Which apps are actually slowed by hyper-threading?

[Ken90630]

Hey, Everyone. Just signed up, and this is my first post to Anandtech.com.

I'm planning on upgrading my motherboard & CPU and am trying to decide whether or not to get a P4 with hyper-threading or without it. The only apps I will be running on this particular computer will be Microsoft Office (mostly Word and PowerPoint) and the Adobe Creative Suite (mostly InDesign and Photoshop). I won't be using the Internet or e-mail on this computer. Other than Photoshop CS, I don't see any of these apps making notable use of hyper-threading. Further, I have no intentions of doing any other multi-tasking while these programs are running (at least not often enough to matter).

So as for Photoshop: According to Intel's Web site, they worked with Adobe to optimize Photoshop CS for P4 chips with hyper-threading. In particular, I think they say there are about 20 or so filters specifically architected to utilize the dual-threading capabilities of the P4 with HT. Sounds great.

My question is this: I've read that some software apps actually take a hit (run slower) when run on a P4 with hyper-threading. When I called Intel and asked which ones, I couldn't get a straight answer (or any answer, actually!). Rather than spend hours scouring the Web for some lab's benchmark results on this, I thought I'd ask if any of you know which software programs out there in computerland actually take a hit with a hyper-threading-enabled P4. (And if there's a reasonably recent [and truly reliable] test any of you can point me towards, that's available online, that'd be great.)

Several test reports and commentaries I've read on hyper-threading mention things like, "some applications run 5-10% slower with hyper-threading enabled, so whether or not HT is for you depends on the applications you'll be running." Well, okay, that's great to know, but that type of statement is useless unless we know WHICH apps take a hit, right?!

I'm basically trying to determine whether or not I'll take a slight performance hit with my MS Office apps, InDesign, Illustrator and Acrobat in exchange for some genuine speed improvement with specific filters in Photoshop CS if I get a P4 with HT.

Ken

PS: I'm also curious about whether the particular optimized filters in Photoshop CS are ones that a typical Photoshop user uses a lot, so if any of you have an opinion on that, feel free to share it with the class as well.

 

[Shimmishim]

PM Duvie or wait for him to respond... he should be able to help you out

 

[Soulkeeper]

interesting

i would suggest that atleast 90% of programs that aren't optimized for multi-threading
would either do the same or slightly worse

the ones that are optimized for dual proc systems or any kinda mult-thread capable system should benefit greatly from the feature

this is why i am so excited about getting a 2.8e to replace my 1.8 P4m northwood
hyperthreading and sse3 can't be bad if intel spent so much money developing them right

generally speaking most apps won't hurt you for not having hyperthreading
but atleast those that are mult threaded will benefit (perhaps more than 15% increases in some cases)

 

[myocardia]

Originally posted by: Shimmishim
PM Duvie or wait for him to respond... he should be able to help you out

No doubt, Duvie knows three times as much about hyperthreading as all of the review sites put together.

 

[Wingznut]

Originally posted by: Ken90630
Several test reports and commentaries I've read on hyper-threading mention things like, "some applications run 5-10% slower with hyper-threading enabled, so whether or not HT is for you depends on the applications you'll be running." Well, okay, that's great to know, but that type of statement is useless unless we know WHICH apps take a hit, right?!

I'm not aware of any apps that are "5-10% slower" with HyperThreading... I don't suppose you happen to have any links to the articles that published such comments?

 

[mikela78]

I expect you are using Windows XP, as you seem to have a taste for recent versions of software (Photoshop CS). In case you are using Windows 2000 as I am, you should know that in most cases enabling hyperthreading slows down the application. This link (Babelfish translated English version) gives some actual benchmarks comparing win2000 vs. winXP with and without hyperthreading. In Windows XP, WinRAR and a few games did very slightly worse with hyperthreading than without, but there are admittedly few application benchmarks in this article.

In case you are using win2000, I have found that rendering (Houdini 6.1) and computational fluid dynamics (Realflow 2.5) are both slower with hyperthreading, combining the results of two threads.

And in case you use Linux 2.4, I have found no difference in rendering times with and without hyperthreading. Kernel 2.6 might be a different matter...I hope to get it all working tomorrow!

 

[mikela78]

rereading your post, I don't think you have to worry about Office, Illustrator, or Acrobat. I've used all three a fair bit, and even if they slow down a little with hyperthreading under XP (which I doubt) I don't think they'd feel sluggish provided you have a fast processor. They're just not processor-intensive like Photoshop can be.

Speeding up Photoshop is not so straightforward, though. I remember reading on the Adobe forums that people were experiencing a slowdown with CS when upgrading from 1 to 2GB of memory, of all things. Also, in many cases PS7 is faster than CS. Perhaps the most effective way to speed up Photoshop is to get a seperate, high-speed hard drive for exclusive use as a Photoshop scratch disk. More info on the Adobe PS forum...

 

[Duvie]

I think the 5-10% and the claims of apps running slower was early on when HT first came on the scene with the 3.06ghz P4....As mentioned above some of the testing was done on windows 2k which is not officially an OS that supports HT...therefore as soulkeeper has stated, and I have tested over 30 some apps and have yet to find one that slows down more then say 1% or in the neglible testing range.

Fact is more and more apps are becoming optimised for HT and the ones you listed including excel and word (office components) if the newer versions, are actually supposed to have some features that also can take advantage of HT. I have yet to confirm those, as I do not have office for XP and use MS works suite for my spreadsheet and word processor needs.

Since I have been testing, I can think of half a dozen apps that have added HT features to them that I use personally. This number will surely increase.

Take a good look at newer reviews of HT from the last 6 months to present, and you are not likely to see mentioning of decreases in performance like that.....Only time you hear this now is by the amd fanboy clan who like to spread FUD....

ONe thing to keep in mind and something I have noticed is that priority settings seem to be quite different. For example what I mean is sometimes apps that are low priority on non HT systems (meaning they steal no cpu cycles from main applications or application with normal or higher thread priority) are not quite as low priority on HT systems. It is conceivable if you have some powerful background apps running concurrently with an application it could slow that application down. I would like to think review sites would have taken this into account, but who knows....I have isolated this in my testing and like I said I never saw one app I tested slow down. 0 out of about 30??? I think that is pretty good odds....

Gaming currently has no HT optimization that I am aware of and therefore the games I tested where exactly the same whether HT was on or off.

 

[Duvie]

Originally posted by: mikela78
I expect you are using Windows XP, as you seem to have a taste for recent versions of software (Photoshop CS). In case you are using Windows 2000 as I am, you should know that in most cases enabling hyperthreading slows down the application. This link (Babelfish translated English version) gives some actual benchmarks comparing win2000 vs. winXP with and without hyperthreading. In Windows XP, WinRAR and a few games did very slightly worse with hyperthreading than without, but there are admittedly few application benchmarks in this article.

In case you are using win2000, I have found that rendering (Houdini 6.1) and computational fluid dynamics (Realflow 2.5) are both slower with hyperthreading, combining the results of two threads.

And in case you use Linux 2.4, I have found no difference in rendering times with and without hyperthreading. Kernel 2.6 might be a different matter...I hope to get it all working tomorrow!

I actually have tested winrar on Winxp and I saw no such slowdown so I will need to check out this review site and their test setup...Winrar v2.90 which I tested is not HT optimized. I also tested winzip and didn't see a decrease either....



In theory if you look at HT I don't see how it actually should slow down any app, unless as mentioned it is being used on an OS software that does not support HT...


Update: You need to read your own link...Winrar winxp with HT was eaxctly the same time as without as well as a another similar type of app winace...It was win2000 that did better then even winxp, but it is clearly the signs of a more mature OS platform.....


Read the article it pretty much states what I have said....

Also notice the application that winxp w/ HT actually did worse then w/o HT....Futuremock pcmark2002...crap synthetic benchmarks...why any site would test these is beyond me...

 

[mikela78]

hmm. No sense editing out my mistake now! Looks like I mixed up the XP-no-hyperthreading and Win2K-no-hyperthreading results...

Anyway, it's good to know you haven't seen any significant slowdown with HT under XP. I hope Linux will be as good to me...

 

[pspada]

I can't see that the 2.6 kernel would cause any slowdown when utilizing HT - Linus and company are not stupid people, if anything it will take advantage of HT.

 

[Soulkeeper]

yeah linux runs faster with HT enabled
apache, db servers, and even compiling the kernel were faster with both 2.4 and 2.6 if i remember correctly from a review i read

 

[Ken90630]

Hey, All:

Sorry it's taken me so long to jump back in. I've been unusually busy the last few weeks.

Thanks to y'all for your thoughts on this subject. To answer Wingznut's question if I have a link to any articles mentioning the "5-10% slower" results w/HT, unfortunately I don't. I actually trotted out a stack of my computer magazines today to try & find that reference, 'cuz I know that's what I read somewhere, but I couldn't find it. As I'm typing this, I'm thinking I might have read it in one of those Australian computer magazines that I stood and read at a newsstand awhile back and then put back without buying .... Either that or maybe I read it on any one of a zillion computer Web sites and didn't happen to print the article out. Anyway, that "5-10% slower" phrase stuck in my mind 'cuz it sounded kinda significant (and a bit improbable). If I do run across it again, I'll post a link (or at least a direct reference) here on the forums and restore what's left of my now likely-tarnished credibility.

Moving right along: What I did come across today is the following, from Maximum PC magazine's "Build the Perfect PC" special issue from fall 2003:

On page 32, Section 9, it states, "Although there are times when Hyper-Threading can hinder performance, for the most part it's pretty useful stuff." And on page 29 of the same issue, the author states, "There are occasions when the CPU can run slower because two floating-point-heavy applications are trying to vie for the same physical resources, but we've mostly found it to be useful."

I also noticed something in the July 2003 issue of PC World magazine: On page 24 they have a test report comparing the AMD Athlon XP 3200+ with Intel's Pentium 3GHz chips on business and graphic applications. In the fine print under the results, they note that they tested all P4 systems with Hyper-Threading turned off. But they don't say why. (All the tests were done w/Windows XP, incidentally.)

I know too that I've read other comments on the Web about certain apps taking a slight hit w/HT, but I don't remember exactly which sites or when. From now on I'll print out those things when I see them.

Cutting to the chase, I now have the following comments and questions, mainly for you, Duvie. I thought I'd post 'em here rather than PM you so any other interested viewers can read 'em too if they want.

1) Re your excellent reply posting on 3/7: The "5-10% slower" claim may very well have been around the time the 3.06GHZ P4 came out, which was the first appearance of HT. Maybe this chip WAS, in fact, 5-10% slower on some apps tested and maybe HT has been refined since to overcome that (?). That would explain a lot, huh? And I don't know if any credible magazine or lab would have been dumb enough to test HT on a known non-HT compliant OS like Windows 2000, but I suppose it's possible. Needless to say, any results of such a test would be useless.

Also, I don't personally know if the new Microsoft Office 2003 suite has any optimizations for HT. I am using Office 2000 Premium and don't have any need to upgrade anytime soon, so it won't matter to me anyway. I can wait for Longhorn for my next upgrade to Office. Besides, I think that Hyper-Threading for Word, Excel, et al. would kinda be overkill, wouldn't it?

2) You also mention in your post, "Since I have been testing, I can think of half a dozen apps that have added HT features to them that I use personally." Cool -- do you mind if I ask which ones? Maybe they're things I might wanna use or try one of these days.

3) Do you happen to know if any of the following are "floating-point-heavy" apps: Adobe Photoshop CS, Illustrator CS, InDesignCS, Microsoft Word, or Microsoft PowerPoint? And what would be some examples of apps that are considered "special instruction" rather than floating point? Are FPU apps mainly math things like Excel, or scientific programs, or ????

4) I think you may have hit on something when you talked about priority settings. This is a good thought. So my obvious question now is, how can I find out (before I buy the chip/motherboard!) what the priorities are gonna be for Photoshop CS, Illustrator CS, InDesign CS, Word and PowerPoint and whether or not they might in fact slow one another down if running concurrently in some combination or another? Geez. (And don't advise me to call Microsoft or Intel -- they'll think I'm out of my mind for asking such a question, and they won't know [or tell me] anyway!) And is such "prioritization" adjustable in the BIOS or something? Based on your theory, a Hyper-Threaded chip could cause a slight slowdown because of prioritization rather than its actual symmetrical processing task execution. Interesting ....

This topic is really getting down to a level of minutia that's probably not gonna make a huge difference in real-world performance, but I would like to figure this out and nail it down so I can make the best buying decision for both now and the near future. The price difference between an HT and a non-HT P4 2.8GHz chip is not huge, but it will impact my chipset and mobo choice as well and everything together could translate into notable extra $$. And if "more and more apps are being optimized for HT" like I keep reading on Intel's Web site and all over the place, I'd sure as heck like to know WHICH ONES ?!!!!!! so I can consider that as well. Gosh, is this too much to ask of the computer gods?

Whew. Too much for a late night posting. Goodnite. Look forward to the replies.

Ken

PS: Why do you think PC World would have run the test I mentioned above with HT turned off? The only reason I can think of is that it would slow the apps down and they didn't want that. (None were optimized for HT, so it couldn't speed them up.) So if HT would slow them down, then what's the deal???????

 

[myocardia]

Since no Athlons are HT-enabled, it wouldn't be fair for them to test an XP3200 against a P4C with HT enabled, plain and simple.

 

[Accord99]

Originally posted by: Ken90630
4) I think you may have hit on something when you talked about priority settings. This is a good thought. So my obvious question now is, how can I find out (before I buy the chip/motherboard!) what the priorities are gonna be for Photoshop CS, Illustrator CS, InDesign CS, Word and PowerPoint and whether or not they might in fact slow one another down if running concurrently in some combination or another? Geez. (And don't advise me to call Microsoft or Intel -- they'll think I'm out of my mind for asking such a question, and they won't know [or tell me] anyway!) And is such "prioritization" adjustable in the BIOS or something? Based on your theory, a Hyper-Threaded chip could cause a slight slowdown because of prioritization rather than its actual symmetrical processing task execution. Interesting ....

Applications almost always run at Normal priority in Windows XP. If you're really running two programs concurrently where they are both performing CPU intensive tasks (and not just running idle), then in a non-HT system, each program will take turns getting access to the CPU and take roughly twice as long to perform a task compared to if they're running by themselves; so there is usually no time difference running them sequentially or concurrently. With a HT system, both programs are executed by the CPU simultaneously, taking better advantage of the available processor resources. Usually, running two applications concurrently is 15-25% faster than running them sequentially on a HT P4.

The priority situation where HT can cause problems is when you have a background application running at Idle priority and is only supposed to run when the system has nothing else to do and pause whenever anything else needs the CPU. In a non-HT system, this is no problem because the OS only has one CPU to worry about. With a HT system, where you have two logical CPUs available, then this background application can still run even when you're running a higher priority application that is demanding 100% of the CPU. Since many resources are shared in HT, the performance of the higher priority application will suffer, but you get work done in the background thread. Turning off HT would give you the full performance for the higher priority application, while HT enabled gives you more total work done per second, if you combine the work done by both threads. Or you can just turn off the background application and get the full performance of the single application with HT still enabled.

 

[LTC8K6]

The hyperthreading that didn't work so good was only on the earlier P4 Xeons and some P4 ES's, back when people were trying to enable it on regular P4's. The 3.06B and everything newer should be okay.

 

[Wingznut]

Originally posted by: myocardia
Since no Athlons are HT-enabled, it wouldn't be fair for them to test an XP3200 against a P4C with HT enabled, plain and simple.

That's pretty ridiculous.

 

[Jeff7181]

Originally posted by: Wingznut

Originally posted by: myocardia
Since no Athlons are HT-enabled, it wouldn't be fair for them to test an XP3200 against a P4C with HT enabled, plain and simple.

That's pretty ridiculous.

Yeah, pretty much. With that reasoning, I guess you'd have to disable the on die memory controller of the A64 for a fair comparison with the Pentium 4 since the P4 doesn't have an on die memory controller and it's not fair to let the A64 use it.

 

[DAPUNISHER]

Originally posted by: Wingznut

Originally posted by: myocardia
Since no Athlons are HT-enabled, it wouldn't be fair for them to test an XP3200 against a P4C with HT enabled, plain and simple.

That's pretty ridiculous.

I agree. Both the CPUs mentioned compete directly with one another for sales and therefore any and all features of both should be tested in order to give a complete evaluation of their performance and capabilities. What isn't fair is depriving their readers of the HT performance data which could play a major role in the purchasing decision of which best suits their needs.

 

[Wingznut]

Originally posted by: Jeff7181

Originally posted by: Wingznut

Originally posted by: myocardia
Since no Athlons are HT-enabled, it wouldn't be fair for them to test an XP3200 against a P4C with HT enabled, plain and simple.

That's pretty ridiculous.

Yeah, pretty much. With that reasoning, I guess you'd have to disable the on die memory controller of the A64 for a fair comparison with the Pentium 4 since the P4 doesn't have an on die memory controller and it's not fair to let the A64 use it.

And we should limit Raptors to only 7200rpm when comparing with typical hd's, right?

 

[Jeff7181]

Originally posted by: Wingznut

Originally posted by: Jeff7181

Originally posted by: Wingznut

Originally posted by: myocardia
Since no Athlons are HT-enabled, it wouldn't be fair for them to test an XP3200 against a P4C with HT enabled, plain and simple.

That's pretty ridiculous.

Yeah, pretty much. With that reasoning, I guess you'd have to disable the on die memory controller of the A64 for a fair comparison with the Pentium 4 since the P4 doesn't have an on die memory controller and it's not fair to let the A64 use it.

And we should limit Raptors to only 7200rpm when comparing with typical hd's, right?

Exactly!

 

[Verdant]

2 reasons that you might get slower performance

1. SMP/HT overheard running any single threaded application in a setup with more than cpu tends to run slightly slower, this is typically due to synchronization. But you will most likely never notice this slowdown, as you are almost always running more than one thread, which improves the cpu scheduling for your "important" thread and reduces the non-yourapp slices being run on the cpu, but its still possible to have an occasional slowdown with some applications.

2. If you applications distributed computing application like folding or seti, as although these processes/threads are usually marked low priority, if there is nothing for the "virtual" cpu to do, they will be given slices on the virtual cpu... and since it is not actually a seperate cpu, they will "steal" cycles from your main application. The reduction in performance will again be minimal, as intel has worked hard to minimize this issue, also if the program is designed to run in multiple threads, then this will not be an issue whatsoever, as any secondary threads will run instead of tha background thread.

Overall those two points are pretty moot, and HT is a great Tech, for your situation i am highly confident you will notice an increase in performance with the HT part.

 

[Ken90630]

Thanks, guys. I wanna respond to some of your comments and also ask a few (hopefully shorter) questions.

Myocardia -- Sorry, I guess I neglected to mention in my last post that the PC World test report was done on non-HT- optimized apps running by themselves (no SMP or multi-tasking going on). So it's not like they would have been asking the Athlons to perform SMP, which they obviously can't do, against HT P4s that could. So from that standpoint the test would still have been 100% fair with HT enabled, in my opinion, and I'd still like to know why it was turned off. Any guesses anyone? [I suppose a conspiracy-theorist might joke that the testers own a bunch of AMD stock or something and thus wanted the Athlons to look better, but I'll assume that wasn't the case. ] If anyone cares, the apps they tested were Photoshop 7.0 (not the newer, HT-optimized CS version I plan to use), AutoCAD, Musicmatch 7.1, VideoWave 1.5, and two games: Unreal Tournament 2002 and Return to Castle Wolfenstein (neither of which matter to me as I'm not a gamer). If the article wasn't so old now, I'd fire off an e-mail to PC World and ask them why they turned the HT off. I still wanna know.

As Dennis Miller says, "And another thing:" I agree with the other guys here that components should be tested at full strength. As long as the prerequisite of comparing apples to apples is taken care of at the outset, all's fair in love & testing. I don't think any of us want to be pressing a finger up against a 10K Raptor platter to slow it down to 7200 rpm, a la Fred Flintstone dragging his feet on the ground to simulate brakes in his prehistoric-mobile, just so a competitor's drive could compete better!

Accord99 -- Great post. I already had a handle on everything in your first paragraph, but you say some things in your second paragraph that I wanna discuss. First, educate me on something: What would be an example of an app running at "idle priority," and which determines the priority -- the CPU or the OS? Here's what's confusing me a bit: Let's say I'm using Photoshop and Illustrator at the same time. I have Photoshop open and onscreen, but Illustrator is minimized (I'm clicking back & forth between the two windows, minimizing one when I bring the other up). So Illustrator is open, but it's not doing anything except sitting there. In this sitch, is it using CPU resources or is it just sitting in RAM or something until I click on its window to bring it back up and do something with it? When Illustrator is minimized, is it receiving "Idle priority" from the CPU or Win XP, or is priority irrelevant 'cuz it's not using CPU resources anyway until I bring its window back up and actually execute a task with it?

As I currently understand muti-tasking, it consists of two or more apps EXECUTING TASKS simultaneously rather than just being open at the same time. Is this accurate?

Again, re my question about "Idle priority" apps, what would be some examples?

What you describe in your paragraph 2 is cutting to the chase of helping me figure this out. What I want to avoid is, like you describe, a situation whereby with a HT-enabled chip, a high priority app could be slowed down a bit because a lower priority app will still be executing tasks instead of being idled. I think for my purposes, if I'm running Photoshop, I want my CPU to be devoting all its resources to Photoshop rather than giving some cycles to a lesser-priority app in some grand (albeit noble) plan to save overall total time executing both.

As long as I can, as you say, "just turn off the background application and get the full performance of the single application with HT still enabled," I should be happy. This point is actually my whole reason for asking all this -- what I wanted to establish from the beginning is whether or not any single apps running by themselves (no SMP going on) would actually slow down if I have a Hyper-Threading chip. I had read (somewhere!) that some would, hence my original post.

Now, this brings up the "synchonization" issue mentioned by Verdant -- I wanna discuss this a bit too. When you say, Verdant, "But you will most likely never notice this slowdown, as you are almost always running more than one thread," that actually might NOT be the case with me -- unless just having two apps open at the same time constitutes running more than one thread. See, I don't do any of what most people cite as examples of multi-tasking: things like running a virus scan while editing photos, or ripping MP3s while downloading files from the Web at the same time, or encoding video files while doing something else at the same time .... For people who do this sort of stuff, Hyper-Threading and SMP are no-brainers. It's ideal technology! But I won't have this need on this computer -- at least not right now or in the foreseeable future.

So for me, when you say, "but its still possible to have an occasional slowdown with some applications," that's making me a bit uneasy. The only use I'll have for HT is with some particular filters in Photoshop CS that are optimized for HT -- that's pretty much it. So I'm back to my original question of WHICH APPLICATIONS WILL I EXPERIENCE A SLOWDOWN with if I'm running them by themselves (no SMP) on a Hyper-Threaded P4? If Word, PowerPoint, Illustrator or InDesign CS running by themselves will be slowed down with a HT chip, I don't want HT. If these and other single-threaded apps will run the SAME speed on a HT chip and my HT-optimized Photoshop CS filters will apply faster as well, then I want it. It's that simple -- everything I've said in these mile-long posts boils down to that.

I'll await you guys' responses, and maybe see if Duvie has two cents to throw in, then make a decision. Thanks a ton for your input and for educating me on this. I appreciate your patience.

Ken

 

[Accord99]

Originally posted by: Ken90630
Accord99 -- Great post. I already had a handle on everything in your first paragraph, but you say some things in your second paragraph that I wanna discuss. First, educate me on something: What would be an example of an app running at "idle priority," and which determines the priority -- the CPU or the OS? Here's what's confusing me a bit: Let's say I'm using Photoshop and Illustrator at the same time. I have Photoshop open and onscreen, but Illustrator is minimized (I'm clicking back & forth between the two windows, minimizing one when I bring the other up). So Illustrator is open, but it's not doing anything except sitting there. In this sitch, is it using CPU resources or is it just sitting in RAM or something until I click on its window to bring it back up and do something with it? When Illustrator is minimized, is it receiving "Idle priority" from the CPU or Win XP, or is priority irrelevant 'cuz it's not using CPU resources anyway until I bring its window back up and actually execute a task with it?

Again, re my question about "Idle priority" apps, what would be some examples?

The initial priority of an app is determined by the app itself, when it starts it will automatically give itself a level of priority (though depending on the user account level, Windows may not allow programs that seek the absolute highest priority levels). If your account has administrative privileges, then you can set it through Task Manager->Processes->right click on the application/process. There really aren't that many commercial/productivity apps that are designed to run in the background with Idle Priority; generally only the Distributed Computing clients like RC5, SETI, Folding@Home, etc will automatically run at Idle Priority.

In your scenario, the priority of Illustrator is not relevant since it is not doing anything and will not be requiring any CPU time, so if your were doing some processing in PS, PS would receive the full speed of the system in both the HT and non-HT cases. As well, since PS is multi-threaded, HT could improve performance for those filters whose workload can be split into two or more threads.

As I currently understand muti-tasking, it consists of two or more apps EXECUTING TASKS simultaneously rather than just being open at the same time. Is this accurate?

Hmm... multi-tasking seems to be a very broad term that's used a lot to cover a lot of things; this is definitely multi-tasking and is one of the scenarios where you should almost always see a performance benefit with HT. Single applications which can split its work into multiple threads (like PS, 3D rendering programs) will typically also benefit from HT.

As long as I can, as you say, "just turn off the background application and get the full performance of the single application with HT still enabled," I should be happy. This point is actually my whole reason for asking all this -- what I wanted to establish from the beginning is whether or not any single apps running by themselves (no SMP going on) would actually slow down if I have a Hyper-Threading chip. I had read (somewhere!) that some would, hence my original post.

If you're just running a single application, then having HT enabled or disabled should still result in virtually identical performance. The P4 processor, combined with Windows XP, is intelligent enough that it will automatically recognize when it is running a single thread or two threads, and allocate its resources accordingly. In the Anand review of the 3.06GHz HT Processor, there's a list of the performance difference in various apps when HT is enabled/disabled. For all but one of the single-threaded benchmarks, the scores are almost identical.

http://www.anandtech.com/cpu/showdoc.html?i=1746&p=6

 

[chsh1ca]

Originally posted by: myocardia
Since no Athlons are HT-enabled, it wouldn't be fair for them to test an XP3200 against a P4C with HT enabled, plain and simple.

As I've said elsewhere, there is a very tangible reason they should. Intel provided HT as a technology to recoup performance they lost IPC-wise when transitioning to a longer staged pipeline. With HT, you can use the execution units on a second process while the first thread being processed is not. Intel is providing a way for you to get the most performance out of your processor, so it is justifiable to compare the two.

Pipeline stage differences are the main reason why you will likely not see a HyperThreading-enabled processor from AMD any time soon. The fact of the matter is, AMD processors don't NEED to have HT in order to give their best performance, since they have a much shorter pipeline and do not take the large penalty incurred on a branch predictor miss. It's not as if they are giving away free performance, or there are two cores on a single die, it's just performance the processor should have, but cannot in a single-thread environment. By making it handle multiple concurrent threads, they can recover the performance and make up for the low IPC of the processor, to an extent.
Anandtech covered the details of pipelines in their latest article on the Prescott: http://www.anandtech.com/cpu/showdoc.html?i=1956&p=2