athlon 64: why is it's poor multitasking ignored/downplayed?

[CaiNaM]

Originally posted by: MartinCracauer
If that's true, then the only reason for all this is that the Windows XP scheduler for the non-SMP case sucks.

well, again, that could certainly be the case, but the wouldn't explain why the athlonXP runs 2 instances of daoc noticeably better than the a64.

Now, I don't say that Hyperthreading is useless, for me it can bring 15-25% on the right workload.

well, yea... to say it's useless is simply fanboy FUD. at the same time, it doesn't benefit all applications either, but this isn't really about the overall viability of HT.

However, the observed effect that interative performance of a foreground process is much worse on the AMD than on the hyperthreaded P4 appears to be purely a case of bad scheduler policy in the case of Windows XP.

again, fair enuff.. but it's still not a good thing for a mainstream cpu to have to rely on the 'bad scheduler policy' of the OS that dominates the mainstream market when the competition (in this case intel) is not affected. the one thing that's out of place tho is why the aXP (in the specific case relating to daoc) performs better than the a64. does windows 'schedule' differently for a64 than aXP? that doesn't make sense....

Can you please run that above test with HT tunred off to verify that theory?

i believe that was discussed earlier in this thread, however I can take another look later when I get the chance.

 

[Etruscan]

Originally posted by: glugglug
I wonder.... does your Athlon64 system have the same Win32K.SYS file as your AthlonXP and P4 systems? It is conceivable that it might have a new HAL to support DEP. The alternate HAL could define whatever quanta sizes it wants. I don't have a 32-bit proc WinXP system to check.

Also, did you try the registry change? (Set HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\PriorityControl\Win32PrioritySeparation to 36 decimal or 0x24 hex)

My Athlon64 system multitasks a lot better than the P4 Williamette I use at work.

I'm using an A64, so this thread was of some interest to me. Since no one had ever seemed to try changing this registry value, I did.

With default values.

The RTHDRIBL Window in the front ran between 25-35 fps, window in the back ran between 15-20.

With 0x24

The RTHDRIBL Window in the front ran between 32-45 fps, window in the back ran between 13-18.

Variability was no different from a visual observation standpoint, which is worth fuckall of course.

For what it's worth, launching firefox did not affect the FPS in the front RTHDRIBL window with either setting, but the rear window (now third window on screen) dropped to around 10-14/8-13 FPS.

Changing settings seemed to make Firefox more responsive, but I wouldn't put any weight into that, as it was the result I wanted.

 

[uOpt]

Originally posted by: CaiNaM

Originally posted by: MartinCracauer

Can you please run that above test with HT tunred off to verify that theory?

i believe that was discussed earlier in this thread, however I can take another look later when I get the chance.

I scanned the whole thread yesterday but couldn't find a posting where somebody would run a set of tests with HT, then turn off HT and run the same tests. If I overlooked something please point me to it.

That would be incredibly interesting to see whether my theory that the different Windows XP schedulers cause a difference holds any water.

Especially now that I started to gather benchmarks on my Linux-running P4.

 

[Lithan]

Originally posted by: CaiNaM
and everquest players out there running 2 EQ clients (was contacted by someone who does this on a p4, so it's obvioulsy done in that game as well) on an a64/aXP/p4 perhaps can share their observations?

Eq was one of the programs I've ran multiple instances of on one machine. Everquests autofollow was godawful at the time, but running twice on one machine A character would get lost less often than other group members would if they trusted the autofollow command. I ran two instances of EQ on a k62 at one point in time.

Originally posted by: CaiNaM
going back to the example of running 2 rthlbl, on the p4 both instances run slower, but smoothly (the fps on both instances are consistent). on the a64, the fps varies greatly on both instances, which would seem to support the fact on the athlon64 both processes are stopping/starting constantly, while on the intel both run continuoulsy, albeit slower than when running a single instance.

You apparently don't understand how quickly modern processors operate.

 

[HeroOfPellinor]

Originally posted by: jiffylube1024

Oh, and I own both an Athlon64 2800+ and a P4 2.8. The only fanboi here is Cainam with this thinly veiled "AMD SUXXORS" thread.

Since he founded the thread on a problem that he was personally experiencing in a program that is very important to him, I fail to see how this is a thinly veiled "AMD SUXXORS" thread. He didn't start up the thread on hearsay or other nonsense 'I hear these Athlons don't multithread worth a dang' but posted an actual problem he was having with his Athlon64 (and rushed to a hasty conclusion).

I would agree.......had the thread been dropped or he merely ceded the fact that the P4's "mulitasking" benefits extend no further than a miniscule realm of multi-account MMORPG players and space-limited professional-DVD-burners. The fact that we're still being asked "what's wrong with the AMD64" should give you some incite into his prejudice.

Look, it's running two instances of DAOC each at 50% processor power so they both look in sync. Splendid, but that is NOT how it SHOULD be behaving. The foreground application SHOULD be receiving priority over the background application. The fact here is that, because you WANT both instances to run at the same speed, this particular quirk of hyperthreading just happens to work for you. Congrats! Cheers! But then taking a logical leap backwards and pondering why the Athlon64 doesn't multitask wrong is ridiculous.

If Mythic had actually DONE any of the programming for the game instead of piecing together other peoples' code, they would be able to design it from the ground up to run as a mulit-instance application and then both instances would run as one application. So you've got a borked application running on a borked processor that miraculously produces the results you want. Thank your lucky stars, stick with the P4, and stop falsly accusing the Athlon64 of being flawed or inferior in some way.

 

[CaiNaM]

Originally posted by: HeroOfPellinorI would agree.......had the thread been dropped or he merely ceded the fact that the P4's "mulitasking" benefits extend no further than a miniscule realm of multi-account MMORPG players and space-limited professional-DVD-burners. The fact that we're still being asked "what's wrong with the AMD64" should give you some incite into his prejudice.

but the fact is it extends further than that, which means you're the one showing prejudice, not me.

Look, it's running two instances of DAOC each at 50% processor power so they both look in sync.

actually, that's incorrect. this is just a guess, but from the behavior it's more like 90/30. yes, this is greater than 100%, but again, the second process is using portions unused by the first thread. part of the reason the athlon is faster even tho running at lower frequencies is that intel HAS to ramp clock speed to make up for the inefficiency of it's longer data stages.

since we're specifically speaking of daoc here, what you fail to consider is that rendering the display for the second client is not necessary. it only needs to be able to exchange data with the server.

Splendid, but that is NOT how it SHOULD be behaving. The foreground application SHOULD be receiving priority over the background application.

according to whom? if that's the case, we wouldn't need multitasking operating systems. we'd simply get by with window swapping...

The fact here is that, because you WANT both instances to run at the same speed, this particular quirk of hyperthreading just happens to work for you. Congrats! Cheers! But then taking a logical leap backwards and pondering why the Athlon64 doesn't multitask wrong is ridiculous.

again, completely inaccurate. i just want the second process to RUN. it doesn't have to render any displays ro anything else; it merely has to keep the data flowing between the client and the server. the intel is able to do so. the a64 is not.

If Mythic had actually DONE any of the programming for the game instead of piecing together other peoples' code, they would be able to design it from the ground up to run as a mulit-instance application and then both instances would run as one application. So you've got a borked application running on a borked processor that miraculously produces the results you want. Thank your lucky stars, stick with the P4, and stop falsly accusing the Athlon64 of being flawed or inferior in some way.

again, an assumption made to support your bias, but based on no facts. this wouldn't explain why an athlonXP is also able to do what the a64 is not.

 

[CaiNaM]

Originally posted by: Lithan
Eq was one of the programs I've ran multiple instances of on one machine. Everquests autofollow was godawful at the time, but running twice on one machine A character would get lost less often than other group members would if they trusted the autofollow command. I ran two instances of EQ on a k62 at one point in time.

but do it now, with processors relevant to this discussion. doesn't matter how 3 yrs ago EQ ran on a k6...

You apparently don't understand how quickly modern processors operate.

heh.. think the shoe is on the wrong foot.. not quickly enough that it can "swap" between 2 tasks seamlessly.

[rant]
what is it with you people? the a64 does many things very well. better than the competition. so it doesn't do one thing as well. does that fact like hurt you so bad you have to defend it by acting like the issue isn't there?

the one thing it doesn't do as well as the intel happens to be something that is important to me. that it isn't important to you is certainly fine, but it doesn't mean the issue isn't there. if it's that important for you to show your love of amd, then praise it's virtues, but don't pretend valid issues don't exist. there are many things where the intel falls behind amd, and frankly there are "corporate" decisions regarding intel i don't like. it doesn't change the fact their cpu does some things well.

if you need to blast intel for their shortcoming, feel free - there are many valid points one could make, but this thread isn't about brand loyalty, it's about an issue that exists, and while originally i was looking for a 'fix', it's apparently not there, so now we're just discussing what the reason is, not whether intel is greater than amd... because the truth is, on many fronts intel is lacking when comparing p4 to a64, or amd to intel for that matter...
[/rant]

 

[HeroOfPellinor]

Originally posted by: CaiNaM

Originally posted by: HeroOfPellinorI would agree.......had the thread been dropped or he merely ceded the fact that the P4's "mulitasking" benefits extend no further than a miniscule realm of multi-account MMORPG players and space-limited professional-DVD-burners. The fact that we're still being asked "what's wrong with the AMD64" should give you some incite into his prejudice.

but the fact is it extends further than that, which means you're the one showing prejudice, not me.

Look, it's running two instances of DAOC each at 50% processor power so they both look in sync.

actually, that's incorrect. this is just a guess, but from the behavior it's more like 90/30. yes, this is greater than 100%, but again, the second process is using portions unused by the first thread. part of the reason the athlon is faster even tho running at lower frequencies is that intel HAS to ramp clock speed to make up for the inefficiency of it's longer data stages.

intel p4 HT enabled
--------------- CPU ---------- MEM -----
game.dll ---- 50 ---------- 305,736k
game.dll ---- 50 ---------- 332.723k

That was you who posted that, right?

The fact here is that, because you WANT both instances to run at the same speed, this particular quirk of hyperthreading just happens to work for you. Congrats! Cheers! But then taking a logical leap backwards and pondering why the Athlon64 doesn't multitask wrong is ridiculous.

again, completely inaccurate. i just want the second process to RUN. it doesn't have to render any displays ro anything else; it merely has to keep the data flowing between the client and the server. the intel is able to do so. the a64 is not.

Others have already posted in this thread they can run two clients on an Athlon64 just fine, so the claim that the second client is receiving 0% processor time is ridiculous. Scheduler is sending it as much as it can spare which will be more than 0%. :roll:

If Mythic had actually DONE any of the programming for the game instead of piecing together other peoples' code, they would be able to design it from the ground up to run as a mulit-instance application and then both instances would run as one application. So you've got a borked application running on a borked processor that miraculously produces the results you want. Thank your lucky stars, stick with the P4, and stop falsly accusing the Athlon64 of being flawed or inferior in some way.

again, an assumption made to support your bias, but based on no facts. this wouldn't explain why an athlonXP is also able to do what the a64 is not.

I'd love to see some of the "facts" you have to backup this assertion.

 

[Etruscan]

I think that "what is with" them is that you have made some sweeping statements about the processor that don't seem to pan out. The specific setup with Windows XP seems to handle multi-tasking poorly in certain scenarios, though not in others. You have made comments about how the problem is that the A64 processor multi-tasks poorly. It's a generalization that doesnt seem to hold up when others (Vee, etcetera) test with other programs. HeroOfPellinor seems to be so biased it hurts though.

I think the other thing that gets them is that you never seemed to try the suggestions regarding changing priorities and quanta(?) length in the registry keys.

Regardless though, I thought I would ask again: Do you get multitasking problems in any situation that doesn't involve running two instances of the same program? I kind of wonder if it's something to do with a stupid way of handling identical priority requests in WinXP task manager.

Also, are the results I posted for RTHDRIBL on my A64 similar to yours, or are our a64 setups having different multi-tasking results.

 

[HeroOfPellinor]

Originally posted by: Etruscan
HeroOfPellinor seems to be so biased it hurts though.

Whew, thanks for reminding me to throw my week old P4 system out the window. :roll:

 

[Zero Plasma]

Athlon 64 90nm 3500+ =Pure Gaming Joy.(When they come out.)

 

[CaiNaM]

Originally posted by: HeroOfPellinor
intel p4 HT enabled
--------------- CPU ---------- MEM -----
game.dll ---- 50 ---------- 305,736k
game.dll ---- 50 ---------- 332.723k

That was you who posted that, right?

yes, it was. and your point?

again, think a bit further; windows "thinks" there are 2 cpus (it doesn't differentiate between logical and physical); half of 2 cpus = 100% of 1 cpu. so no, it's not running the first process at 50%, rather somewhere between 50-100% cumilatively between physical and logical cpus.

also, when you switch task manager to "performance" view, it shows 2 cpus, both running 90-100% with 2 instances of daoc running.

Others have already posted in this thread they can run two clients on an Athlon64 just fine, so the claim that the second client is receiving 0% processor time is ridiculous. Scheduler is sending it as much as it can spare which will be more than 0%. :roll:

yet another fallacy, and another example of your spreading FUD to try and support your position. please show me where others have ran two clients "just fine".

as to your completely ignorant (and i mean that in terms of your lack of any first hand knowledge) statement of "the claim that the second client is receiving 0% processor time is ridiculous", check this video, and watch the cpu % hit zero at the beginning of the video. also, watch not only the forground client does not run smoothly, but char on the background client has pauses/stops for several seconds and cannot "stick" with the char running in the foreground cleint.

this video shows the same two clients running smoothly on a (much slower vs the a64 running 2.4ghz) p4 2.6c. the char on the second client follows smoothly and does not 'lag'.

i'll have to apologize for the poor 'production values' of the shaky cam video, but you can easily see the issues the athlon64 (first video) has trying to run dual clients. a picture is worth a thousand words, as they say....

I'd love to see some of the "facts" you have to backup this assertion.

yet another. it's obvious you don't want to see them, as this thread is full of them, however you appear to choose to ignore that which doesn't back you your brand bias.

Originally posted by: Etruscan
I think that "what is with" them is that you have made some sweeping statements about the processor that don't seem to pan out. The specific setup with Windows XP seems to handle multi-tasking poorly in certain scenarios, though not in others. You have made comments about how the problem is that the A64 processor multi-tasks poorly. It's a generalization that doesnt seem to hold up when others (Vee, etcetera) test with other programs. HeroOfPellinor seems to be so biased it hurts though.

i'm not sure they're 'sweeping', however it's been shown to affect many multi-tasking scenarios, and it's not just been observed by me, but from review sites as well (those that chose to compare it; most didn't even look at those scenarios).

I think the other thing that gets them is that you never seemed to try the suggestions regarding changing priorities and quanta(?) length in the registry keys.

actually i did; most had no effect, and some actually caused instability. these were discussed way earlier in the thread.

Regardless though, I thought I would ask again: Do you get multitasking problems in any situation that doesn't involve running two instances of the same program? I kind of wonder if it's something to do with a stupid way of handling identical priority requests in WinXP task manager.

again referenced earlier in the thread, regarding games, norton av, encoding, etc...

Also, are the results I posted for RTHDRIBL on my A64 similar to yours, or are our a64 setups having different multi-tasking results.[/quote]

i kind of thought your showed the same tendencies? fps varies wildly, with forground higher fps than background....

 

[CaiNaM]

Originally posted by: Smoke0
Athlon 64 90nm 3500+ =Pure Gaming Joy.(When they come out.)

could be. i'm sincerely interested in seeing how/if dual channel affects the situations discussed here. altho why not get a 90mm 3000+ now and oc it > 3500+

 

[Etruscan]

My RTHDRIBL results were run with the following setup.

MSI K8N Neo2 Platinum
A64 3000+ @2520 MHz
1 GB OCZ Platinum2 ram at 282 (2.5-3-3-10),2.8V
Geforce 6800GT @420/1150

So, they are dual channel results. Sorry, should have mentioned that.

 

[HeroOfPellinor]

1. Your video links are broken.
2. There are plenty of instances of the A64 being competitive with the P4 in "mulit-tasking".http://reviews.zdnet.co.uk/har...4015,39164010-5,00.htm Though I'll take your experiences as more authoritative than their synthetic benchmarks.
3. Though the scheduler sees two CPUs there's still only one L2 cache which IS a bottleneck at times. There are times when people are disabling hyperthreading because it does, in fact, cause both applications to slow down.

In any case, what are we arguing about here? Everyone admits that the P4's HT gives it slight advantages while bi-tasking. Having a special scheduler customized for you will do that for a CPU. Nevertheless, case closed on that issue.

The other part of your contention is that NOT ONLY is the Athlon64 inferior to the P4 for bi-tasking, but ALSO inferior to the AthonXP, correct? I think we're going to need some kind of hard evidence and then you may be on to something.

 

[jiffylube1024]

Originally posted by: HeroOfPellinor

Originally posted by: jiffylube1024

Oh, and I own both an Athlon64 2800+ and a P4 2.8. The only fanboi here is Cainam with this thinly veiled "AMD SUXXORS" thread.

Since he founded the thread on a problem that he was personally experiencing in a program that is very important to him, I fail to see how this is a thinly veiled "AMD SUXXORS" thread. He didn't start up the thread on hearsay or other nonsense 'I hear these Athlons don't multithread worth a dang' but posted an actual problem he was having with his Athlon64 (and rushed to a hasty conclusion).

I would agree.......had the thread been dropped or he merely ceded the fact that the P4's "mulitasking" benefits extend no further than a miniscule realm of multi-account MMORPG players and space-limited professional-DVD-burners. The fact that we're still being asked "what's wrong with the AMD64" should give you some incite into his prejudice.

Look, it's running two instances of DAOC each at 50% processor power so they both look in sync. Splendid, but that is NOT how it SHOULD be behaving. The foreground application SHOULD be receiving priority over the background application. The fact here is that, because you WANT both instances to run at the same speed, this particular quirk of hyperthreading just happens to work for you. Congrats! Cheers! But then taking a logical leap backwards and pondering why the Athlon64 doesn't multitask wrong is ridiculous.

No, HeroOfPelinor, you're getting things confused. Just because the P4 HT makes it look like two virtual CPU's with an equal (balanced) load, that is not how it actually works, and just because the first virtual CPU gets loaded to 100% and it looks like the CPU is 'half' being used, this is not the case.

If this was a dual core or dual processor motherboard, then yes, the two CPU's would reflect their respective activity, but in HT, it's not like half of the CPU is used as HT processor #1 and half as HT processor #2.

3. Though the scheduler sees two CPUs there's still only one L2 cache which IS a bottleneck at times. There are times when people are disabling hyperthreading because it does, in fact, cause both applications to slow down.

You're doing what Lithan was doing before and that is introducing irrelevant arguments to the discussion. First of all, I don't really know of anyone who find HT to slow down their computer (it would have to be a VERY specific example) and second, this is not what the discussion is about. Even if HT was 20% slower in a single threaded app (and it is not), that would have no bearing on our argument because we are talking about running multiple demanding apps on HT.

 

[Lithan]

Originally posted by: jiffylube1024
You're doing what Lithan was doing before and that is introducing irrelevant arguments to the discussion.

Example? Or are you just being a prick?
And in a thread that exists to badmouth the A64's multitasking, pointing out situations where it performs better at it is an obviously relevant argument. We've got fifteen pages of Cainam talking out his ass to justify his irrational belief that his situation (which is very unusual) means that "Athlon 64... (has) poor multitasking..."

Origionally posted by: Cainam

Origionally posted by: Lithan
You apparently don't understand how quickly modern processors operate.

think the shoe is on the wrong foot.. not quickly enough that it can "swap" between 2 tasks seamlessly.

Haha. That is absolutely priceless. OH MY GOD! SERVICES.EXE IS RUNNING! SMSS.EXE MUST BE FROZEN! OH GOD NO! WINLOGON IS RUNNING TOO! OH GOD NO! AND GAIM! AND MOZILLA! LORD HELP ME!

 

[jiffylube1024]

Originally posted by: Lithan

Originally posted by: jiffylube1024
You're doing what Lithan was doing before and that is introducing irrelevant arguments to the discussion.

Example? Or are you just being a prick?

No, you and Pelinor are both taking offense to Cainam's thread because it claims the Athlon64 is 'bad' at something, which is apparently a sore spot for both of you guys. Pelinor, at least, is open to reason, and only was confused of a couple things (like thinking HT splits the CPU load into two equal sizes, which it doesnt).

Just to give you what you want, I'll repost some of your quotes again to show you how you changed the topic multiple times.

Originally posted by: Lithan
It's not a matter of "filling" it's a matter of the longer pipeline causing cache misses so the cpu must wait while what it wants is fetched, HT allows the cpu to be doing something instead of just waiting. That's what I've been told at least. Again, we are talking in submilliseconds here. Only noticable in performance benchmarks. It wouldn't account for any noticable feelings of responsiveness.

Originally posted by: Lithan
I doubt mem bandwidth is the issue. In most instances the limits at which the cpu can address/recieve the data the memory sends hits a wall before or at the same time as the memory interfaces bandwidth. And since dcDDR does so little for A64's I'd guess that it's safe to say that it's in the same boat as Axp's and in short doesn't really need the extra memory bandwidth the way P4's do.

But Iwant, I agree that the reason your p4b trails is that it is clocked slower, along with it's much lower FSB, and no dcDDR. My p4b @ 3800mhz 224fsb dcDDR multitasks and singletasks just fine.

Originally posted by: Lithan
Superpi is a big intel-leaning benchmark. I'm refering to testing that found intel minus HT outperformed Intel + HT when running a single instance. It's only relavance is defending the statement that HT can cause a performance hit in instances where only one heavy-load process is running.

Originally posted by: Lithan
1. HT can degrade system performance. I seem to remember one of the pi calculaters takes (or took at one time) a significant hit when HT was enabled. I also remember some people testing and finding that disabling HT boosted 3dmarks slightly. There are other examples of this. Also, as motherboard/psu/etc manufacturers start producing parts designed to handle the load of Dual core cpu's, I would wager the dual cores will follow only slightly behind single core cpu's. Much like Opterons are right on the tail of FX these days.

Originally posted by: Lithan
BDSM. I agree with you on compatibility (...with ram, although I hear 939 goes a long way to fixing how well you can overclock with a gig+ of ram). But it's been known for a long long time that there is no stability problem with any current cpu's. AMD or Intel.

And actually. Intel gave me a lot more problems with drivers than AMD did. With AMD I can get chipset drivers for the nf2/nf3 from either the mobo manufacturer or Nvidia themselves. With intel I had to download the drivers for the Intel865PERL (A board which I do not own) to get chipset drivers for my p4p800. Asus (for whatever reason) doesn't have a driver download for my board, and Intel doesn't have a driver download for the 865 chipset (at least not marked as such). To get 865 drivers I had to download the drivers specifically for Intels 865 board and install them (No workaround was needed, they installed fine).

Stuff like 'HT can slow down the CPU sometimes', my 'non-HT multitasks just fine' (end of discussion), 'Intel had more problems for me' wasn't at all constructive and seemed to be a kneejerk reaction to the claim that AMD64 is a weaker multitasker ("bi-tasker") than the P4 HT, and even (according to Cainam, which I haven't seen other evidence for) other non HT CPU's.

And in a thread that exists to badmouth the A64's multitasking, pointing out situations where it performs better at it is an obviously relevant argument. We've got fifteen pages of Cainam talking out his ass to justify his irrational belief that his situation (which is very unusual) means that "Athlon 64... (has) poor multitasking..."

Again, I don't see how you think this is a thread "to badmouth the A64's multitasking" . Cainam made a finding on his own personal setup, people got pissed off because apparently, the P4 was 'better' at something than the A64, which they paid good money for, and now it's just the usual fanboy trench warfare.

I bet you Cainam even thinks that the Athlon64 is a better chip than the P4 C/E series overall (I know I do), but he's just annoyed at this problem he's having running something that worked for him before on his other machines.

 

[Lithan]

Those were all responses to someone elses comments brainiac. And read the thread title. This isn't a thread about 'Why is my system having trouble running 2 instances of DAOC'. This is a thread about "Athlon 64... it's poor multitasking...".

What I'm taking offense to is misinformation being spread. Of particular note is when Vee explained to Cainam how useless the review of HT vs A64 was due to it's asinine testing procedure, which he acknowledged then later refered to again as though it were relevant. He's making an arguement that "Athlon 64 (has) poor multitasking (capabilities)" and basing it on blatantly biased reviews and his own assumptions. All the while hiding his fanboyism under a very thin attempt at troubleshooting a very particular problem he has encountered (Which even if it were widespread (of which he's offered no evidence) STILL would require a massive leap of faith to hold the A64 at fault. Facts and reasoning simply do not support the assertions he has made. And so he argues defensively that he's right unless you prove him wrong. Then you come in and start flaming people as "introducing irrelevant arguements" if they dare to take a position other than; 'OMG you're so totally right! Hyperthreading is like the seed of Jesus baptised my processor!' There is a reason that people with years of experience with single processor, multiprocessor, and alternate-OS systems in home, business, workstation and server applications are at best luke-warm about Hyperthreading. And I promise you that it isn't that they are all Amd fanboys.

 

[piasabird]

When they write games sometimes the game is compiled in such a way that it runs better on one platform i.e Intel P4 and may not run as well on another platform. It just depends on which platform the software was devoloped on. Also if a program was designed to take advantage of HT it might run better on a HT capable P4.

 

[HeroOfPellinor]

Originally posted by: jiffylube1024

Originally posted by: HeroOfPellinor

Originally posted by: jiffylube1024

Oh, and I own both an Athlon64 2800+ and a P4 2.8. The only fanboi here is Cainam with this thinly veiled "AMD SUXXORS" thread.

Since he founded the thread on a problem that he was personally experiencing in a program that is very important to him, I fail to see how this is a thinly veiled "AMD SUXXORS" thread. He didn't start up the thread on hearsay or other nonsense 'I hear these Athlons don't multithread worth a dang' but posted an actual problem he was having with his Athlon64 (and rushed to a hasty conclusion).

I would agree.......had the thread been dropped or he merely ceded the fact that the P4's "mulitasking" benefits extend no further than a miniscule realm of multi-account MMORPG players and space-limited professional-DVD-burners. The fact that we're still being asked "what's wrong with the AMD64" should give you some incite into his prejudice.

Look, it's running two instances of DAOC each at 50% processor power so they both look in sync. Splendid, but that is NOT how it SHOULD be behaving. The foreground application SHOULD be receiving priority over the background application. The fact here is that, because you WANT both instances to run at the same speed, this particular quirk of hyperthreading just happens to work for you. Congrats! Cheers! But then taking a logical leap backwards and pondering why the Athlon64 doesn't multitask wrong is ridiculous.

No, HeroOfPelinor, you're getting things confused. Just because the P4 HT makes it look like two virtual CPU's with an equal (balanced) load, that is not how it actually works, and just because the first virtual CPU gets loaded to 100% and it looks like the CPU is 'half' being used, this is not the case.

If this was a dual core or dual processor motherboard, then yes, the two CPU's would reflect their respective activity, but in HT, it's not like half of the CPU is used as HT processor #1 and half as HT processor #2.

If both processes are utilizing 50% of the physical CPU that does tell you something....i.e. that neither is taking priority as a foreground app. Imagine a third app starts running in the background. Would one game stay at 50% or would both drop? I'll bet you the latter. Think about it......hyperthreading helps, but it's not essentially two CPUs as is being suggested.....you can't run two apps the same speed together as they would run independently.

I'm still waiting for some evidence concerning the XP outperforming the 64 in "mulittasking".....that is the assertion on which this thread title is based.

And, yes, cainam is biased, as he has already concluded in his mind that the A64 is bad at muilti-tasking. There's no, "IS the A64 bad?", it's "It's bad, now WHY." That annoys me.

 

[Tom]

"I'm totally in love with my A64 3400+. After sogging knee and thigh deep through swamps, for years with laggy Pentium 4s, that were management's favorites, but poorly suited for the technical work, A64 feels like a ride in a Bentley. It performs an incredible 91% (I've done my own application benchmarking) better than a 3.2 P4C. "

This is supposed to represent an unbiased opinion ?

 

[CaiNaM]

Originally posted by: HeroOfPellinor
1. Your video links are broken.

links are fine from 2 different networks i've tried (comcast dsl and my office DS3

2. There are plenty of instances of the A64 being competitive with the P4 in "mulit-tasking".http://reviews.zdnet.co.uk/har...4015,39164010-5,00.htm Though I'll take your experiences as more authoritative than their synthetic benchmarks.

yes, and links to where it's not. if indeed HT is the reason intel is better, it's also a fact HT is not beneficial in every circumstance.

3. Though the scheduler sees two CPUs there's still only one L2 cache which IS a bottleneck at times. There are times when people are disabling hyperthreading because it does, in fact, cause both applications to slow down.

that's certinaly possible, i haven't run into a situation with that scenario, so i can't confirm or deny this.

[/quote]In any case, what are we arguing about here? Everyone admits that the P4's HT gives it slight advantages while bi-tasking. Having a special scheduler customized for you will do that for a CPU. Nevertheless, case closed on that issue.

The other part of your contention is that NOT ONLY is the Athlon64 inferior to the P4 for bi-tasking, but ALSO inferior to the AthonXP, correct? I think we're going to need some kind of hard evidence and then you may be on to something.[/quote]

it doesn't make sense to me either... it's hard to explain. one thing tho is this (unlike the athlon 64) is based only on the observation of one particular task, and not really an 'overall' impression, altho the little barton has been a very good cpu, and i've been quite happy with it.

Originally posted by: Lithan
Those were all responses to someone elses comments brainiac. And read the thread title. This isn't a thread about 'Why is my system having trouble running 2 instances of DAOC'. This is a thread about "Athlon 64... it's poor multitasking...". <blah blah blah.. snipped in the interest of space>

and if you actually followed the logic than looking thru your rose colored glasses, you would see we've discussed performance on various levels with various apps. the doac issue is hardly the only example, altho it is what annoys me the most.

Originally posted by: piasabird
When they write games sometimes the game is compiled in such a way that it runs better on one platform i.e Intel P4 and may not run as well on another platform. It just depends on which platform the software was devoloped on. Also if a program was designed to take advantage of HT it might run better on a HT capable P4.

that's a good point, altho if we are to apply this specifically to daoc, why can the athlonXP handle two instances of that app better than the a64? is there something fundamentally different in the way the a64 handles 32bit apps differently than the aXP which something in daoc code would not run as well?

Originally posted by: HeroOfPellinor
And, yes, cainam is biased, as he has already concluded in his mind that the A64 is bad at muilti-tasking. There's no, "IS the A64 bad?", it's "It's bad, now WHY." That annoys me.

because it IS bad, at least with the various combinations of tasks that's been discussed here. facts shouldn't annoy you unless you have other agendas; they're just facts, nothing more, nothing less.

if you can show me which apps specifically run well in combination with other simultaneous cpu intensive tasks, i'd certainly mention those situations as well. problem is, i just haven't found any. it's fast in doing just about anything i've thrown at it (including encoding, long viewed as a weak point of amd), until i add another cpu intensive task - as the performance degrades noticeably.

if there's a case where it does something well and you want me to try it, i'm more than open to giving it a shot.

wow... and the BoSox finally won...

 

[CaiNaM]

well, it's certainly not the difference between an 2754 ans s939. new msi k8n platinum/3200+ combo does not perform any better...

 

[Thermalrock]

the athlon64, the opteron, and the xeon with em64t are 'real' 64 bit processors to say they arent is stupid.
they perform very well, they are starting to build supercomputers with insane numbers of opterons. if you go x86-64 you get a 'real' 64 bit processor that runs current software as good as anything else and your not chosing current over future software either as its killing off all over architectures. its the only architecture around that is safe to stay and the only architecture that is certain to have future software developed to run on it. they are starting to use x86 chips for every possible task. its also gonna evolve a lot faster than anything else because its what sells the most and and therefore what the most money for research can be put into. its less likely to die than it has ever been. saying opterons arent 64 bit cpus is so stupid, ignorant and unjustified. you cant even call em cisc cpus anymore either cuz they work just like risc processors on the inside. is the architecture perfect? nope. could they come up with a better architecture today if they put enough research money into it and started fron scratch? definitely. would that be the way to go? arguable. will it happen? well the one time someone took some money earned with x86 chips to develop a totally new architecture was a failure. is there one single thing that would justify to not call a64 or xeon em64t cpus 64 bit chips? no. why would someone do that? cuz the architecture of an alpha was more elegant? because the first 32 bits of the cpu are compatible with an existing 32 bit cpu? wtf? dec's alpha was the better cpu concept. the alpha was the faster cpu. the x86 architecture made enough money for research to take it to the performance level of the alpha, that was the point when the alpha lost its right to exist. there was no reason to buy an alpha for more money than a cpu just as capable. the itanium is alil different it was designed with money intel made with x86 cpus and they keep putting money they make with p4s into it. all those billions. think of it. all those billion the itanium cost intel could have been used to make p4s better or simple to sell p4 cheaper. amd would have had a lot harder time. intel wont keep doing this forever and even the itanium that had a head start compared to other cpu architectures because its also intels will fail and go away probably not that long from now.