anandtech Prescott review NOW UP! (anand inside)

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

FishTankX

Platinum Member
Oct 6, 2001
2,738
0
0
Originally posted by: Tiorapatea
From the horse's mouth, Prescott specifications and thermal design considerations...

Specs

Thermal and mechanical design

I wish I could understand a little more in these documents. It seems to me like Prescott, well, uh, runs a little hot and, like, there's lots of stuff you have to do to keep it happy. Also, it looks like 2.8A (the 533 FSB proc.), 2.8E and 3.0E aren't designed with the same electrical specs as 3.2E and 3.4E.

Anyone care to explain how VID, Vccmax, Icc etc. move around and interact when overclocking. Am I correct in saying that with these newer multiple VID processors, you want to get lucky with a low voltage chip for its rated speed to give you more headroom to overclock?

References please in layman's terms as to how you stay in touch with Intel's design specs for processors (more or less) while o'cing. I mean I keep reading about upping the FSB and voltage and waiting to see if your system crashes but isn't there a slightly more scientific way of estimating what you might get out of a CPU (without frying it) ex ante., i.e. along the lines of saying: ok, I have a 2.4C, M0 stepping with VID at 1.425V so if I overclock by amount x, increasing voltage by delta, and then take y and z measurements, then I'll have an idea that the CPU could maybe go as far as [some extrapolated GHz] or I can at least predict some safe further increment to the overclock/voltage while staying within Intel's thermal design guidelines? Hope this makes sense.

No. Supply and demand could easily force chips that normally performed much higher, into a lower bin, because of demand. Or demand for a higher clocked processor could have lower binned processors capable of higher speeds put up into the upper speed bracket.
 

formulav8

Diamond Member
Sep 18, 2000
7,004
522
126
Watch everyone still put AMD in the boiler category. I don't believe AMD has ever made a CPU with those heat numbers. But everyone will still refer to AMD as the boiler chips.


Jason


Originally posted by: Jeraden
According to the review over at Sudhian, check out these chip temps (using stock cooling):
http://www.sudhian.com/showdocs.cfm?aid=489&pid=1815

Athlon 64 FX-51
Idle: 39C
Full Load: 49C

P4 3.4 GHz EE
Idle: 36C
Full-Load: 61C

Prescott
Idle: 53C
Full-Load: 71C

 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Intel is officially the heat king.

I still strongly believe that temps would be much lower (~50c load) with a high end heatsink.
 

jdogg707

Diamond Member
Jun 24, 2002
6,098
0
76
Originally posted by: Acanthus
Intel is officially the heat king.

I still strongly believe that temps would be much lower (~50c load) with a high end heatsink.

Isn't that the point of buying a high-end heatsink?

It breaches into the "duh" category...haha
 

sellmen

Senior member
May 4, 2003
459
0
0
Originally posted by: Acanthus
Intel is officially the heat king.

I still strongly believe that temps would be much lower (~50c load) with a high end heatsink.

Maybe, but those temps are still pretty absurd.

The stock Intel cooler for 3.0+ GHZ Pentium 4 processors is nothing to scoff at; it is large, has a copper contact area, and has a large fan running at high RPM. Those things are loud...and the idle temp for prescott is still 54 degrees?! The Athlon 64 line runs at a lower load temp than the prescotts idle temp, using a cooler that's 1/2 the size and much quieter.

The thermal difference between A64 and Prescott is huge.
 

zephyrprime

Diamond Member
Feb 18, 2001
7,512
2
81
This is a beef I have with the A64. The 128K L1 has been the baseline since the original K7 core. Its L1 design has now survived several process and architecture changes, and its obvious that the L1 is not efficient enough for the A64 - the larger L2 caches in A64 makes a noticeable difference in benchmarks that have even relatively small data sets. Nor is A64's L2 cache, with its 64-bit pathway, an improvement over old K7 technology. AMD should have moved to 192K or 256K L1, or even a trace cache, when they released the Opteron.
You can't just increase cache size like that without a cost. Bigger cache means higher latency and latency is a killer. Look at how much the latency has increased with the Prescott and its larger cache. In the future, I think L1 & L2 caches will be essentially the same size as now but we may have more cache layers. L1-L4 maybe?
 

GNY

Member
Sep 1, 2000
93
0
0
Looks like a Job for a SP94 and a Vantec Tornado to cool these baby's down !
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Originally posted by: jdogg707
Originally posted by: Acanthus
Intel is officially the heat king.

I still strongly believe that temps would be much lower (~50c load) with a high end heatsink.

Isn't that the point of buying a high-end heatsink?

It breaches into the "duh" category...haha

Of course its a duh category... but people are flipping out about the high temps. Thats with a crappy stock cooler, spend $40 on a good aftermarket and it wont run that hot.
 

zShowtimez

Senior member
Nov 20, 2001
544
0
76
Personally it looks really tempting to me, as an overclocker. If the 2.8s can hit 4 on a regular basis, im gonna jump all over it.
 

ST4RCUTTER

Platinum Member
Feb 13, 2001
2,841
0
0
I think overclockers.com pretty much hit the mark when they said, "31-stage pipelines do you no good if the processor melts before you can take full advantage of it."

With thermal characteristics like this Intel will have their hands full trying to improve the process and take advantage of that 31-stage pipe. I sure hope it drops the Northie prices though!
 

cqcc

Member
Sep 26, 2003
41
0
0
Let me tell u this pal, is it EXTREMELY impressive to increase the pipline 20% from AthlonXP's 10 stages to Athlon64's 12 stages and totally BEAT whatever Prescott and Notthwood P4 in most benchmarks while stupid Prescot only can catch up with Northwood??

Originally posted by: Acanthus
Originally posted by: Pandaren
Performance is pretty mediocre. I've looked at the AnandTech benches as well as the benches from Tom's and Xbitlabs.

Clock for clock performance seems roughly equivalent. Prescott's technological tricks mainly keep the IPC from taking a big hit from the 31-stage pipeline.

OEMs will like the chip because they won't have to pay extra $ for it. I think Intel knows enthusiasts won't be happy with the numbers, and this is why it will take time for the CPU to ship in volume to resellers.

It is EXTREMELY impressive to increase the pipeline 55% and keep performance the same. I'm going to wait for the chip to get ironed out then get a good step 3.2ghz to run at 4.0ghz on default voltage :D just like i did with northwoods.

:beer::clock:
 

Eug

Lifer
Mar 11, 2000
24,048
1,679
126
Hmmm... Even after the 0.09 um shrink, Intel still holds the crown for this list. ;)

It will be interesting to see what AMD gets when they go 0.09 um SOI this year.
 

MadRat

Lifer
Oct 14, 1999
11,971
287
126
It would be interesting to see how quick these puppies burn out due to fatigue in the interconnects. Remember how sensitive T-Birds were to hot and cold cycles?
 

dawks

Diamond Member
Oct 9, 1999
5,071
2
81
I have yet to read through the entire thread, but I did read the entire article.

I have one question. At one point in the article (mid-way through page 9) it says:
The LDDQU instruction is one Intel is particularly proud of as it helps accelerate video encoding and it is implemented in the DivX 5.1.1 codec. More information on how it is used can be found in Intel?s developer documentation here.
But theres no link. And immediatly following it goes on to talk about Developers opinions on the new instructions. I guess I could go over to Intel and look for it myself, but their site is so bloody big and cluttered, and poorly designed.. :)

Thanks Anand! Your writing is easy and enjoyable to read! :)
 

Acanthus

Lifer
Aug 28, 2001
19,915
2
76
ostif.org
Originally posted by: cqcc
Let me tell u this pal, is it EXTREMELY impressive to increase the pipline 20% from AthlonXP's 10 stages to Athlon64's 12 stages and totally BEAT whatever Prescott and Notthwood P4 in most benchmarks while stupid Prescot only can catch up with Northwood??

Originally posted by: Acanthus
Originally posted by: Pandaren
Performance is pretty mediocre. I've looked at the AnandTech benches as well as the benches from Tom's and Xbitlabs.

Clock for clock performance seems roughly equivalent. Prescott's technological tricks mainly keep the IPC from taking a big hit from the 31-stage pipeline.

OEMs will like the chip because they won't have to pay extra $ for it. I think Intel knows enthusiasts won't be happy with the numbers, and this is why it will take time for the CPU to ship in volume to resellers.

It is EXTREMELY impressive to increase the pipeline 55% and keep performance the same. I'm going to wait for the chip to get ironed out then get a good step 3.2ghz to run at 4.0ghz on default voltage :D just like i did with northwoods.

:beer::clock:

Ok "pal" but did intel double the number of registers on the chip? implement SOI? on die memory controller?

Increasing the pipeline by 20% doesnt allow them much room for clocking, however, SOI technology has allowed the chip to maintain far cooler operation. We will see how much smack you talk on prescott as it ramps up in clockspeed ;)
 

FishTankX

Platinum Member
Oct 6, 2001
2,738
0
0
Originally posted by: Acanthus
Originally posted by: cqcc
Let me tell u this pal, is it EXTREMELY impressive to increase the pipline 20% from AthlonXP's 10 stages to Athlon64's 12 stages and totally BEAT whatever Prescott and Notthwood P4 in most benchmarks while stupid Prescot only can catch up with Northwood??

Originally posted by: Acanthus
Originally posted by: Pandaren
Performance is pretty mediocre. I've looked at the AnandTech benches as well as the benches from Tom's and Xbitlabs.

Clock for clock performance seems roughly equivalent. Prescott's technological tricks mainly keep the IPC from taking a big hit from the 31-stage pipeline.

OEMs will like the chip because they won't have to pay extra $ for it. I think Intel knows enthusiasts won't be happy with the numbers, and this is why it will take time for the CPU to ship in volume to resellers.

It is EXTREMELY impressive to increase the pipeline 55% and keep performance the same. I'm going to wait for the chip to get ironed out then get a good step 3.2ghz to run at 4.0ghz on default voltage :D just like i did with northwoods.

:beer::clock:

Ok "pal" but did intel double the number of registers on the chip? implement SOI? on die memory controller?

Increasing the pipeline by 20% doesnt allow them much room for clocking, however, SOI technology has allowed the chip to maintain far cooler operation. We will see how much smack you talk on prescott as it ramps up in clockspeed ;)

Ondie mem conrollers tie Intel to one memory tech. Remember the i820? *chuckles*

 

mchammer187

Diamond Member
Nov 26, 2000
9,114
0
76
prescott now is kinda meh but i dont think that Athlon 64's have truly one - upped Northwoods yet

I was far more disappointed with the Williamette PIVs than I am with prescott