Does a P4 really run multi apps better than AMD?

[Capt Caveman]

I think it's b/c people don't want to spend an evening encoding a video and not being able to do anything else with their pc, a dual-core allows you to do other things. With a X2, I can run more background apps without having to go into task manager and shutting some of them down to gain performance. I can do dumb things like run adware or do a virus scan without having to walk away from the pc but can watch a video that I recorded on my pvr. You gain a lot more performance/efficiency if you're a power-user or heavy multi-tasker. Not everyone needs it. I love it.

 

[Viditor]

The biggest problem in multitasking is the ineficiency of the Windows scheduler...the question is what the Longhorn (sorry, Vista) version will do...

 

[clarkey01]

Originally posted by: Viditor
The biggest problem in multitasking is the ineficiency of the Windows scheduler...the question is what the Longhorn (sorry, Vista) version will do...

Has HT got long left ?

With dual cores and the lack of HT (up to now) on the Intel models (Cept the EE) is there really a need for it ? Im not sure if Yonah & future generations which use shorter pipelines will use a form of HT.

I thought HT would play a big part in the future, looks like shorter pipelines due to higher IPC desired by both sides its not going to have such a big impact.

 

[BitByBit]

Neos:
I wasn't implying HT gives the P4 an unfair advantage.
I was merely stating that without HT, the P4 would have no multitasking advantage over the Athlon. I've seen many posts that claim or imply that the P4 is just a better multitasker, with or without HT, which isn't true.

Zebo:
I'd imagine the idea that HT only benefits deep pipelines has arisen from the fact that deeper pipelines take longer to flush and refill than shorter pipelines.
While this is true, context switching latency is actually more dependent on memory interface speed than a processor's instruction latency. A 3.0GHz Prescott's instruction latency is around 10ns, while its memory access latency is many times that.
What HT allows is two threads to run concurrently without having to continually switch between them, which would involve memory access.
The second benefit to HT is when running multithreaded applications.
All modern processor's are superscalar. Superscalar execution allows a processor to execute more than one instruction simultaneously, but only if the pick stage(s) can find non-dependent instructions to execute in parallel. If this isn't the case, then there are going to be redundant execution units.
HT simply allows the processor to pick from more instructions, so that the probability of not being able to find two or more instructions to execute in parallel has been reduced, which has resulted in the P4 getting up to a 20% performance boost when running multithreaded applications.
'Wider' designs, such as the Athlon, would benefit from SMT atleast as much as the P4 has. In order to utilise its execution units efficiently, the Athlon must find more instructions to execute in parallel. With SMT, it would have more instructions to pick from, and therefore would be able to sustain a higher execution rate.

As far as I know, Intel intends to enable HT on Conroe, which is supposedly going to be a 4-issue design (wider than the Athlon).
That makes sense, because without HT, it would have a hard time keeping those execution units busy.
I doubt Dothan/Yonah will ever make use of HT. Despite Dothan's high IPC, it is in fact a 'narrow', but already highly efficient design.

 

[Keysplayr]

Originally posted by: JasonE4
The P4 2.4C is the processor with the lowest frequency that supports HT. All 800 MHz fsb P4s support HT, while the P4 3.06 533 fsb is the only proc other than the those that supports HT.

BINGO

 

[Lord Banshee]

Originally posted by: stevty2889

Originally posted by: munky
I'm using a p4 prescott at work, and an a64 at home. From everyday use of multi-tasking, I personally cant tell the difference - my a64 never seems bogged down, even with multiple apps open. The p4, as far as I can tell, has no apparent advantages over the a64. However, I'm using win2k at work, so does HT work under win2k or only in win xp? When I open task manager, I can only see one processor, and everywhere I look, it seems to show up as a single processor. Does win2k not support HT or not?

Windows 2000 doesn't support hyperthreading.

Seems to work for me?? P4 2.8C at work using win2k. Multithreaded apps use 100% CPU ( 1 real core 1 HT) usage as single threaded uses 50% (1 real core no HT)

 

[AntiStatic]

Security flaw in Intel Hyper-Threading CPUs

Colin Percival, a FreeBSD committer and security team member, has found a local exploit against the current implementation of Intel's Hyper-Threading Technology. "Hyper-Threading, as currently implemented on Intel Pentium Extreme Edition, Pentium 4, Mobile Pentium 4, and Xeon processors, suffers from a serious security flaw," Colin explains. "This flaw permits local information disclosure, including allowing an unprivileged user to steal an RSA private key being used on the same machine. Administrators of multi-user systems are strongly advised to take action to disable Hyper-Threading immediately."


not so good.

 

[angryswede]

Originally posted by: BitByBit
Neos:
I wasn't implying HT gives the P4 an unfair advantage.
I was merely stating that without HT, the P4 would have no multitasking advantage over the Athlon. I've seen many posts that claim or imply that the P4 is just a better multitasker, with or without HT, which isn't true.

Zebo:
I'd imagine the idea that HT only benefits deep pipelines has arisen from the fact that deeper pipelines take longer to flush and refill than shorter pipelines.
While this is true, context switching latency is actually more dependent on memory interface speed than a processor's instruction latency. A 3.0GHz Prescott's instruction latency is around 10ns, while its memory access latency is many times that.
What HT allows is two threads to run concurrently without having to continually switch between them, which would involve memory access.
The second benefit to HT is when running multithreaded applications.
All modern processor's are superscalar. Superscalar execution allows a processor to execute more than one instruction simultaneously, but only if the pick stage(s) can find non-dependent instructions to execute in parallel. If this isn't the case, then there are going to be redundant execution units.
HT simply allows the processor to pick from more instructions, so that the probability of not being able to find two or more instructions to execute in parallel has been reduced, which has resulted in the P4 getting up to a 20% performance boost when running multithreaded applications.
'Wider' designs, such as the Athlon, would benefit from SMT atleast as much as the P4 has. In order to utilise its execution units efficiently, the Athlon must find more instructions to execute in parallel. With SMT, it would have more instructions to pick from, and therefore would be able to sustain a higher execution rate.

As far as I know, Intel intends to enable HT on Conroe, which is supposedly going to be a 4-issue design (wider than the Athlon).
That makes sense, because without HT, it would have a hard time keeping those execution units busy.
I doubt Dothan/Yonah will ever make use of HT. Despite Dothan's high IPC, it is in fact a 'narrow', but already highly efficient design.

when people say the processor is wider, do they mean the number of execution units? and when they say its longer that refers to pipelines stages right?

 

[Acanthus]

Originally posted by: Zebo
Hmm.. seems to me at every price point Intel is more money.

Look at that $170 3400 beating $600 660 on the first test and last, tying $250 640 most of the time.

Your memory is still there😉

Actually if you look at my experience im not seeing the price advantage of AMD, i have to use much more expensive memory to get the aggressive O/C i was on a P4 with corsair value select. (this is mainly because i chose the DFI Ultra-D for its stellar overclocking reputation).

The CVS doesnt seem to like any memory dividers on the Ultra-D except for the 333, so im stuck at 250htt.

Which would roughly peg me at the exact same performance as the rig in my sig. Only the motherboard costs significantly more for AMD, and the CPU is more expensive for Intel.

To actually get the ~2.7ghz overclock i was expecting, i would have to move to OCZ memory at a significant cost increase.

Im not saying i cant afford it, or wont do it, Im just saying price isnt as clear cut with AMD as some imply.

 

[Acanthus]

Originally posted by: Lord Banshee

Originally posted by: stevty2889

Originally posted by: munky
I'm using a p4 prescott at work, and an a64 at home. From everyday use of multi-tasking, I personally cant tell the difference - my a64 never seems bogged down, even with multiple apps open. The p4, as far as I can tell, has no apparent advantages over the a64. However, I'm using win2k at work, so does HT work under win2k or only in win xp? When I open task manager, I can only see one processor, and everywhere I look, it seems to show up as a single processor. Does win2k not support HT or not?

Windows 2000 doesn't support hyperthreading.

Seems to work for me?? P4 2.8C at work using win2k. Multithreaded apps use 100% CPU ( 1 real core 1 HT) usage as single threaded uses 50% (1 real core no HT)

Its just not officially supported, it does in fact work, ive been using HT on 2k for over a year.

 

[imported_nocturne]

bla bla bla bla...

I only care about performance when I need it. No matter which apps I have open (besides games) my 1ghz tbird handles perfectly good enough. When you're talking a jump of a few years ahead in technology, what little latency there is doesn't matter much unless you're actually using your cpu to its potential.

just my $.02

 

[stevty2889]

Originally posted by: Lord Banshee

Originally posted by: stevty2889

Originally posted by: munky
I'm using a p4 prescott at work, and an a64 at home. From everyday use of multi-tasking, I personally cant tell the difference - my a64 never seems bogged down, even with multiple apps open. The p4, as far as I can tell, has no apparent advantages over the a64. However, I'm using win2k at work, so does HT work under win2k or only in win xp? When I open task manager, I can only see one processor, and everywhere I look, it seems to show up as a single processor. Does win2k not support HT or not?

Windows 2000 doesn't support hyperthreading.

Seems to work for me?? P4 2.8C at work using win2k. Multithreaded apps use 100% CPU ( 1 real core 1 HT) usage as single threaded uses 50% (1 real core no HT)

The problem with win2k is that it recognizes the HT as an actual second processor, so it doesn't handle it quite the same, which can cause it to use it inefficiently or cause some strange glitches at times.

 

[stevty2889]

Originally posted by: clarkey01

Originally posted by: Viditor
The biggest problem in multitasking is the ineficiency of the Windows scheduler...the question is what the Longhorn (sorry, Vista) version will do...

Has HT got long left ?

With dual cores and the lack of HT (up to now) on the Intel models (Cept the EE) is there really a need for it ? Im not sure if Yonah & future generations which use shorter pipelines will use a form of HT.

I thought HT would play a big part in the future, looks like shorter pipelines due to higher IPC desired by both sides its not going to have such a big impact.

I think HT will die with netburst. It would appear that the dualcore xeons MP's will still have HT.

 

[anandtechrocks]

I've got a 3700+ SD and I actually do notice a difference between it and my parents socket 478 3.2 Ghz P4 w/HT as far as multitasking goes.

A recent example: I was defragmenting my A64 and decided to do my parents computer at the same time. The Athlon lagged noticably during the defrag while I tried to do other things like surf the web. However, the P4 stayed very smooth and I did not notice any drop in performance. Just something I noticed.

 

[Hacp]

Originally posted by: Acanthus

Originally posted by: Zebo
Hmm.. seems to me at every price point Intel is more money.

Look at that $170 3400 beating $600 660 on the first test and last, tying $250 640 most of the time.

Your memory is still there😉

Actually if you look at my experience im not seeing the price advantage of AMD, i have to use much more expensive memory to get the aggressive O/C i was on a P4 with corsair value select. (this is mainly because i chose the DFI Ultra-D for its stellar overclocking reputation).

The CVS doesnt seem to like any memory dividers on the Ultra-D except for the 333, so im stuck at 250htt.

Which would roughly peg me at the exact same performance as the rig in my sig. Only the motherboard costs significantly more for AMD, and the CPU is more expensive for Intel.

To actually get the ~2.7ghz overclock i was expecting, i would have to move to OCZ memory at a significant cost increase.

Im not saying i cant afford it, or wont do it, Im just saying price isnt as clear cut with AMD as some imply.

Thats cause your using the DFI Lan Party board. If you use many other boards, the ram divider will be fine. The DFI board just sucks with value ram. The price advantage is there.

 

[BillyBobJoel71]

i think that intel single cpus are better at it then amd cpus, but the dual cores i don't know. considering they have hyper threading.

 

[Hacp]

Hyperthreading is currently bugged with the Intel EE dual core Cpus I think. This is because when you have two independant threads running simotaneously, Windows will use one core and one "virtual" core instead of two cores.

Also, does anyone know where you can get the 2.4C retail or has Intel stopped making them?

 

[stevty2889]

Originally posted by: Hacp
Also, does anyone know where you can get the 2.4C retail or has Intel stopped making them?

Wow, I actualy couldn't find one. They stopped producing northwoods quite some time ago, but I know last month at least they still had some 2.4c's on newegg, can't seem to find any now though.

 

[lordsaytor]

Read Anandtech's article on the 4200+ dual core cpu

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2452

I'm basically regurgitating what Anandtech says. And I'm just repeating the obvious.

The dual core AMD cpu isn't really going to benefit your son much unless he's doing media encoding. And I don't know if he intends to upgrade or what. Now it depends on what your son's cpu is. If it's an Athlon64 4000+ then the price difference between that and the 4200+ (the cheapest dual core) is like fifty bucks. If it's a 3500+ now, then that difference jumped to $300.

Again, your son probably has a really capable computer already. He really ought to ask himself what he's comparing to. If everyone in his office has computers made by Falcon Northwest or Alienware, then what's the point?

Now this is just my opinion. But I find that the point of having a really fast PC is for games. If you were really into digital art and media editing, you'd probably be buying a PowerMac in the first place.

Yeah I'm pretty ignorant at times, forgive me.

 

[BitByBit]

Originally posted by: angryswede
when people say the processor is wider, do they mean the number of execution units? and when they say its longer that refers to pipelines stages right?

Correct.
It is execution core width that directly determines IPC, and not pipeline depth.

 

[coomar]

i've had no problem with multitasking on an a64 single core, alt-tabing out of games is near instanteous as is going back in (a second or two)

 

[Gamingphreek]

Originally posted by: BitByBit
Neos:
I wasn't implying HT gives the P4 an unfair advantage.
I was merely stating that without HT, the P4 would have no multitasking advantage over the Athlon. I've seen many posts that claim or imply that the P4 is just a better multitasker, with or without HT, which isn't true.

Zebo:
I'd imagine the idea that HT only benefits deep pipelines has arisen from the fact that deeper pipelines take longer to flush and refill than shorter pipelines.
While this is true, context switching latency is actually more dependent on memory interface speed than a processor's instruction latency. A 3.0GHz Prescott's instruction latency is around 10ns, while its memory access latency is many times that.
What HT allows is two threads to run concurrently without having to continually switch between them, which would involve memory access.
The second benefit to HT is when running multithreaded applications.
All modern processor's are superscalar. Superscalar execution allows a processor to execute more than one instruction simultaneously, but only if the pick stage(s) can find non-dependent instructions to execute in parallel. If this isn't the case, then there are going to be redundant execution units.
HT simply allows the processor to pick from more instructions, so that the probability of not being able to find two or more instructions to execute in parallel has been reduced, which has resulted in the P4 getting up to a 20% performance boost when running multithreaded applications.
'Wider' designs, such as the Athlon, would benefit from SMT atleast as much as the P4 has. In order to utilise its execution units efficiently, the Athlon must find more instructions to execute in parallel. With SMT, it would have more instructions to pick from, and therefore would be able to sustain a higher execution rate.

As far as I know, Intel intends to enable HT on Conroe, which is supposedly going to be a 4-issue design (wider than the Athlon).
That makes sense, because without HT, it would have a hard time keeping those execution units busy.
I doubt Dothan/Yonah will ever make use of HT. Despite Dothan's high IPC, it is in fact a 'narrow', but already highly efficient design.

Ummm im not sure about that.

Hyperthreading, as some of you would describe it, is not a miracle technology. It is merely a VERY effective way at covering up a processors weakness.

Hyperthreading, merely sends another packet of data through the instruction pipeline while the first packet is already in there, instead of waiting for the packet to finish fully. Additionally this also helps in the event of a cache miss or something similiar because it would hault the entire stream if the packet had to go back through. With HT, you already have the data on its way through.

Additionally, the A64's are not bad at multi-tasking. The P4's only hold an advantage in heavy multitasking. Having a web browser open and playing a game, there will be no difference. However, as it was stated earlier, playing a game, encoding some video, and surfing the web at the same time is heavy and the P4 handily beats the A64. In all fairness however, while the P4 does do MUCH better in that scenario, it still doesn't give the user playable framerates or anything, you would need a Dual Proc/Core system to make that happen.

-Kevin

 

[BitByBit]

Originally posted by: Gamingphreek

Originally posted by: BitByBit
Neos:
I wasn't implying HT gives the P4 an unfair advantage.
I was merely stating that without HT, the P4 would have no multitasking advantage over the Athlon. I've seen many posts that claim or imply that the P4 is just a better multitasker, with or without HT, which isn't true.

Zebo:
I'd imagine the idea that HT only benefits deep pipelines has arisen from the fact that deeper pipelines take longer to flush and refill than shorter pipelines.
While this is true, context switching latency is actually more dependent on memory interface speed than a processor's instruction latency. A 3.0GHz Prescott's instruction latency is around 10ns, while its memory access latency is many times that.
What HT allows is two threads to run concurrently without having to continually switch between them, which would involve memory access.
The second benefit to HT is when running multithreaded applications.
All modern processor's are superscalar. Superscalar execution allows a processor to execute more than one instruction simultaneously, but only if the pick stage(s) can find non-dependent instructions to execute in parallel. If this isn't the case, then there are going to be redundant execution units.
HT simply allows the processor to pick from more instructions, so that the probability of not being able to find two or more instructions to execute in parallel has been reduced, which has resulted in the P4 getting up to a 20% performance boost when running multithreaded applications.
'Wider' designs, such as the Athlon, would benefit from SMT atleast as much as the P4 has. In order to utilise its execution units efficiently, the Athlon must find more instructions to execute in parallel. With SMT, it would have more instructions to pick from, and therefore would be able to sustain a higher execution rate.

As far as I know, Intel intends to enable HT on Conroe, which is supposedly going to be a 4-issue design (wider than the Athlon).
That makes sense, because without HT, it would have a hard time keeping those execution units busy.
I doubt Dothan/Yonah will ever make use of HT. Despite Dothan's high IPC, it is in fact a 'narrow', but already highly efficient design.

Ummm im not sure about that.

Not sure about what exactly?

Hyperthreading, as some of you would describe it, is not a miracle technology. It is merely a VERY effective way at covering up a processors weakness.

It has nothing to do with covering up any weakness. I've already outlined what it does. You seem to be under the very common misconception that the only benefit of HT is to mask the impact of a deep pipeline.

Hyperthreading, merely sends another packet of data through the instruction pipeline while the first packet is already in there, instead of waiting for the packet to finish fully. Additionally this also helps in the event of a cache miss or something similiar because it would hault the entire stream if the packet had to go back through. With HT, you already have the data on its way through.

As I've already stated, the biggest problem when context switching is fetching the next thread from main memory. What Hyperthreading does is allow two threads to run concurrently without having to continually swtich between them, which would involve lots of memory access. You have only reiterated what I've said already.

Additionally, the A64's are not bad at multi-tasking. The P4's only hold an advantage in heavy multitasking. Having a web browser open and playing a game, there will be no difference. However, as it was stated earlier, playing a game, encoding some video, and surfing the web at the same time is heavy and the P4 handily beats the A64. In all fairness however, while the P4 does do MUCH better in that scenario, it still doesn't give the user playable framerates or anything, you would need a Dual Proc/Core system to make that happen.

And that proves my earlier point that many seem to think the P4 is inherently a better multitasker than the Athlon. You made no reference to Hyperthreading.

With respect, I think there are quite a few things you need to read up on before you can really contribute to this subject.
Try reading my posts fully before you try to discredit them.

 

[Gamingphreek]

Not sure about what exactly?

If you went on to read, you would see that i wasn't sure about some of your comments.

It has nothing to do with covering up any weakness. I've already outlined what it does. You seem to be under the very common misconception that the only benefit of HT is to mask the impact of a deep pipeline.

Yes it does. A Long Instruction Pipelines weakness is Cache Misses, and Multi-tasking among other things. HT is merely a solution, or a VERY VERY effective mask for this. YOu said it yourself... without it the P4 sucks at multi-tasking.

And that proves my earlier point that many seem to think the P4 is inherently a better multitasker than the Athlon. You made no reference to Hyperthreading.

It is not inherently a better multi-tasker. The A64 would wipe the floor with the Pentiums in multi-tasking without hyperthreading. The HT is an extra feature, it does not come with a long pipeline architecture.

With respect, I think there are quite a few things you need to read up on before you can really contribute to this subject.
Try reading my posts fully before you try to discredit them.

Why in the world would i have to read up. You just agreed with almost every thing i said. Furthermore you said that i merely reiterated what you said (which i didn't). It would seem to me that you need some basic comprehension skills.

I NEVER discredited your post. I never said you were wrong. I said im not sure if you are right, which means I AM NOT POSITIVE. It also means that i am not sure i am right about everything either (in this case i was). Once again, reading comprehension.

-Kevin

 

[BitByBit]

Originally posted by: Gamingphreek

It has nothing to do with covering up any weakness. I've already outlined what it does. You seem to be under the very common misconception that the only benefit of HT is to mask the impact of a deep pipeline.

Yes it does. A Long Instruction Pipelines weakness is Cache Misses, and Multi-tasking among other things. HT is merely a solution, or a VERY VERY effective mask for this. YOu said it yourself... without it the P4 sucks at multi-tasking.

No!
Multitasking performance on non-SMT processors is determined by memory interface speed!
What HT allows is two threads to run in parallel without having to continually swap between them and without having to continually access main memory.
A 3.0GHz Prescott's instruction latency is 10ns. The latency of the memory subsystem is around 80ns, I believe. So you can see that the biggest performance impact when context switching is accessing memory, and not filling the pipeline.

Secondly, HT does not mask the impact of pipeline stalls or flushes for a particular single-threaded application. Think about it: if you're gaming, and the branch predictor mispredicts a branch in the game's code, how is being able to run an unrelated thread going to help? The benefit in terms of masking branch mispredictions is only apparent when running multithreaded code.

I've already stated the other benefit of HT previously in this thread.