with higher latency to RAM, they still run faster on longer pipe line processors.

The way you say that is like the length of the pipeline has something to do with the application. That's what I was arguing against. Sure, on the PARTICULAR long-pipe processor that we have today, the P4, streaming apps run quickly. And on the PARTICULAR short-pipe processors that we have today, Athlon64 and Pentium-M, streaming is not that fast.
That doesn't mean that the long pipeline is the thing making the difference. No, the design of the P4 makes the difference. It is optimized for stream processing, and as a result, it has a long pipeline. Just wait for Conroe, and you'll see a processor with a shorter pipeline, which will be as good, if not better, at stream processing.

See MaxPC Premiere Pro 7 that has SSE2/3, Hyperthread and is miltithreaded?

But I'm talking about a theoretical ideal application, that gets the maximum possible performance out of the system. You're giving an application that has so many optimizations that it gets performance increase from various different technologies. That's not the same. It doesn't prove that a good SSE2 application automatically runs faster on a dual-core system, or that a good multithreaded application automatically runs faster on an SSE2 system. See the difference?

That's one I'd have to see to believe, hehehehe! Same goes for the P3 before it. The same "Cutting Pony Legs" that make the processor stop a dime, it what kills it when it tries to run long distances. Get two pairs of short legs and it will do a lot better LOL! Draw your own analogy.

I think you're getting lost in the analogies, and not trying to understand what is actually going on inside a streaming application and the processor it is running on.

But weren't talking about the L1 and double the core or do I have to explain everything? Of course I'm talking about within the core and not to main memory.

Well I specifically gave some examples that pointed out situations where the cache would no longer be effective. Obviously as long as the L1 cache is in play, the P4 is the fastest processor in the world, I thought we wouldn't even have to debate that, since it's so obvious. I wanted to go into the cases where things go wrong with the P4, and why, and how to improve on that.

True! But yet Games down't stream and streaming apps need very little interaction like games need.

That doesn't mean that you can't optimize games to run well on a P4 processor. That's what I mean.

With WinXP, it does affect gaming but how can you tell on a clean system? I'm trying to a associate HT with SYSTEM PERFORMANCE not game perform. (Capps here denote importance not shouting).

But then you're arguing something else. Ofcourse HT improves multitasking. That's the whole point. Same with dualcore. But in both cases a single application will not run faster, unless it is rewritten to take advantage of it.
And that's the difference with a pipeline. All instructions must pass through the pipeline, so any changes to that pipeline will immediately affect performance. It is 'transparent'.
So I didn't understand the way you were trying to compare the two.