![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
-
Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.
difference between a C and E
- Thread starter ruthless-
- Start date
MemoryInAGarden
2[H]4U
- Joined
- Oct 26, 2003
- Messages
- 4,067
1MB L2 cache, SSE3 extensions, heat, longer pipelines, and more heat.
- Joined
- Mar 19, 2003
- Messages
- 11,086
2.8C > 2.8E see sig 
Aircooled and only 1.55Volts!
Aircooled and only 1.55Volts!
No it isn't. The Prescott "E" cores have longer pipelines which actually cause branch prediction issues. This is a penalty more than a help. However this move is theoretically going to help the processor scale better with higher frequencies. Also the 1MB cache in my opinion is what helps to offeset this problem.
If you actually look at the test scores you'll see that Prescott is no faster than Northwood. And in alot of cases slightly slower. However memory throughput has increased ALOT. This is probably due to the increase in L2 cache.
Prescott scores better than Northwood in very few tests. With the so called SSE3 instructions and future software optimizations Prescott will probably improve. That was how it was with SSE2. It took awhile but eventually the Northwood designed proved to be a good one. Prescott will probably too.
The same whining about it's performance occured back when the switch was made from P3 to P4. Only time will tell however. I personally wouldn't hesitate to install a Prescott in my system. However I am confident I can handle the heat problems. But right now Northwoods are a better buy. Less heat means that they are less dangerous to overclock.
BillLeeLee
[H]F Junkie
- Joined
- Jul 2, 2003
- Messages
- 13,486
The longer pipeline helps the Prescott scale to higher speeds, but since the Prescotts are same speeds as current Northwoods, this is a hinderance because you need higher speeds to take advantage of the longer pipelines. This is why the early generation Willamette P4s showed lackluster performance (I owned a P4 1.3); they weren't that much faster than the P3s of the time, but they had a 21 stage pipeline compared to the P3's 10 (?).
If longer pipelines kept translating to better performance, we'd all be using Prescotts and the Athlons wouldn't stand a chance. But as most of us know, the 10 stage Athlons are either dead even or can outperform P4s (like the Ath64 in gaming).
If longer pipelines kept translating to better performance, we'd all be using Prescotts and the Athlons wouldn't stand a chance. But as most of us know, the 10 stage Athlons are either dead even or can outperform P4s (like the Ath64 in gaming).
chrisf6969
[H]F Junkie
- Joined
- Oct 27, 2003
- Messages
- 9,011
Basic:
C = cooler - (NORTHWOOD)
E = extremely HOT - (PRESCOTT)
Advanced:
C = .13 micron, 1.525v, 512KB L2 cache, 20 stage pipe, MMX, SSE, SSE2, HT
E = .09 micron, 1.380v, 1024KB L2 cache, 32 stage pipe, MMX, SSE, SSE2, SSE3, HT + some HT improvements
The only good improvement to the PRescott is the 1Mb L2 cache.
Everything else either cant be used yet b/c of software (SSE3, HT improvement) or isnt mature yet (.09 process = too much heat)
32 stage pipeline should make it easier to reach 4+ Ghz... but the .09 process is leaking too much current/heat.
C = cooler - (NORTHWOOD)
E = extremely HOT - (PRESCOTT)
Advanced:
C = .13 micron, 1.525v, 512KB L2 cache, 20 stage pipe, MMX, SSE, SSE2, HT
E = .09 micron, 1.380v, 1024KB L2 cache, 32 stage pipe, MMX, SSE, SSE2, SSE3, HT + some HT improvements
The only good improvement to the PRescott is the 1Mb L2 cache.
Everything else either cant be used yet b/c of software (SSE3, HT improvement) or isnt mature yet (.09 process = too much heat)
32 stage pipeline should make it easier to reach 4+ Ghz... but the .09 process is leaking too much current/heat.
The Pentium III had a 10 stage pipeline.
The Willamette and the Northwood had a 20 stage pipeline.
The Prescott has a 31 stage pipeline.
Thanks to the doubled L1, doubled L2, decreased integer multiply latency, faster shift/rotate, improved HT, and improved branch predictor, the Prescott manages to perform just about the same clock-for-clock as the Northwood. All of those goodies compensate for the nastier branch misprediction penalty.
It seems that the Prescott on S478 can't scale worth a damn. I'm still waiting for my God Damned Prescott 3.4; I think on Monday I'll break down and get the Prescott 3.0. When we switch to LGA775, power distribution to the processor will be improved, and (so Intel says) frequencies will start ramping up significantly. They're still talking about a Prescott 4.0 by the end of this year. The problem is not heat - you can always slap an MCX-4000 or the LGA775 equivalent on these bastards - but power.
Also, it does appear that 90 nanometers is not the magical heat reducing process that 130 nanometers was for Intel. But recall that even though the Northwood rocked, the NV30 ran pretty hot - that's because Intel's shift to 130 was simultaneous with low-k, whereas Nvidia went to 130 without low-k. Now here we are - P4s already have low-k, and they're not going to get it again. Strained silicon helps, of course.
We're reaching the end of the silicon road. Ramping up is going to become harder and harder.
The Willamette and the Northwood had a 20 stage pipeline.
The Prescott has a 31 stage pipeline.
Thanks to the doubled L1, doubled L2, decreased integer multiply latency, faster shift/rotate, improved HT, and improved branch predictor, the Prescott manages to perform just about the same clock-for-clock as the Northwood. All of those goodies compensate for the nastier branch misprediction penalty.
It seems that the Prescott on S478 can't scale worth a damn. I'm still waiting for my God Damned Prescott 3.4; I think on Monday I'll break down and get the Prescott 3.0. When we switch to LGA775, power distribution to the processor will be improved, and (so Intel says) frequencies will start ramping up significantly. They're still talking about a Prescott 4.0 by the end of this year. The problem is not heat - you can always slap an MCX-4000 or the LGA775 equivalent on these bastards - but power.
Also, it does appear that 90 nanometers is not the magical heat reducing process that 130 nanometers was for Intel. But recall that even though the Northwood rocked, the NV30 ran pretty hot - that's because Intel's shift to 130 was simultaneous with low-k, whereas Nvidia went to 130 without low-k. Now here we are - P4s already have low-k, and they're not going to get it again. Strained silicon helps, of course.
We're reaching the end of the silicon road. Ramping up is going to become harder and harder.
The Silicon Freeway *is* running out of steam unless Intel either comes up with (or licenses) a process similar to IBM's SOI (which IBM has already licensed to AMD) or adopts Gallium Arsenide (GaAs) which has never even been demonstrated in a CPU before.
However, a lot of the current lithography processes don't translate well to non-silicon substrates (with the exception of X-ray lithography, which started with GaAs and came to silicon later) so unless Intel can scale X-ray litho to 90 nm or smaller using either GaAs, SOI, low-k strained Si, or some combo, we have seen the last major die-shrink for a while.
However, a lot of the current lithography processes don't translate well to non-silicon substrates (with the exception of X-ray lithography, which started with GaAs and came to silicon later) so unless Intel can scale X-ray litho to 90 nm or smaller using either GaAs, SOI, low-k strained Si, or some combo, we have seen the last major die-shrink for a while.
[PGHammer]
> The Silicon Freeway *is* running out of steam unless
No, period. There is an ultimate limit to how small you can make transistors on silicon. And we'll hit it, or something close to it, by 2020 or thereabouts.
> Gallium Arsenide (GaAs) which has never even been demonstrated in a
> CPU before.
The Cray-I used gallium arsenide chips, I believe.
It's not used anymore for microprocessors, though.
> we have seen the last major die-shrink for a while.
65 nm is up next, and Intel /will/ move to it. I forget what year they say they're going to do it, but they're obsessed with shrinking.
> The Silicon Freeway *is* running out of steam unless
No, period. There is an ultimate limit to how small you can make transistors on silicon. And we'll hit it, or something close to it, by 2020 or thereabouts.
> Gallium Arsenide (GaAs) which has never even been demonstrated in a
> CPU before.
The Cray-I used gallium arsenide chips, I believe.
It's not used anymore for microprocessors, though.
> we have seen the last major die-shrink for a while.
65 nm is up next, and Intel /will/ move to it. I forget what year they say they're going to do it, but they're obsessed with shrinking.
hey, by that time, how fast would you guess processors will be? I always wondered about this. If it could just keep going and going and going. How could it never end, that would be crazy man. So by 2020 how fast in terms of ghz. do you think we'll be up to?
Black Morty Rackham
Supreme [H]ardness
- Joined
- Jan 8, 2004
- Messages
- 4,728
Even Intel has realized you cannot just keep upping the clock frequencies to make faster processors, so the clock frequency of a top-of-the-line processor in 2020 isn't all that interesting...