3.2GHz Dual-Core Extreme Edition Reviews

64bit_is_here said:
Exactly what struck my mind when i thought about this. See it like this:

previously it was 1thread vs 2threads
now it would be 2 threads vs 4 threads

while its true that the ratio is the same, intel wont have the same amount of performance advantage in the dual core scenario.
Click to expand...

It doesn't work that way though.
You'll get more responsiveness from HT... but in terms of actual performance, HT doesn't do a whole lot.

Intel would never want to put the mem controller in the die of the already huge prescott core, the prescott with no mem controller on the die is still quite a bit larger than a A64 WITH the mem controller. and now with dual core the size has just doubled... :p
Click to expand...

A memory controller is not that large. Besides, it doesn't have to run at such a high clockspeed. So if they can add a second core to a Prescott, surely they can add at least a memory controller to the single-core versions...
Besides, Intel already has working 65 nm samples in their lab... That should also help to put some extra transistors on those dies.
 
Dear God, this thread is growing faster than I can read it! :D
But seriously, I hope dual-core is adopted fairly quickly. Hopefully, the huge OEMs like Dell, etc, will start selling them pretty soon. The sooner DC hits the mainstream, the sooner programs will take advantage of it. And since ILP has pretty much hit the wall, we need something like dual-core or SMP to continue to gain preformance.
 
Scali said:
It doesn't work that way though.
You'll get more responsiveness from HT... but in terms of actual performance, HT doesn't do a whole lot.
Click to expand...
What? HT is great for keeping a large and inefficient pipeline full...but the fact is that you still have 2 threads in one execution pipe. Dual core (and indeed SMP) is superior in every way.
 
Scali said:
And you don't think I know the difference? Geez.
I was just making the point that P4s clean up in pretty much every SSE/SSE2/SSE3 benchmark that is multimedia-related, in a response to you saying I infer that Intel is superior.
Just re-read my posts and see. Perhaps you've not been following the actual conversation closely enough.
Click to expand...
No, I don't think you know the difference. Because when I said the only video renderer I knew of was GridIron for Adobe, you shot back with a smug, 'Well then you've never seen a Pixar film before' and mentioned Renderman.

Renderfarms used for 3d renders are USELESS for film, and Renderman has exactly jack squat to do with Adobe. You were going down a completely irrelevant tangent in regards to clustered computing when I started talking about video editing.

I could go into several other reasons why our Opteron platforms perform better in HD (superior I/O from the PCI-X bus, superior memory performance), but I don't think you'd listen. You still insist on talking about Athlon vs Pentium in regards to video editing. Neither of those can hold a candle to the requirements of a studio's digital intermediary workflow.

Oh, and here's something for you to chew on:

http://www.gamepc.com/labs/view_content.asp?id=x36o252&page=8
 
Morley said:
What? HT is great for keeping a large and inefficient pipeline full...but the fact is that you still have 2 threads in one execution pipe. Dual core (and indeed SMP) is superior in every way.
Click to expand...

I've discussed this before.
SMT and SMP are not mutually exclusive. And the point of SMT is to make the cores more efficient, not to run two threads at the same time... Obviously the second thread would never run very fast... unless your scheduler is not an SMT one, but an SMP one, in which case your SMT will generally get decreased performance rather than increased performance.
You can have SMT AND SMP, which will be superior to dual core, obviously.
So you can't say SMP is superior to SMT, because it doesn't make sense. SMT can be used to improve SMP performance, which would not be possible if SMP was superior to SMT in the first place. That just doesn't make sense.
 
Scali said:
I've discussed this before.
SMT and SMP are not mutually exclusive. And the point of SMT is to make the cores more efficient, not to run two threads at the same time... Obviously the second thread would never run very fast... unless your scheduler is not an SMT one, but an SMP one, in which case your SMT will generally get decreased performance rather than increased performance.
You can have SMT AND SMP, which will be superior to dual core, obviously.
So you can't say SMP is superior to SMT, because it doesn't make sense. SMT can be used to improve SMP performance, which would not be possible if SMP was superior to SMT in the first place. That just doesn't make sense.
Click to expand...

What he was saying is that while SMT does have tangable benefits, it is nothing compared to the performance boost of SMP.



I'm excited about dual core because it means that video encoding(specifically DVD trancoding) while playing games is a low performance hit. AMD will really benefit from Dual core because it will give AMD systems the responsiveness that Intel has enjoyed with HT.
 
Scali said:
I've discussed this before.
SMT and SMP are not mutually exclusive. And the point of SMT is to make the cores more efficient, not to run two threads at the same time... Obviously the second thread would never run very fast... unless your scheduler is not an SMT one, but an SMP one, in which case your SMT will generally get decreased performance rather than increased performance.
You can have SMT AND SMP, which will be superior to dual core, obviously.
So you can't say SMP is superior to SMT, because it doesn't make sense. SMT can be used to improve SMP performance, which would not be possible if SMP was superior to SMT in the first place. That just doesn't make sense.
Click to expand...
HT makes up for an inefficient pipeline. If you don't have a long pipeline to begin with, then HT won't do much for you.
 
I am sure that I don't know what I am talking about but here goes.

Can someone tell me why NO hyperthreading in the std dual cores?
Have to go EE to get it ?

WTHELL they were bragging for years how they implemented it in all their processors
NOW you have to pay double to get it?

what is going on...if you don't go EE you might as well go AMD

Sparks
 
Morley said:
HT makes up for an inefficient pipeline. If you don't have a long pipeline to begin with, then HT won't do much for you.
Click to expand...

I've seen people say this for a while now... I wonder where it came from, as I recall, the PIII was rumored to have a version of HT that was never enabled. Supposedly, it made the biggest difference on slower clocked machines. If HT could be on a PIII, I don't see why it only has to be on longer pipelined processors.
 
Morley said:
No, I don't think you know the difference. Because when I said the only video renderer I knew of was GridIron for Adobe, you shot back with a smug, 'Well then you've never seen a Pixar film before' and mentioned Renderman.

Renderfarms used for 3d renders are USELESS for film, and Renderman has exactly jack squat to do with Adobe. You were going down a completely irrelevant tangent in regards to clustered computing when I started talking about video editing.
Click to expand...

So just because you were not formulating very clearly.... you just assume that I don't know the difference? I find that both arrogant and insulting.
You simply talked about "clustered video rendering"... and that is exactly what Renderman does. What you actually meant was not specifically rendering of video, but rather (post-)processing, and perhaps editing and/or encoding too... Am I right?
What Renderman does, is obviously called rendering, hence the name. Or are you going to argue that Pixar doesn't know the difference either?
 
sparks said:
I am sure that I don't know what I am talking about but here goes.

Can someone tell me why NO hyperthreading in the std dual cores?
Have to go EE to get it ?

WTHELL they were bragging for years how they implemented it in all their processors
NOW you have to pay double to get it?

what is going on...if you don't go EE you might as well go AMD

Sparks
Click to expand...

I think they're having a hard time finding enough features to differentiate the EE from the regular processors.
 
Morley said:
HT makes up for an inefficient pipeline. If you don't have a long pipeline to begin with, then HT won't do much for you.
Click to expand...

Again you are not formulating clearly.
To be exact, HT makes up for 'bubbles' in the pipeline by taking extra instructions from a second thread (these are obviously guaranteed to be independent from the first thread, and are guaranteed to be scheduled in parallel, which obviously was no longer possible wiht the first thread, hence the bubbles). Whether you consider this an inefficient pipeline or not, is debatable. Depends on the amount of bubbles, compared to other pipelines. You could call the P4 pipeline inefficient, compared to other x86 pipelines.

However, the length of the pipeline doesn't have anything to do with its efficiency directly.
You should say "if you don't have a lot of bubbles to fill in the first place, then HT won't do much for you".
Look at the IBM POWER architecture for example... the pipeline is very short, much closer to Athlon64 than to P4, yet it uses SMT.

Anyway, none of this applies to what I said. What I said is that it's nonsense to compare SMP against SMT. Obviously two cores are better than one... But SMT is a core-specific feature, so the amount of cores has no effect on it.. Therefore comparing SMT vs SMP is silly. Completely different technologies, with completely different goals, and in a completely different pricerange.
 
Scali said:
So just because you were not formulating very clearly.... you just assume that I don't know the difference? I find that both arrogant and insulting.
You simply talked about "clustered video rendering"... and that is exactly what Renderman does. What you actually meant was not specifically rendering of video, but rather (post-)processing, and perhaps editing and/or encoding too... Am I right?
What Renderman does, is obviously called rendering, hence the name. Or are you going to argue that Pixar doesn't know the difference either?
Click to expand...
What was there to formulate? As soon as I said ADOBE PREMIERE PRO you should have realized that I was not discussing 3D rendering. Your infamiliarity with this industry is not a failure of communication on my part. Renderman is a renderer used in the 3D world. It has NOTHING to do with video editing. So when I started talking about the use of dual processors for video, and you started to talk about clustered computing for video, I explained that it was pointless, as only one part of the video workflow (RAM previews) can benefit from clustered computing.

How hard is this for you to understand? If you need me to 'clarify' any further, please let me know.
 
Morley said:
As soon as I said ADOBE PREMIERE PRO you should have realized that I was not discussing 3D rendering.
Click to expand...

I never said you were!
I was though.

Your infamiliarity with this industry is not a failure of communication on my part.
Click to expand...

Haha, the arrogance continues!
For your information, I've actually worked on developing video codecs and post-processing filters and such, as far as my "infamiliarity with this industry" goes.
Who the hell are you anyway?!
 
Scali said:
I never said you were!
I was though.
Click to expand...
I know you were, and you were talking about it as a 'rebuttal' to my comment that it was pointless for video. THAT was incorrect.


Haha, the arrogance continues!
For your information, I've actually worked on developing video codecs and post-processing filters and such, as far as my "infamiliarity with this industry" goes.
Who the hell are you anyway?!
Click to expand...
Odds are if you've seen a movie lately, our workstations were used at some point during the production process.
 
Morley said:
I know you were, and you were talking about it as a 'rebuttal' to my comment that it was pointless for video. THAT was incorrect.
Click to expand...

There you go again... It's not pointless for video, a lot of 2d/3d video effects can be accelerated with clusters, and software like Renderman does exactly that... it's pointless for your specific software/tasks. Let's face it, my statement was not incorrect, it just didn't apply to you.
Now if you weren't so self-centered, you'd realise the difference.

Odds are if you've seen a movie lately, our workstations were used at some point during the production process.
Click to expand...

And what contribution did you make to all that?
 
Scali said:
There you go again... It's not pointless for video, a lot of 2d/3d video effects can be accelerated with clusters, and software like Renderman does exactly that... it's pointless for your specific software/tasks. Let's face it, my statement was not incorrect, it just didn't apply to you.
Now if you weren't so self-centered, you'd realise the difference.
Click to expand...
Self centered? We were talking about video editing. Renderman cannot accelerate anything from an Adobe Premiere Pro workflow. Period. You even admit it above, and that's all I've been trying to say!



And what contribution did you make to all that?
Click to expand...
You asked who I was, evidently you want to make this personal. If you'd like to see my resume, I'd be glad to email it to you.
 
Morley said:
Self centered? We were talking about video editing. Renderman cannot accelerate anything from an Adobe Premiere Pro workflow. Period. You even admit it above, and that's all I've been trying to say!
Click to expand...

Nono, you said a whole lot more... about me not knowing the difference, and not being familiar with the industry, and whatever else ad-hominem I forgot.

You asked who I was, evidently you want to make this personal. If you'd like to see my resume, I'd be glad to email it to you.
Click to expand...

No, you made it personal when you started throwing the ad-hominems in my general direction. But yes, go ahead and send me your resume. It had better be good, because I don't tolerate remarks like "you don't know the difference between video editing and 3d rendering" from people who have not achieved a whole lot in that industry themselves.
 
Scali said:
Nono, you said a whole lot more... about me not knowing the difference, and not being familiar with the industry, and whatever else ad-hominem I forgot.
Click to expand...
I'm not going to get into a hissy fit with you. You said I couldn't formulate an argument. I said you were unfamiliar with the terms. If I hurt your feelings, I apologize.


No, you made it personal when you started throwing the ad-hominems in my general direction. But yes, go ahead and send me your resume. It had better be good, because I don't tolerate remarks like "you don't know the difference between video editing and 3d rendering" from people who have not achieved a whole lot in that industry themselves.
Click to expand...
ROFL, pm me your email addy.
 
One more thing people are saying the die just doubled. BUT they are about 50 million short of double on the transistor count...where did they go...if we are short 50 million then there are going to be a lot of commonality here and that sounds like ...one processor trying to tell the other one....ITS MY TURN....whoops

but back to my question why NO hyperthreading on the std editions.
IF Jonsey is correct
"I think they're having a hard time finding enough features to differentiate the EE from the regular processors."

then $500 for EE that has it enabled.So what they bragged about for so long is being dropped unless you pay double. Sounds like a BIG RIP OFF TO ME.

but what do I know...maybe another tax break or something will persuade them to give us this 50 cent mod on our $500 processors in the future.

HOPE that someone will come out with the wire wrap mod for the dual cores.

Sparks
 
Morley said:
You said I couldn't formulate an argument.
Click to expand...

No, I said you were rather unclear, that's different. Basically you're doing the same again.
Do you see what I mean?
What you're saying is not the same as what you're trying to say, I suppose. I can only respond to what you're actually saying, that's all I see.

I said you were unfamiliar with the terms.
Click to expand...

Which I am not, and I would like you to recognise this.
 
sithspit said:
OK, I don't know much about how operating systems allocate threads to available processors, but something that I think would be beneficial to real world performance on multicore processors would be something I saw in relation to the design of Sun Microsystem's multicore processors. Their design, however, required the cooperation of the operating system. In XP you can set processor affinity for applications. However, you have no control over system threads and it could be tedious to manually set affinity for every program.
Click to expand...

Sun has done a number of wonderful things that just never piqued Intel/AMD's interest.
Two essential problems is that Intel and Microsoft are 800lbs gorillas. Neither one is in a position to easily dictate terms to the other. Sun is different. The same CEO presides over the Solaris team and the UltraSPARC team. If he says "Work together", they work together. Doing things the way you described would really require a lot of low level work, not to mention certain architectural departures. AMD is closer to something like this with HyperTransport.

On top of all this, it's not cheap. Cost plays into this whole scene quite a bit. No doubt that Sun's stuff is superior, but it's so damned expensive that in many situations that you can buy 3-4 Dell servers for every 1 Sun machine, thus making up for performance delta.

And cut the flame crap before the thread gets locked. Go put some ointment on that bruised ego and let it go. (Not directed to you, sithspit)
 
sparks said:
HOPE that someone will come out with the wire wrap mod for the dual cores.

Sparks
Click to expand...

If I had to guess, I'd say they've cut HT off on the die directly, like they did with the Celeron Northwoods.
 
Morley said:
Ok, was just told April 22nd.
Click to expand...
NDA lift, or actual release date?

if thats the case, im gonna have to rethink my wanting a Venice, as thats only a few weeks away, and i still have my winchester to deal with till then...
 
lithium726 said:
NDA lift, or actual release date?

if thats the case, im gonna have to rethink my wanting a Venice, as thats only a few weeks away, and i still have my winchester to deal with till then...
Click to expand...
We're supposed to announce our product simultaneously with AMD on April 22nd. I have no idea when the NDA is lifted. Note: I don't know availability.
 
Believe it or not, Tomshardware has one of the better previews of the the new EE. They overclocked theirs to 4GHz, and they pointed out another feature of the EE: it has unlocked multipliers from 14-60. I couldn't tell if the unlocked multipliers were a new feature, like toms said, or just because the chip was an ES.
 
Jonsey said:
Believe it or not, Tomshardware has one of the better previews of the the new EE. They overclocked theirs to 4GHz, and they pointed out another feature of the EE: it has unlocked multipliers from 14-60. I couldn't tell if the unlocked multipliers were a new feature, like toms said, or just because the chip was an ES.
Click to expand...
If the EE had unlocked multipliers, that'd be enough to differentiate it from the Pentium D, and make it worth the extra $$$ for those that can afford $1000 processors. At least, that works for AMD and the Athlon FX.
Also, I agree with those that said that it's a rip for intel to put HT on only the EE. Kinda a dick move on their (intel's) part...
 
I'm just finally excited to see some real evolution again in this market...
 
Jonsey said:
Believe it or not, Tomshardware has one of the better previews of the the new EE. They overclocked theirs to 4GHz, and they pointed out another feature of the EE: it has unlocked multipliers from 14-60. I couldn't tell if the unlocked multipliers were a new feature, like toms said, or just because the chip was an ES.
Click to expand...
Are you sure they just dont have a Engineering Sample?
 
USMC2Hard4U said:
Are you sure they just dont have a Engineering Sample?
Click to expand...

As I said before, no. But they do make this statement on the last page:

Similarly, the new Extreme Edition will come with a considerably lower price/performance ratio, since its price is doubled again, while offering only the advantages of Hyper Threading and open multipliers for overclocking.

http://www.tomshardware.com/cpu/20050405/pentium_d-19.html

It's pretty clear they think unlocked multipliers are a feature, whether that's the case or not is another matter. I would think Tom's has been doing reviews long enough to know the difference between an ES and a production sample though. Also, note the maximum multiplier shown in cpu-z, 60. Seems a bit high for a normal ES. For a grand, it seems like opening up the multipliers is the least Intel could do.
 
This is very cool !!! I can see a awesomely cheap rocket for co-location gaming server for my unit.
 
Im not impressed. The P4 in general is just flawed & boring.

Im interested to see what AMD has up there sleeves, they continue to impress.
 
I'm curious how various OS's will be able to handle dual core and the dual core with HT. I thought there was a two processor limit in XP Professional. I'm uncertain if that means two cores, two physical processors, two cores but you can have HT. Anyone know which OS's will handle which combinations?
 
Rhys said:
I'm curious how various OS's will be able to handle dual core and the dual core with HT. I thought there was a two processor limit in XP Professional. I'm uncertain if that means two cores, two physical processors, two cores but you can have HT. Anyone know which OS's will handle which combinations?
Click to expand...

Windows XP licenses physical and logical processors differently. You will be able to have two cores w/HT on XP Pro, as people already use dual Xeons w/HT on XP Pro.
 
Scali said:
It doesn't work that way though.
You'll get more responsiveness from HT... but in terms of actual performance, HT doesn't do a whole lot.
Click to expand...

I know it doesnt, but i was just explaining in simple terms that intel's advantage which they had in hyperthreading would not be there in the same way as before, since AMD's main flaw would be lessened by the introduction of dual cores.

And not giving HT to the non EE P4's is just retarded. That would further cut down intel's lead in the mainstream segment as basically there would be that large inefficient pipeline with no HT of the P4 Dual Core competing with the A64 Dual core. Bad decision it really is. Intel loses out in the mainstream segment.

Scali said:
A memory controller is not that large. Besides, it doesn't have to run at such a high clockspeed. So if they can add a second core to a Prescott, surely they can add at least a memory controller to the single-core versions...
Besides, Intel already has working 65 nm samples in their lab... That should also help to put some extra transistors on those dies.
Click to expand...
But intel seems to have no plans to bring mem controllers to the P4, and i know the mem controller is not that large but it still would increase the die size considerably for the single core P4's for which u seem to recommend it. and i thought they would want to phase out single core versions slowly, why would they want to spend R&D on them further?
 
64bit_is_here said:
I know it doesnt, but i was just explaining in simple terms that intel's advantage which they had in hyperthreading would not be there in the same way as before, since AMD's main flaw would be lessened by the introduction of dual cores.
Click to expand...

So what's "not a lot"?

Hexus saw a very nice "performance" boost with Hyperthreading, so did 2CPU and even GamePC, what do you mean? 2CPU even used Linux LOL!

If this were a Gimmick, no one would care about HTT not being part of the i840. It seems like you guys want to have your cake and eat it to. On the one hand, you guys are saying "Hyperthreading is nothing worth having because it's no big deal" then OTOH, you're saying "Intel are Dicks for charging folks for it". Something ain't right here. Either it sucks and it's a Gimmick, so who cares if it's not part of 840. or; It is a useful feature and Intel is retarded for trying to charge for it.

Here

and

here

So are these tests rigged, are these respected sites lying? I've gotten use to Hyperthreading just as much as I have two monitors. It's not groudbreaking or earth shattering but it does work, well, most of the time on my rig=P

I'm not buying an Intel Dual Core. If I get an Intel Survey, I'll be sure to let them know how sleezy Pentium D XE 840 really is. If they wanted to make an XE model, maybe they should have used an L3 or Doubled the L2 Cache or something. But the 3.73GHz, something I can do myself, and the HT enabled i840 sold as Xtreme sucks, oh IMHO. I'll buy AMD to vote with my Wallet. I had planned to buy Intel.

Donnie27
 
You must log in or register to reply here.
Back
Top