Intel Proclaims Hyper-Threading Volume Breakthrough

[FelixDeCat]

Originally posted by: Pete84

Originally posted by: Adn4n

Originally posted by: crazySOB297
50 million people elected hitler into office

They were actually bullied by the SA into doing it. Just some FYI.

Although Hitler did say that they were a superior race and by and large the Germans believed it. Ask anyone about hyperthreading and they say it makes 'stuff' run faster.

That really reaching. You have to back to world war 2 to criticize hyperthreading? 😕

 

[DAPUNISHER]

Originally posted by: FelixDeKat

Originally posted by: Pete84

Originally posted by: Adn4n

Originally posted by: crazySOB297
50 million people elected hitler into office

They were actually bullied by the SA into doing it. Just some FYI.

Although Hitler did say that they were a superior race and by and large the Germans believed it. Ask anyone about hyperthreading and they say it makes 'stuff' run faster.

That really reaching. You have to back to world war 2 to criticize hyperthreading? 😕

LMAO! Where's that rule about the first person who mentions Hitler or Nazis losing the debate by default? I understand the idea is to equate the propaganda but it's still silly. HT works, and can offer far more than 5% gains in some situations. Just off the cuff I can point out that DC projects and WMV HD content are 2 areas where HT makes a significantly bigger impact than 5%. I swear this place is AMDMB v.2.0 anymore.

Try running the 1080p trailer of Step Into Liquid from the WMV HD content showcase with a AMD 2.2ghz XP or A64 with nV40 based card since it doesn't help decode the content for shat, and watch your CPU usage shoot to 75-100% and frames get dropped like crazy. Then try it with a 3ghz P4HT and observe how it can easily handle it with CPU never getting close to 70% and with silky smooth playback. Argue all you want but HT is real and it works, just not in all circumstances. I would say it's showing more immediate benefits to the typical OEM desktop user than the 64bit features of A64 does though.

I'm glad Felix is being a PITA and pushing Intel hard of late, this forum is too much of a Homogeneous
AMD lovefest anymore and it's needs some contrasting perspective to balance things out IMO. :beer:

 

[FelixDeCat]

lol! No kidding!

I can easily think of one real world application where hyperthreading makes of difference (that most people probably use or have used) and thats GTA:Vice City. Sure I use Office 2003, Active Sync for my HP iPAQ 6315 etc., but on Vice city you notice a big difference in frame rates if you disable HT.

And you know, Im going to regret saying this, but with all they hype about a64s I actually want to try one. But if my games are going to struggle, fageddaboutit! 😉

 

[Tom]

"exactly true. hyperthreading is at most a 5% performance gain WHILE RUNNING CONTINUOUS MULTITHREADED APPS. "

that isn't right, HT can give very large improvements in overall speed..

"In short, while there's no difference between the two systems with only one application running, Hyper-Threading gives no less than a 40% improvement with two running. Impressive."

http://www.hardwareanalysis.com/content/article/1557.7/

 

[LTC8K6]

Those of you saying that servers are slowed down by hyperthreading are probably thinking of the early Xeons which had an older version of HT that did not work as well as newer chips. Servers with those chips can indeed have problems with HT, but they would be old servers by now.

All desktop chips have the improved HT, as it was implemented with the 3.06B.

I am not sure when the Xeons got the improved HT, but HT was definitely improved after the initial Xeons came out with it and were disappointing.

If you will read AT's review of the 3.06B chip, you will read about the improved HT.

 

[CaiNaM]

if all you do is task swap between web browsers while listening to mp3's, or use your pc for nothing more than a gaming system, chances are the a64 will do everything just fine, and even excel over comparably rated intel chips.

run more than one app which requires large amounts of cpu time/processes however, and the a64s weakness becomes quite obvious.

all depends on how you use your pc. for many, HT is a valuable feature, for others it is not.

 

[cbehnken]

like I said, compiling (which is an A64 strong point) two apps at once runs very smoothly on my A64.

Felix, you better get an FX53 (or 55) if you want to beat your 3.9ee in much of anything.

No offense, but I don't see how you can worry about game performance from a processor standpoint if you are running a 5600u. A 2.66 P4 non HT will be GPU limited.

 

[CaiNaM]

well, don't take my word for it. here's an easy way for anyone to bring out the amd 64's weakness.

download RthDribl (Real Time High Dynamic Range Image-Based Lighting demo).

open up internet explorer. mine defaults to msn homepage, and takes maybe 1 sec. to load. now run rthdribl. open up ie again, and it took about 9 sec. to load and display the msn homepage. this is on an

athlon 64 3000+ (newcastle CG rev)
dfi lanparty ut nf3-25gb
1gb kingston hyperX CL2.5
fresh install winxp sp2
nvidia chipset driver v5.10

the same sequecne on a

p4c (northwood) 3.0
865PE Neo2-FIS2R
1GB Geil Value CL2.5

w/o rthdribl: 1 sec.
while running rthdribl: less than 2 sec.

making a 20mb zip file while running rthdribl, the p4 shows no framerate drop in rhtdribl (other than the normal 2-3 fps variance), the athlon drops from 70+fps down to 15-30fps, then back up to 70+ fps once winzip is done.

downloading a file using cftp pro @ 3.7mb/s, start rthdribl, and the d/l rate drops down to 1.7mb/s on the athlon 64. the p4? no change in d/l rate.

or even running 2 copies of rthdribl at the same time. on the a64, one client varies wildly from 27-45fps. the second one varies from 19-34fps.

do the same on the the p4c, and both clients run a pretty consistend 33-35fps.

blame the athlon 64. blame windows. maybe it's the nf3 chipset (same thing happens using a chaintech vnf3 250 mb)? i don't really care. regardless, the p4 platform handles running these multiple apps much better than the a64 platform. the difference is night and day, and glaringly obvious.

try it and see for yourself. even better if you have both platforms to compare.

 

[cbehnken]

Takes 3-4 secs compared to normal 1-2.

System remains completely responsive, no prob.

I can compile a seven million line program, run that program you showed me and I am typing right now. No jerks, no pauses, smooth running with all apps. What's the problem? Pulling over 30fps with Real Time High Dynamic Range Image-Based Lighting demo at the same time as compile. It does pull 50 when not compiling, but this is expected when we are dividing the cpu up. Program takes 106.03 seconds to compile compared to usual 85.

Maybe you should check your system out...

 

[cbehnken]

Originally posted by: CaiNaM

or even running 2 copies of rthdribl at the same time. on the a64, one client varies wildly from 27-45fps. the second one varies from 19-34fps.

This is not happening here. Maybe you have your foreground priorities too high. I am running two copies right now and typing and they are both running about 24 fps.

Sounds like an even split to me.

 

[SimsFreak]

Originally posted by: DAPUNISHER

Originally posted by: FelixDeKat

Originally posted by: Pete84

Originally posted by: Adn4n

Originally posted by: crazySOB297
50 million people elected hitler into office

They were actually bullied by the SA into doing it. Just some FYI.

Although Hitler did say that they were a superior race and by and large the Germans believed it. Ask anyone about hyperthreading and they say it makes 'stuff' run faster.

That really reaching. You have to back to world war 2 to criticize hyperthreading? 😕

LMAO! Where's that rule about the first person who mentions Hitler or Nazis losing the debate by default? I understand the idea is to equate the propaganda but it's still silly. HT works, and can offer far more than 5% gains in some situations. Just off the cuff I can point out that DC projects and WMV HD content are 2 areas where HT makes a significantly bigger impact than 5%. I swear this place is AMDMB v.2.0 anymore.

Try running the 1080p trailer of Step Into Liquid from the WMV HD content showcase with a AMD 2.2ghz XP or A64 with nV40 based card since it doesn't help decode the content for shat, and watch your CPU usage shoot to 75-100% and frames get dropped like crazy. Then try it with a 3ghz P4HT and observe how it can easily handle it with CPU never getting close to 70% and with silky smooth playback. Argue all you want but HT is real and it works, just not in all circumstances. I would say it's showing more immediate benefits to the typical OEM desktop user than the 64bit features of A64 does though.

I'm glad Felix is being a PITA and pushing Intel hard of late, this forum is too much of a Homogeneous
AMD lovefest anymore and it's needs some contrasting perspective to balance things out IMO. :beer:

That's why people invented ECC ram, and specific Server boards. I personal experience, you will 9/10 times get better performance from a server board then a regular board. Server boards are optamized for ECC, and for server functions, unlike regular boards. You cannot company an AMD2.2Ghz and a P43Ghz, why does everyone do that?

 

[Markfw]

I'm glad Felix is being a PITA and pushing Intel hard of late, this forum is too much of a Homogeneous
AMD lovefest anymore and it's needs some contrasting perspective to balance things out IMO.

Well maybe that has something to do with the fact that Intel has fallen behind by almost everyones estimation.

 

[CaiNaM]

Originally posted by: cbehnken

Originally posted by: CaiNaM

or even running 2 copies of rthdribl at the same time. on the a64, one client varies wildly from 27-45fps. the second one varies from 19-34fps.

This is not happening here. Maybe you have your foreground priorities too high. I am running two copies right now and typing and they are both running about 24 fps.

Sounds like an even split to me.

how does one change the "degree" of forground priority?

the only thing i am familiar with is the default proc selection scheduling between programs or background services, which are default (programs).

 

[KDKPSJ]

I personally think HT is one of the best and the most useful thing Intel ever brought to the consumer market. I encode video files a lot, and when I was using Barton, I just couldn't do any other thing while Virtualdub or TMPEGEnc runs. Of course I could make it lower priority, but then encoding rate dropped dramatically. So, I had to let the encoding runs while I sleep.

And now, I have 2.8C system, and all the sudden, it feels like I am in heaven. I can play around, even 3D games, while Virtual Dub with Real time priority eats up one of logical CPUs without dropping any rate at all. And while I sleep, I run two Virtual Dubs, which help me saving time a lot. (at least compare to Barton)

Well, I am actually thinking about getting Athlon64 as well, hoping it will do as good or better as 2.8C. Can someone with A64 please try to run two Virtual Dubs at once, and let me know how encoding rates are? Also, I want to know how tough multitask you can do while one task of Virtual Dub runs Real time priority. (If you can test with WinXP 64bit beta version, that would be even more awesome)

 

[cbehnken]

Originally posted by: CaiNaM

Originally posted by: cbehnken

Originally posted by: CaiNaM

or even running 2 copies of rthdribl at the same time. on the a64, one client varies wildly from 27-45fps. the second one varies from 19-34fps.

This is not happening here. Maybe you have your foreground priorities too high. I am running two copies right now and typing and they are both running about 24 fps.

Sounds like an even split to me.

how does one change the "degree" of forground priority?

the only thing i am familiar with is the default proc selection scheduling between programs or background services, which are default (programs).

That is what I am talking about. Set the scehduling to favor background services and things you run in the background will run smoother

 

[DAPUNISHER]

That's why people invented ECC ram, and specific Server boards. I personal experience, you will 9/10 times get better performance from a server board then a regular board. Server boards are optamized for ECC, and for server functions, unlike regular boards. You cannot company an AMD2.2Ghz and a P43Ghz, why does everyone do that?

You are correct of course and you get no argument from me on that, but servers are a different market than the desktop. HT is on Intel servers so it is still pertinent to the discussion. However, I am addressing desktop systems and usage and not the server side. My bad though, I should have made that clear 🙂

I also specifically stated A64 when suggesting the 1080p test not just the XP system. Besides, if AMD is going to give it a 3200+ rating than I see zero reason it should not be included. Sure, you and I know it isn't what it is touted to be but since it is performance rated so highly that makes it fair game for any comparisons to comparably clocked Intel desktop parts. Particularly given that many here are trashing HT! Why the hell should I give the XP3200+ a pass when it's a far more deserving of being bashed for it's lackluster performance compared to it's rating? It isn't really relevant to me what CPU the rating is obstensibly based on, the fact is it was the flagship desktop part@one point, or is my memory shot? Want to throw in 3ghz C/Deleron? I have no issue with that and would enjoy seeing those results as well 🙂

 

[packmule]

Blah blah blah... Intel stroking themseleves to bump up earnings.

 

[SimsFreak]

Well we started on servers and that's what I know. I work with desktops but I don't use HT technology on desktops so you got me on that. Yes Xeon's were the first to have problems, and still do. Most if not all versions with HT do not work well as a server. It's just a known fact, but for PC you do have me.

 

[carlosd]

Originally posted by: everydae
I personally think HT is one of the best and the most useful thing Intel ever brought to the consumer market. I encode video files a lot, and when I was using Barton, I just couldn't do any other thing while Virtualdub or TMPEGEnc runs. Of course I could make it lower priority, but then encoding rate dropped dramatically. So, I had to let the encoding runs while I sleep.

And now, I have 2.8C system, and all the sudden, it feels like I am in heaven. I can play around, even 3D games, while Virtual Dub with Real time priority eats up one of logical CPUs without dropping any rate at all. And while I sleep, I run two Virtual Dubs, which help me saving time a lot. (at least compare to Barton)

No way. I cannot do anything else in P4 HT while encoding in realtime, not even in normal priority, and the less hungry game will execute at 1 or 2 FPS. In my experience HT is completely useless if the applications are not optimized.