Pentium 4 3.2 Extreme Edition

[Dr.2K]

Very interesting :)

http://www.anandtech.com/IT/showdoc.html?i=1870

[pointreyes]

Very interesting. On the gamer side of things this should prove to be very interesting. I wonder if the 3.2 HT EE will match or seriously exceed the Athlon64? Looks like the fall season will be a fun season for people with the pocketfuls of money. :)

[Dr.2K]

Agreed! Very interesting indeed. I will just have get my paws on one if the reviews shows that it is faster then the prescott. Hmm...will prescott be worth the wait or will this be better for my system? Decisions... :scratch: :confused:

[ThugsRook]

yes, this is very interesting :chit:

[eva2000]

woohoo my wish has come true! More cache is good :cool: :D :yummy:

[Jim]

Yummy, little more (http://www.theregister.co.uk/content/3/32863.html)

[pointreyes]

What I find very interesting is the way AMD and Intel are using the larger cache. Still one pipe on each of these procs. The AMD Opteron has three pipes but only if you use the correct motherboard that allows it. For gaming this is not important but for a server, it is very important-at least that is the gist I'm getting from watching IA64 and x86-64 development. Intel is doing a lot on the IA64 side, the Deerfield will run at 1Ghz with 1.5Meg of L3 cache. And the Tanglewood (also an IA64 proc) will have multiple cores on a single piece of silicon. Whoops, sorry I seem to have gone to the server-side. :rolleyes:

[DivideBYZero]

Originally posted by eva2000
woohoo my wish has come true! More cache is good :cool: :D :yummy:

It is good, but don't forget this is L3.

[SupermanCK]

can we say $1000+???

does L3 cache helps that much more?

now if they increase L2 cache...that would really be interesting

[pointreyes]

Originally posted by DivideBYZero
It is good, but don't forget this is L3.

Good point.

Here's a comparsion of the Opteron/Xeon/IA64-this might help see the differences between the different caches between these procs. I need to find this comparsion chart for the P4/AMD64. :)

http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_8796_8799,00.html

[pointreyes]

I did find one with the Opteron and P4 but this is an unfair comparsion to me. The Opteron should be compared to the Xeon and IA64 not the P4. I was hoping to find an Athlon64/P4 comparsion. :( Maybe on the 23rd that will happen.

http://www.amd.com/us-en/Processors/TechnicalResources/0,,30_182_865_9014,00.html

[Hawk]

Sounds like a cheaper alternative to the Prescott right now. . . .

[ThugsRook]

Originally posted by Hawk
Sounds like a cheaper alternative to the Prescott right now. . . .

i was thinking the same thing.
rather just replace a chip, then replace an entire platform again and again.

[Dr.2K]

Well it was mentioned that the price of the EE would be a bit more than the current 3.2c ($650-$680 maybe). But if the current boards out are able to support prescott at that time, then it will be a tough decision as to which way to go and then cost will only be measured in terms of the chip and not a whole platform upgrade. But, I'm thinking that intel has hedged its bet with this updated 3.2 EE cpu against the new athlon64 just in case or perhaps knowing that prescott will not be compatible with the current 875/865 mobo's.

[pointreyes]

Originally posted by Dr.2K
Well it was mentioned that the price of the EE would be a bit more than the current 3.2c ($650-$680 maybe). But if the current boards out are able to support prescott at that time, then it will be a tough decision as to which way to go and then cost will only be measured in terms of the chip and not a whole platform upgrade. But, I'm thinking that intel has hedged its bet with this updated 3.2 EE cpu against the new athlon64 just in case or perhaps knowing that prescott will not be compatible with the current 875/865 mobo's.

Personally, wouldn't it make more sense to save up enough to get the Prescott coupled to PCI Express and start taking advantage of the throughput you will get with the new bus? Note: that for 2004, that PCI-X is currently being considered faster then PCI Express until there are more 'wires' in the PCI Express bus.

2004 should be much more interesting than this year because bottlenecked pipes on the technology is finally getting addressed better. I don't see much logic in having a super fast proc that is bottlenecked with slow pipes. Haha, I took the different route, slow procs with superfast pipes. :lol: :rolleyes:

[SupermanCK]

1st Benchmark of 3.4EE and 3.2EE compare to 3.2c1 (http://www.aceshardware.com/#75000423)


As we expected the 2 MB L3-cache can not do any wonders for Games, which are somewhat cacheable but also need streams of new data (memory intensive). The best result is a 14% in Comanche, which is of course quite impressive, as it means the L3-cache pushes the 3.2 GHz to the level of a 3.7 GHz Pentium 4. Quite a few games give only a 5-6% performance increase. Whether or not the P4 3.2 GHz will be an interesting option for hard core games will depend on the pricetag.

[kONGO]

It's just proof that Intel are a bit nervous regarding the Athlon 64. That the NDA for benchmark results ends on the launch date for the Athlon 64 makes it so obvious it's silly :).

From a purely technical standpoint, the Athlon 64 looks nicer, or at least from my technical standpoint ;). On-die memory controller means fast memory access and also leaves the HT link to the AGP tunnel/southbridge a lot less crowded. The HT link still has a very high bandwidth which means it should perform nicely with the headroom AMD has given it using an on-die memory controller.

It'll be interesting for sure. Competition is always nice... Let's hope AMD gets it right this time.

-kONGO

[pointreyes]

Originally posted by kONGO
From a purely technical standpoint, the Athlon 64 looks nicer, or at least from my technical standpoint ;). On-die memory controller means fast memory access and also leaves the HT link to the AGP tunnel/southbridge a lot less crowded. The HT link still has a very high bandwidth which means it should perform nicely with the headroom AMD has given it using an on-die memory controller.

It'll be interesting for sure. Competition is always nice... Let's hope AMD gets it right this time.

-kONGO

Yep. I agree completely. :)

[SupermanCK]

they got it right with the athlon...
just wasn't right enuff...
but still good enuff to drive prices to where we can buy buy buy...
remember those days where 1GHz > $1000

[Duvie]

I kept my k6-2 for 2 years because the prices were too high...The I was able to get a 750 tbird for about 125 and the race began!!!!


I think the AMD is technically superior to the P4ee for sue...i will have to see all the advancements of the prescott but it doesn't represnet that much of a change from the p4c chips and likely wont have the guts to ramp very far...

if athlon64 can OC and come down in price I may have found a new home....The multimedia performance needs to get much better and hopefully fx chips and 2.4-2.6ghz speeds with some 64bit os and programs could help that eventually...

[ThugsRook]

review (http://www.x86-secret.com/popups/articleswindow.php?id=91)

[Duvie]

Originally posted by ThugsRook
review (http://www.x86-secret.com/popups/articleswindow.php?id=91)

Yep I read that one...I know they couldn't get the fx or the 3200+ very high oc'd, but in this earlky stage how much of that is the chipset?? They said no amount of vcore could change that which makes me think the cpu is not the limiter here...

The 103 watts at default with the prescott worries me as it really seems like the p4c's may be the best we may get for ocers for a while...

The multimedia and cadd test reiterates my response the a64 for now is a gamers delight but lacks in the MM world...

[pointreyes]

Yeah, the MM world still belongs to Intel. I was bummed to see that. Fortunately, I'm not doing MM so the Opteron route I took made the most sense.

1) The current specs on the Opteron boards are that they only have to handle a max voltage of 89w. This does not mean that the Opterons today are at 89w, it means that AMD made the specs so that people can upgrade to faster Opterons and not worry about the voltage. So far, AMD does not plan to exceed 89watts. And yet the Prescott is at 103w.

2) PCI-X, from what I can tell only the server-class motherboards are providing this wonderful technology. Even though PCI Express is coming very soon it appears that on the server side it will be PCI-X for a little while.

3) I get more 'pipes' of throughput on my 64-bit based serverboard. You won't even get that with the Athlon64 (FX).

4) L2 instead of L3 cache.

5) Finally there is a dual Opteron board that does not require ECC RAM. There should be more of these type of boards showing up shortly, this should make for some interesting competition between the Opterons and Xeons.

Duvie, for MM I can't but wonder what you would get with a dual Xeon at 800fsb. Imagine what an IWill DPI800 (there isn't one but was hoping there will be) would give ya. :)

[Dr.2K]

HardOCP Review: Athlon 64 Vs. Pentium 4

http://www.hardocp.com/article.html?art=NTI0

http://www.xbitlabs.com/articles/cpu/display/athlon64-fx51.html

[kONGO]

Originally posted by pointreyes
1) The current specs on the Opteron boards are that they only have to handle a max voltage of 89w. This does not mean that the Opterons today are at 89w, it means that AMD made the specs so that people can upgrade to faster Opterons and not worry about the voltage. So far, AMD does not plan to exceed 89watts. And yet the Prescott is at 103w.Yes, it's worrysome. Prescott's Thermal Design Power is 103W while the Opteron's maximum power (worst-case, IDD_max) is 89W, and that seems to be a global figure (for all Opterons) too. AMD has the edge in that respect. SOI is great for decreasing leakage currents and thus lessening the effect of static power dissipation.
Originally posted by pointreyes
2) PCI-X, from what I can tell only the server-class motherboards are providing this wonderful technology. Even though PCI Express is coming very soon it appears that on the server side it will be PCI-X for a little while.PCI-X is nice, but since it's never taken off as a consumer technology, it's still pricey. PCI Express is rooted in the consumer spectrum and should not be much more expensive than regular PCI when it's no longer a new technology (say 1 year after intro).
Server motherboards will use PCI Express sooner or later. Its much smaller footprint, point-to-point topology and good scaling will be something server/workstation motherboards will use without a doubt.
Originally posted by pointreyes
3) I get more 'pipes' of throughput on my 64-bit based serverboard. You won't even get that with the Athlon64 (FX).The Athlon 64 doesn't need more than one 16-bit HyperTransport link. The Opteron is designed with 3 since it uses HT between CPUs in multi-CPU environments and also because PCI-X requires a huge chunk of bandwidth when running at full steam.

Originally posted by pointreyes
4) L2 instead of L3 cache.Prescott will be 1MB L2 cache too.
Originally posted by pointreyes
5) Finally there is a dual Opteron board that does not require ECC RAM. There should be more of these type of boards showing up shortly, this should make for some interesting competition between the Opterons and Xeons.It'll still require registered RAM, though. And if you buy registered, there's no reason not to buy ECC...

-kONGO

[RotorHead]

Don’t know if you guys saw this one yet. It’s not in English but…

Looks like the P4EE is doing pretty good

Linky (http://www.x86-secret.com/popups/articleswindow.php?id=91)

RH

[pointreyes]

Originally posted by RotorHead
Don’t know if you guys saw this one yet. It’s not in English but…

Looks like the P4EE is doing pretty good

Linky (http://www.x86-secret.com/popups/articleswindow.php?id=91)

RH

Yeah, that seems to be the main site that people point to. Hopefully there will be more after this week (and in English :p ).

[DivideBYZero]

Originally posted by RotorHead
Don’t know if you guys saw this one yet. It’s not in English but…

Looks like the P4EE is doing pretty good

Linky (http://www.x86-secret.com/popups/articleswindow.php?id=91)

RH

Buy me one, then.

Bah, humbug. Intel paper launching to steal AMD's thunder.

[Dr.2K]

Well it seems that it is a paper launch al over lol. On the one end you have intel trying to one up on amd and on the other side you have amd it seems without enough supply :p. And Reviews galore everywhere! This is just gettin good :yummy:

[pastorjay]

Just posted ours not to long ago... here it is..

FX at Legit Reviews (http://legitreviews.com/Reviews/64fx_1.shtml)


PJ

[pointreyes]

Originally posted by pastorjay
Just posted ours not to long ago... here it is..

FX at Legit Reviews (http://legitreviews.com/Reviews/64fx_1.shtml)


PJ

Nice. I found the information on the CPU-Z display interesting. See how it differs from mine. Both of the displays show the procs as Sledgehammer but there is more info on my proc. Interesting.

[kONGO]

On the power issue. What I failed to include in my previous post is that comparing numbers the way I've done is very iffy. You can compare the TDP value from Intel to Intel's other CPUs, but carrying that over to compare with the AMD platform requires speculation on one level or another.

I still think the Prescott sounds too hot compared to the Opteron, but a lot of it all has to do with the software being run, how the CPU manages its power consumption and things like that (if it can shut parts of itself down, etc.).

This example is enlightening. The Opteron is hotter than the Xeon 3.06 (there is of course variables in these numbers too, but it's good for perspective).

http://www.aceshardware.com/forum?read=105039651

-kONGO

[pointreyes]

Solution 2: 66 x335 (2 x 3.06 Xeon), 2 Racks: 54,933 BTUs/hr (416 BTUs/CPU)
Solution 3: 42 e325 (2 x Opteron 246), 1 Rack: 57,322 BTUs/hr (682 BTUs/CPU)
Solution 4: 84 x335 (2 x 2.8 Xeon), 2 Racks: 69,946 BTUs/hr (416 BTUs/CPU)
Solution 5: 149 x305 (1 x 2.8 P4), 4 Racks: 69,264 BTUs/hr (464 BTUs/CPU)

Ok, I'm confused-why are the 2.8 procs so much hotter? Is it the Racks that make the difference? If so, then yes, there are other things that might have to be taken into account.