Intel Prescott EE Thread
Straight_Man
Geeky, in my own wayNaples, FL Icrontian
Prescott EE, per se, is a transitional chip. Pentium M, the new ones, are a combo of a processor that scales back when OH'd and uses a Prescott Pipe core processing structure.
It hates 16 bit code and preSSE ALU calcs, it is tuned for 32 and 64 bit code. Zeons are coming online with basically two Prescott\Zeon blended structure pipes per CPU core, they will be on open market shortly. Two dual-pipe ZEONS on the newest ZEON HT boards outperform Prescott by huge amounts. Better proven core design than Prescott has.
I can give links if wanted, each will have some of what I say, and expect chipset and CPU type docs from Intel's dev area to be part of links. Tom's Hardware will have an article on this, so I am starting a processor type thread here that can go wide-open within type and comparatively. This is not a pure 32 bit CPU.
It hates 16 bit code and preSSE ALU calcs, it is tuned for 32 and 64 bit code. Zeons are coming online with basically two Prescott\Zeon blended structure pipes per CPU core, they will be on open market shortly. Two dual-pipe ZEONS on the newest ZEON HT boards outperform Prescott by huge amounts. Better proven core design than Prescott has.
I can give links if wanted, each will have some of what I say, and expect chipset and CPU type docs from Intel's dev area to be part of links. Tom's Hardware will have an article on this, so I am starting a processor type thread here that can go wide-open within type and comparatively. This is not a pure 32 bit CPU.
0
Comments
I've _got_ a Prescott. It's a piece of crap. It's not a slow cpu, and it overclocks well, but it's a piece of crap. It's just like anything else- there's a right way to do something, and a wrong way to do something. With the P4, and the Prescott in particular, Intel did it the wrong way. Sacrificing efficiency for speed is a Bad Thing.
Furthermore, the P4 is no more stable at high temperatures than the athlon is (which you've said before). Your P4 board(s) is (are) just reading the temperature wrong, I'm sure. Mine does. The 2.8E idles at 30*C with a POS Thermaltake heatsink. That's a far cry from the 53*C the BIOS registers, and it explains why the heatsink doesn't even get warm, and the back of the board only gets lukewarm. The fact of the matter is that the Prescott puts out more heat than the Athlons, and it's equally unhappy at temperatures above 65*C. It just has clock throttling to prevent any serious damage. That is the only advantage it has as far as heat-related things go.
The bottom line (and this is coming from someone who's got a P4-E, a P4 laptop, and a dual s604 board waiting for 2 xeons) is that the P4 and P4-based CPUs are Crap.
My HSs are HOT. This is not just BIOS temps I am talking. CPU is not throttling at all. but this CPU IS NOT pure 32 bit. Period. It is a transitional.
No.
If it can run it, it's a 64 bit chip. It can't, so it's not. Period.
As for the heat thing, I'll have to play with mine when I get the board fixed. In the meantime, I can tell you that at idle, the cpu does not run very hot at all. When it's at full load is another matter, of course, but the heatsink that Intel packaged with it was an aluminum/copper hybrid, so...
I'll measure the temperature under load after I get the board fixed. I suspect it'll be under 65*C
//EDIT:
The new Pentium M is not at all based on the Prescott pipe. It's a 90nm Banias architecture with increased cached. In fact, it's more like an Athlon.
Given that 99% of the app code out there in use is 32 bit, and given that Prescott literally has two simultaneous 32 bit processing pipes that are NOT run at decreased speed each, what is the net processing capability of an Opteron single pipe Athlon Fx compared to a dual pipe Prescott??? For production use running a business??
Why can Prescott fold and yield 1.7X-1.8X what my slightly faster clocked Non-HT Northwood with twice the RAM amount clocked totally the same does while Prescott runs more apps??? What is the NET effect of two processor pipes tuned specifically for the code out there versus a processor that has to translate to handle 32 bit in order to process it through a native 64 bit pipe, or process half-words with the code now available to do most things??? The 64 bit XP and Linux gens come with and have available for them, mostly 32 bit code applications--Longhorn will have 64 bit apps with it. Intel's 64 bit compiler, while it was still in final testing, was used to write these O\Ss in large part. Crap, my EE spins wheels around my Northwood and the dang Northwood has more RAM available at same exact timings as the Prescott's RAM has for it. Prescott, at 3 GHz, is happiest under so close to full load that it is not funny-- it typically is under 99.3% load, and is more stable that way than my Barton or My Northwood under same load. Unless Prescott makes XP run out of RAM, and at half a gig that takes some doing and ram load is about 60% average, XP Pro SP1 plus security packs is happy as heck to run it full blast. Oh, it helps that I am NOT runnign relaxed timing JEDEC RAM. I am running low latency (intended for Opteron use) at 2.5,3,3,7 on both boxes at DDR400 rates. Both boxes can run weeks at a time.
IS 32+32 simultaneously run close to equal to 64 given 32 bit code??? In performance??? Its about 90% of the way there. My EE ran $179.95 shipped. Crap, no, not as good as Opteron will be when the code is out. Transitional processor. Waiting on large amounts of code rewrites before 64 bit truely takes hold-- in about 2007 given Microsoft's longhorn projections. Hint, apply Missleman's HT patch archive to XP pro SP1 after gettignt eh latest inf files for Prescott, on your Prescott box. Even the HT drivers are not fully out yet, they will be in SP2 final when that hits.
All CPUs have dual 32 bit processing pipes. This is how <b>EVERY DDR BOARD IN EXISTENCE HAS A 64 BIT BUS</b>. The Opteron isn't single pipe.
Because the 1.7x-1.8x are derived from the Prescott's hyperthreading capability. You do understand that it's the slim basics of a dual-core system, and responds quite well to folding, therefore has almost but not exactly twice the performance of a non-HT CPU for folding? Apparently you don't understand this.
Invalid question. There is no such CPU currently marketed that must do this beyond the Itanium. The Itanium is the only pure 64 bit chip out there, and it is the ONLY CHIP that must translate 64 bit code to 32bit via emulation to run it. AMD's X86-64 architecture can interpret 32bit and 64bit at the same time, because 32/64bit is NATIVE to the pipe. There is no translation.
We're not talking about this. Don't mention things that are irrelevant, and that we're not talking about.
You don't have an EE, you have a Pentium 4-E (Prescott). The EE is entirely different from the P4E. The RAM means little to <b>nothing</b> in folding. But this is entirely irrelevant to our discussion your botchup of understanding what a Pentium M is. So stop mentioning this too.
You have no statistical evidence to support your stability claims.
1. We're not talking about RAM
2. We're not talking about SP1
2. 2.5/3/3 isn't low latency.
This question has miserable syntactical elements to it. So I won't touch it. Suffice it to say, today's 64 bit chips and their accompanying 32bit code operation is significantly superior to yesterday's 32bit engines (Pentium 4, Athlon XP, etc.).
Again, you don't have an EE. And it isn't a transitional processor, it was supposed to be a complete evolution, but it's a miserable ****up.
Irrelevant. Don't mention it.
You don't seem to know what the difference is between a 64bit bus via 2 32bit pipes and a 64 bit CPU is?
Let's make this very clear:
1. We're not talking about Linux or 64 Bit windows
2. We're not talking about Intel's 64 bit compiler
3. We're not talking about Longhorn
4. You don't have an EE. You have a Pentium 4-E, which is a prescott. An EE is an entirely different chip. You don't even know what you bought?
5. We're not talking about security packs.
6. 2.5/3/3 isn't low latency
7. We're not talking about stability.
You're eventually going to break someone's computer by appearing to know what you're talking about, when in reality, you're 95.5623454% wrong. They're going to take your suggestion, implement it, and something is going to go seriously wrong.
P4EE: P4 Extreme Edition (Gallatin core, 130nm)
but i suppose that IS how you put up a good amd fight, by knowing the cons of each
but i suppose that IS how you put up a good amd fight, by knowing the cons of each