Help choosing a chip for Asus P4S533.

[Mediawaves]

I am building a watercooled rig w/mods. I have heard some bad things about certain (Willamette/Northwood) chips, is there a particular chip that would be very good for overclocking this board?
Although I have built the rigs you see below, the P4 is new to me. And overclocking anything is really new to me.

[JaY_III]

you wana a Northwood for sure...
and the best CPU toget would be the slowest speed C1 stepping you can find.

2.4Ghz is about as low as you can currently find (unless you wana rip apart a dell)

[Mediawaves]

Thank you, can I just walk into my local dealer and tell him I only want a Northwood 2.4 gig pentium?

[JaY_III]

all 2.0+GHz P4's are Northwoods....
so no problem getting one.
It is the C1 stepping that may be a little harder todo.

But if they let you look at the BOX, it is easy to pick a C1 out.
They have defualt Vcore of 1.525V.
The ones you dont want have a defualt Vcore of 1.5V, they are the older B0 stepping.

[larva]

Usually the P4S533 won't run stably past 150MHz fsb. I have one. The same cpu and ram that would only run at 150fsb on the P4S533 ran fine at 162MHz fsb on the Abit BD7-II I bought to replace it. As such you want a higher multiplier than the 1.8a/2.4b. They have 18 multiplier, which yields 2700MHz with a 2.4b at 150fsb. This is not much for a watercooled rig. I would recommed a chip with a higher multiplier, like a 2.0a. These are showing up in C1 stepping now, and at 150fsb you would have 3000MHz if the chip will do it. You could also use a 2.53, which has a multiplier of 19, yielding 2850MHz at 150fsb.

Your motherboard has settings for up to 166fsb, but only about 1 out 10 operates stably above 150MHz.

[Ec]-[oMaN]

also,very important u will need a mobo with pci/agp lock if u want high fsb

[larva]

Not necessarily true. While the 845e/g/pe/ge chipsets provide an AGP/PCI lock it is not what allows them to out-fsb the SiS 645dx boards like the P4S533. I can lock my PCI at 44MHz and it works fine. The P4S533 is only running 37.5MHz at 150fsb. The chipset is just routinely unstable at higher rates. PCI speed is usually not the devil people make it out to be when looking for an easy answer as to why they can't clock higher. Unless you have a SCSI controller in the machine, the limits almost always stem from other sources like the chipset, CPU, or ram. People would rather blame the PCI speed than admit their chip or ram is at its limits, or the motherboard they chose becomes unstable at lower clock rates than they had hoped for.

The later SiS 648 chipest (P4S8X) has a PCI/AGP lock, but has the same (actually even more) difficulty as the unlocked 645 exceeding 150fsb.

[lutjens]

Skip the SiS chipset and get a motherboard with an Intel chipset on it. Go with the 845PE or the upcoming Granite Bay (E7205 chipset).

Intel chipsets have better overclocking, better stability, better engineering and are an overall better product.

If you want a SiS or VIA chipset, get an Athlon.:)

[Mediawaves]

I took the P4S533 back and exchanged it for a P4PE, I hope this board is better for overclocking.
I plan to put a CORSAIR MEMORY XMS Extreme Memory Speed Series 512MB 32MX8 PC-2700C2PT on the board.
Does anyone know anywhere to get this for less than newegg?

[larva]

Good move on the motherboard. Spend the extra few bucks on Corsair XMS PC3200C2 ram, or just get the Kingston Value Ram PC2700. These modules use the 6ns Winbond chips and will give you the overhead to maximize your FSB while running the 3:4 cpu:memory ratio. This is an important factor. A 1.8a or 2.4b is very nice in combination with the P4PE and the Kingston or Corsair 3200 ram.

I built a P4PE machine with a late model 2.4b and the Kingston ram and it did 162fsb at 3:4 for 2.9GHz. This chip would do 3GHz, but I had to drop the multiplier down to avoid the RAM's limts, resulting in a slower system. Spending the extra cash on Corsair's handpicked version of the ram may allow the extra clock speed to reach 3GHz with the 3:4 enabled, should your CPU have it in it. If you want to save some money, get the Kingston PC2700 instead of the Corsair 2700. It's better.

[Mediawaves]

Would there be another processor that would give me possibly higher speeds such as a 2.8, sorry I don't know too much about the multipliers yet either. Thanks for all your help.

[larva]

One clarification on my post above. The Corsair XMSPC3200C2 ram is the one that uses the same chips as the Kingston Value Ram PC2700. I would get one of those two, with the Corsair version having handpicked chips that usually yield a bit more clock speed. But it is 60 bucks more. Under no circumstance would I get the Corsair 2700, unless I had a stick already I was trying to match.

As far as the CPU goes, it's hard to say definately. CPU's are all individuals, and you can't state what results you will get from the particular example you will end up with. Again for the sake of clarity, let me point out that there are two multipliers to be considerd. One is the fixed multiplier of the CPU, determining the ratio between the board speed and the chip's clock rate. The second is the memory multiplier, or more accurately the ratio between the board speed and the memory's base frequency (which is effective doubled by the properties of DDR). Lets look at a few cases:

1.8a/2.4b's have a 18 multiplier. Late BO stepping 1.8a's usually deliver 2.8GHz, with the best ones achieving 3GHz. Most late B0 2.4b's will get to 3GHz. With our 18 multiplier, this means that 162fsb yields 2.9GHz, and 167 gets us all the way to 3GHz. Since we would like to use the 845pe's 3:4 memory ratio (1.33 multiplier), this coincides pretty closely with our ram's realistic clock rates. 162 fsb at 3:4 yields 216MHz, also termed a 432MHz DDR effective data rate. The inexpensive Kingston ram will do this, plus or minus 2MHz on the fsb. Going up to a `167 fsb at 3:4 yields 222MHz, for 444MHz DDR EDR. This is a touch beyond the Kingston's capability at 2.7V, but the P4PE can put more voltage on the ram than my BD7-II can. This may allow the Kingston to get to 444MHz. The Corsair XMS PC3200C2 is a safer bet if you just have to have 3GHz. Bear in mind that only benchmarks can tell the difference between 2.9 an 3GHz.

If we look at a 2.8, the CPU multiplier is now 21. While getting a 2.8 chip may give you 3.1-3.2GHz capability, some also only reach 3GHz. In any even the multiplier is too high for 845 based systems. Even if we get an outstanding 3.2GHz, this dictates a fsb of only ~152MHz. While easy on the ram, this will create a performance level virtually indistinguishable from a 2.9GHz machine utilizing a fsb 10MHz faster. And if our 2.8 is just average and only does 3GHz, our fsb drops another 10MHz to ~142.

It is true that buying a 2.8 may get you a tad better core than the 2.4b, and the 2.4b may get you a touch better example than a 1.8a. But this is not certain. Stepping up to a 2.4b has a certain attraction as they are commonly C1 steppings now, and don't penalize in terms of the multiplier. But early C1 2.4b's have been dissapointing, many yielding less than the 3GHz the latest B0 2.4b's produce more often than not. So in the end a little luck is a lot more valuable than a 2.8 in terms of the performance envelope. I don't recommend the 2.8 because of the increased cost and the unfavorable muliplier. Use the extra money on good ram, video card, power supply, and hard drive, as investment in these areas will give you better return on your money.

This is the biggest challenge for Intel systems. We must figure out what clock rate the chip will yield and combine this assumption with the fsb and memory limitations in order to pick an optimal chip. As previously mentioned, CPU's are individuals so there is no guarantee our assumption will prove 100% correct. As always in overclocking we must either try many examples or just get lucky on the first to produce all the overclock one might want. It is better to make the best educated guess we can and realize the difference between 2.8GHz and 3.2GHz is tiny. It is better to error on the side of caution and make sure our multiplier is low enough to maximize fsb and memory considerations, secure in the knowledge that they are more important than whether or not we can boast of 3GHz or a bit more.

[Mediawaves]

Thanks for the hand up, out of the darkness of my ignorance. Your the best, Larva. I understand it much better when explained to me like that, rather than trying to get it from a bunch of figures on a page. Thanks for the practical knowledge. I really appriciate that you take the time to explain stuff to new people like myself, without men like you - sites like this would'nt exist.

[Rave]

Originally posted by larva
Not necessarily true. While the 845e/g/pe/ge chipsets provide an AGP/PCI lock it is not what allows them to out-fsb the SiS 645dx boards like the P4S533. I can lock my PCI at 44MHz and it works fine. The P4S533 is only running 37.5MHz at 150fsb. The chipset is just routinely unstable at higher rates. PCI speed is usually not the devil people make it out to be when looking for an easy answer as to why they can't clock higher. Unless you have a SCSI controller in the machine, the limits almost always stem from other sources like the chipset, CPU, or ram. People would rather blame the PCI speed than admit their chip or ram is at its limits, or the motherboard they chose becomes unstable at lower clock rates than they had hoped for.

The later SiS 648 chipest (P4S8X) has a PCI/AGP lock, but has the same (actually even more) difficulty as the unlocked 645 exceeding 150fsb.

I have an ASUS P4S533-E and it is fine to run at 160 FSB completely stable with the p4 2.4b giving 2880. Can run at 166FSB but RAM is funny about it and will do 400 in 160FSB but will not do 400 in 166FSB so ram has to be backed off to 333 which is not an option due to the bandwidth requirements of the p4.

All in all I am pretty happy with this MB for the price it does a great job.

[Rave]

Originally posted by larva
One clarification on my post above. The Corsair XMSPC3200C2 ram is the one that uses the same chips as the Kingston Value Ram PC2700. I would get one of those two, with the Corsair version having handpicked chips that usually yield a bit more clock speed. But it is 60 bucks more. Under no circumstance would I get the Corsair 2700, unless I had a stick already I was trying to match.

As far as the CPU goes, it's hard to say definately. CPU's are all individuals, and you can't state what results you will get from the particular example you will end up with. Again for the sake of clarity, let me point out that there are two multipliers to be considerd. One is the fixed multiplier of the CPU, determining the ratio between the board speed and the chip's clock rate. The second is the memory multiplier, or more accurately the ratio between the board speed and the memory's base frequency (which is effective doubled by the properties of DDR). Lets look at a few cases:

1.8a/2.4b's have a 18 multiplier. Late BO stepping 1.8a's usually deliver 2.8GHz, with the best ones achieving 3GHz. Most late B0 2.4b's will get to 3GHz. With our 18 multiplier, this means that 162fsb yields 2.9GHz, and 167 gets us all the way to 3GHz. Since we would like to use the 845pe's 3:4 memory ratio (1.33 multiplier), this coincides pretty closely with our ram's realistic clock rates. 162 fsb at 3:4 yields 216MHz, also termed a 432MHz DDR effective data rate. The inexpensive Kingston ram will do this, plus or minus 2MHz on the fsb. Going up to a `167 fsb at 3:4 yields 222MHz, for 444MHz DDR EDR. This is a touch beyond the Kingston's capability at 2.7V, but the P4PE can put more voltage on the ram than my BD7-II can. This may allow the Kingston to get to 444MHz. The Corsair XMS PC3200C2 is a safer bet if you just have to have 3GHz. Bear in mind that only benchmarks can tell the difference between 2.9 an 3GHz.

If we look at a 2.8, the CPU multiplier is now 21. While getting a 2.8 chip may give you 3.1-3.2GHz capability, some also only reach 3GHz. In any even the multiplier is too high for 845 based systems. Even if we get an outstanding 3.2GHz, this dictates a fsb of only ~152MHz. While easy on the ram, this will create a performance level virtually indistinguishable from a 2.9GHz machine utilizing a fsb 10MHz faster. And if our 2.8 is just average and only does 3GHz, our fsb drops another 10MHz to ~142.

It is true that buying a 2.8 may get you a tad better core than the 2.4b, and the 2.4b may get you a touch better example than a 1.8a. But this is not certain. Stepping up to a 2.4b has a certain attraction as they are commonly C1 steppings now, and don't penalize in terms of the multiplier. But early C1 2.4b's have been dissapointing, many yielding less than the 3GHz the latest B0 2.4b's produce more often than not. So in the end a little luck is a lot more valuable than a 2.8 in terms of the performance envelope. I don't recommend the 2.8 because of the increased cost and the unfavorable muliplier. Use the extra money on good ram, video card, power supply, and hard drive, as investment in these areas will give you better return on your money.

This is the biggest challenge for Intel systems. We must figure out what clock rate the chip will yield and combine this assumption with the fsb and memory limitations in order to pick an optimal chip. As previously mentioned, CPU's are individuals so there is no guarantee our assumption will prove 100% correct. As always in overclocking we must either try many examples or just get lucky on the first to produce all the overclock one might want. It is better to make the best educated guess we can and realize the difference between 2.8GHz and 3.2GHz is tiny. It is better to error on the side of caution and make sure our multiplier is low enough to maximize fsb and memory considerations, secure in the knowledge that they are more important than whether or not we can boast of 3GHz or a bit more.

Completely agree great post... should be sticky!

[larva]

Originally posted by Rave


I have an ASUS P4S533-E and it is fine to run at 160 FSB completely stable with the p4 2.4b giving 2880. Can run at 166FSB but RAM is funny about it and will do 400 in 160FSB but will not do 400 in 166FSB so ram has to be backed off to 333 which is not an option due to the bandwidth requirements of the p4.

All in all I am pretty happy with this MB for the price it does a great job.

If mine had done 160 I wouldn't have replaced it. But mine would only do 150. Yours is one of the few I've seen get to 160. Only a handful get to 166. All good 845e and pe's get to 175+, and the 3:4 cpu:memory ratio works out better than the 4:6 of the SiS.

I do like the P4S533, bot mine (and most) just turns into a pumpkin at 151fsb. I put the bone cheapest 845e board in it I could find and lo and behold the same chip did 162fsb. What we will do for 12MHz of fsb :D

[Mediawaves]

Originally posted by larva
One clarification on my post above. The Corsair XMSPC3200C2 ram is the one that uses the same chips as the Kingston Value Ram PC2700. I would get one of those two, with the Corsair version having handpicked chips that usually yield a bit more clock speed. But it is 60 bucks more. Under no circumstance would I get the Corsair 2700, unless I had a stick already I was trying to match.

As far as the CPU goes, it's hard to say definately. CPU's are all individuals, and you can't state what results you will get from the particular example you will end up with. Again for the sake of clarity, let me point out that there are two multipliers to be considerd. One is the fixed multiplier of the CPU, determining the ratio between the board speed and the chip's clock rate. The second is the memory multiplier, or more accurately the ratio between the board speed and the memory's base frequency (which is effective doubled by the properties of DDR). Lets look at a few cases:

1.8a/2.4b's have a 18 multiplier. Late BO stepping 1.8a's usually deliver 2.8GHz, with the best ones achieving 3GHz. Most late B0 2.4b's will get to 3GHz. With our 18 multiplier, this means that 162fsb yields 2.9GHz, and 167 gets us all the way to 3GHz. Since we would like to use the 845pe's 3:4 memory ratio (1.33 multiplier), this coincides pretty closely with our ram's realistic clock rates. 162 fsb at 3:4 yields 216MHz, also termed a 432MHz DDR effective data rate. The inexpensive Kingston ram will do this, plus or minus 2MHz on the fsb. Going up to a `167 fsb at 3:4 yields 222MHz, for 444MHz DDR EDR. This is a touch beyond the Kingston's capability at 2.7V, but the P4PE can put more voltage on the ram than my BD7-II can. This may allow the Kingston to get to 444MHz. The Corsair XMS PC3200C2 is a safer bet if you just have to have 3GHz. Bear in mind that only benchmarks can tell the difference between 2.9 an 3GHz.

If we look at a 2.8, the CPU multiplier is now 21. While getting a 2.8 chip may give you 3.1-3.2GHz capability, some also only reach 3GHz. In any even the multiplier is too high for 845 based systems. Even if we get an outstanding 3.2GHz, this dictates a fsb of only ~152MHz. While easy on the ram, this will create a performance level virtually indistinguishable from a 2.9GHz machine utilizing a fsb 10MHz faster. And if our 2.8 is just average and only does 3GHz, our fsb drops another 10MHz to ~142.

It is true that buying a 2.8 may get you a tad better core than the 2.4b, and the 2.4b may get you a touch better example than a 1.8a. But this is not certain. Stepping up to a 2.4b has a certain attraction as they are commonly C1 steppings now, and don't penalize in terms of the multiplier. But early C1 2.4b's have been dissapointing, many yielding less than the 3GHz the latest B0 2.4b's produce more often than not. So in the end a little luck is a lot more valuable than a 2.8 in terms of the performance envelope. I don't recommend the 2.8 because of the increased cost and the unfavorable muliplier. Use the extra money on good ram, video card, power supply, and hard drive, as investment in these areas will give you better return on your money.

This is the biggest challenge for Intel systems. We must figure out what clock rate the chip will yield and combine this assumption with the fsb and memory limitations in order to pick an optimal chip. As previously mentioned, CPU's are individuals so there is no guarantee our assumption will prove 100% correct. As always in overclocking we must either try many examples or just get lucky on the first to produce all the overclock one might want. It is better to make the best educated guess we can and realize the difference between 2.8GHz and 3.2GHz is tiny. It is better to error on the side of caution and make sure our multiplier is low enough to maximize fsb and memory considerations, secure in the knowledge that they are more important than whether or not we can boast of 3GHz or a bit more.
Sticky material, if my dim light understood it, anybody's will.

[myarse]

I have an asus p4s533 m/b.
My first cpu was a 1.8a with 400 bus. I ran the m/b at 140fsb with a vid mod on the cpu giving me the 1.7volt range. I use kingmax pc2700 ddr using the 4:5 cpu/mem divider. This setup gave me 2.5ghz but if i went any higher it would fail on 3dmark. I think this was video card cooling problems.

I upgraded from air cooled to a koolance exos system and at the same time thought i would upgrade the cpu so i could get over 3ghz so I bought a 2.4a again with 400 bus.

I was very disapointed with the new cpu overclocking was very poor. To get the cpu to boot at 133fsb , 3.2ghz i needed to vid mod and use the asus +0.2v jumper giving me a whopping 1.9v in the asus monitor. Windows would load but crashed all the time so i just run it at 2.6ghz at the moment.

I will say the exos is a very good pice of kit and could not belive how quite it is after running a delta fan on a cooermaster fujiyama heat pipe for so long.

so as you can see not all cpu's will overclock and maybe i was just unlucky with the one i have. I do like the asus mobo but wish i could lock the pci/agp bus to see how far i can push the fsb on the new cpu.

myarse

[Mediawaves]

I changed that board for an Asus P4PE and will be using a P2.4b when I can . I was just looking at the Koolance Exos but seem to have my heart set on the Chip-Con™ Prometeia™, it's twice the money I know, but it's soooo coooool...