VOGONS


P3-450 or K6-2+ 450, which to use?

Topic actions

First post, by precaud

User metadata
Rank Member
Rank
Member

I have decided to reconstruct a hard disk recording/editing system I once used. I have two mobo/cpu combos to choose from:

: 450MHz AMD K6-2+ cpu in an Asus S1590S "Trinity 100AT"
: P3 450MHz in an FIC VB-609 (pretty much the same as VB-601)

Both have 100MHz bus; would have the same memory, and run Win 98SE. I know that the software makes heavy use of the FPU.

Any reason(s) to prefer either one over the other in this kind of application?

Reply 1 of 33, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

Depends what you want to achieve with that. Obviously PIII will be more responsive, especially if paired with 440BX chipset.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 2 of 33, by SirNickity

User metadata
Rank Oldbie
Rank
Oldbie

I'm curious -- what software? I know we all have our own various interests around here but I can't imagine wanting to go back in time for audio stuff when today's laptops run circles around them. I don't miss having to pre-render tracks, or waiting for bounces. So it must be something unique and interesting, and that has me intrigued.

Reply 3 of 33, by precaud

User metadata
Rank Member
Rank
Member
The Serpent Rider wrote:

Obviously PIII will be more responsive, especially if paired with 440BX chipset.

I'm biased toward the 440BX as well.

SirNickity wrote:

I'm curious -- what software?

I have SAW Plus 32.

I know we all have our own various interests around here but I can't imagine wanting to go back in time for audio stuff when today's laptops run circles around them. I don't miss having to pre-render tracks, or waiting for bounces. So it must be something unique and interesting, and that has me intrigued.

I will probably also need access to my old MTU Microsound DAW hardware/software from the mid-90's, hence the need for an ISA slot.

Reply 4 of 33, by RaverX

User metadata
Rank Member
Rank
Member

PIII 450 is much faster, especially in FPU.

Reply 5 of 33, by dionb

User metadata
Rank l33t++
Rank
l33t++
precaud wrote:
I'm biased toward the 440BX as well. […]
Show full quote
The Serpent Rider wrote:

Obviously PIII will be more responsive, especially if paired with 440BX chipset.

I'm biased toward the 440BX as well.

SirNickity wrote:

I'm curious -- what software?

I have SAW Plus 32.

I know we all have our own various interests around here but I can't imagine wanting to go back in time for audio stuff when today's laptops run circles around them. I don't miss having to pre-render tracks, or waiting for bounces. So it must be something unique and interesting, and that has me intrigued.

I will probably also need access to my old MTU Microsound DAW hardware/software from the mid-90's, hence the need for an ISA slot.

Unless I'm mistaken, SAW is ALU-only. In that case the FPU of the P3 is irrelevant and the full-speed L2 of the K6-2+ probably compensates for the larger 1/2 speed L2 on the P3.

In that case I'd say raw clock speed probably wins. How far can you clock the CPUs? Try the K6-2+ at 6x100MHz (or 4x133MHz if the motherboard can handle it) at 2.1V. If it manages that, it will run rings around the P3. At stock I'd suspect a tie with this software.

Reply 6 of 33, by precaud

User metadata
Rank Member
Rank
Member
dionb wrote:

Unless I'm mistaken, SAW is ALU-only.

I'd be very surprised if that was true. An EQ algoithm (for exmple) running in int or long int pewcision would be awful.

In that case the FPU of the P3 is irrelevant and the full-speed L2 of the K6-2+ probably compensates for the larger 1/2 speed L2 on the P3.

Except.. the L2 cache of the P3 runs at full cpu speed, not 1/2 speed, and there's twice as much of it.

Reply 7 of 33, by dionb

User metadata
Rank l33t++
Rank
l33t++
precaud wrote:
dionb wrote:

Unless I'm mistaken, SAW is ALU-only.

I'd be very surprised if that was true. An EQ algoithm (for exmple) running in int or long int pewcision would be awful.

Don't know the software, if that's the case the P3 (or indeed P2 or PPro) would make mincemeat of the K6.

In that case the FPU of the P3 is irrelevant and the full-speed L2 of the K6-2+ probably compensates for the larger 1/2 speed L2 on the P3.

Except.. the L2 cache of the P3 runs at full cpu speed, not 1/2 speed, and there's twice as much of it.

This is Katmai, not Coppermine. The cache runs at 1/2 CPU speed. But there's four times as much of it (512kB vs 128kB). If stuff fits in cache, the K6-2+ wins hard, if not the P3 wins. Mendocino experience shows that end result is about equal.

Reply 8 of 33, by mothergoose729

User metadata
Rank Oldbie
Rank
Oldbie

Pick whichever one gives you the most nostalgia tickles. It doesn't really matter. If you cared about performance, you would use faster parts than either CPU.

You said you are recreating a machine you once owned. Did it have a P3 or a K6-2?

Reply 9 of 33, by The Serpent Rider

User metadata
Rank l33t++
Rank
l33t++

If stuff fits in cache, the K6-2+ wins hard, if not the P3 wins.

It's not that simple. Also it's not necessarily 2x cache speed advantage, when it comes to K6.

I must be some kind of standard: the anonymous gangbanger of the 21st century.

Reply 10 of 33, by precaud

User metadata
Rank Member
Rank
Member
dionb wrote:

This is Katmai, not Coppermine. The cache runs at 1/2 CPU speed. But there's four times as much of it (512kB vs 128kB). If stuff fits in cache, the K6-2+ wins hard, if not the P3 wins. Mendocino experience shows that end result is about equal.

Ah, right, my bad. I wonder if a CuMine willrun in it? If so, I have a P3-667 that would run at 500 in it...

mothergoose729 wrote:

If you cared about performance, you would use faster parts than either CPU.

Not necessarily. Performance is just one of several considerations.

You said you are recreating a machine you once owned. Did it have a P3 or a K6-2?

It last had the S1590 w/ K6-2+.

Reply 11 of 33, by shamino

User metadata
Rank l33t
Rank
l33t

I would expect the 440BX to be a lot better than a VIA MVP3 for I/O. I had a good experience with a 440GX 450MHz system doing video capture a long time ago, and that's really the same thing as a 440BX.
My system before that was a Tyan S1590 board like yours, but I never tried using it for an I/O heavy situation.

Main memory throughput is definitely faster with the 440BX. I believe it also has better PCI throughput, which might matter if you make heavy use of the PCI bus.
Intel P2/P3 has a faster FPU than K6 series.
K6-2+/3 do have faster L2 than Katmai though, so that raises a little doubt.

You could try both and do a comparison, but I'd be surprised if the winner wasn't the 440BX setup.

I wonder if a CuMine willrun in it? If so, I have a P3-667 that would run at 500 in it...

First thing to worry about with Coppermine is whether the motherboard's VRM supports those voltages. The chip that regulates voltage is probably about 20 pins and located near the CPU slot, near the MOSFETs, inductors and capacitors in that area. If you can read the part number on that chip (or take a photo) we might be able to look up what voltage range it supports.

If the voltage regulator is compatible, then you could plug in your Coppermine to see if it boots, which will tell you if the BIOS supports them. Even at 500/100 it should be faster than a Katmai 450MHz so you could just roll with that.
It's unlikely for the Katmai to be faster than the Coppermine, but it could happen if the application really wants the larger cache.
There's definitely more CPU upgrade options on the Slot-1 440BX board than there is on the S1590.

Reply 12 of 33, by precaud

User metadata
Rank Member
Rank
Member
shamino wrote:

I would expect the 440BX to be a lot better than a VIA MVP3 for I/O. I had a good experience with a 440GX 450MHz system doing video capture a long time ago, and that's really the same thing as a 440BX.
My system before that was a Tyan S1590 board like yours, but I never tried using it for an I/O heavy situation.

That's excellent input, shamino, thanks.

Main memory throughput is definitely faster with the 440BX. I believe it also has better PCI throughput, which might matter if you make heavy use of the PCI bus.

I'm thinking worse-case PCI use here will be while recording four channels of audio at once from two PCI soundcards with SPDIF inputs, and spooling it out to a wide SCSI hard drive through a 2940UW. I never tried that with the S1590 setup; I only did two channels of record. That worked fine even with an earlier P100 setup.

First thing to worry about with Coppermine is whether the motherboard's VRM supports those voltages. The chip that regulates voltage is probably about 20 pins and located near the CPU slot, near the MOSFETs, inductors and capacitors in that area. If you can read the part number on that chip (or take a photo) we might be able to look up what voltage range it supports.

I'm pretty sure it doesn't, this board is in an HP Vectra and they never released a coppermine-compatible BIOS for it. I thought maybe a 667 in a slocket with voltage jumpers would do the trick?

If the voltage regulator is compatible, then you could plug in your Coppermine to see if it boots, which will tell you if the BIOS supports them. Even at 500/100 it should be faster than a Katmai 450MHz so you could just roll with that.

Yeah, if it does, that would probably be the best possible outcome inthis situation.

Reply 13 of 33, by shamino

User metadata
Rank l33t
Rank
l33t
precaud wrote:
shamino wrote:

First thing to worry about with Coppermine is whether the motherboard's VRM supports those voltages. The chip that regulates voltage is probably about 20 pins and located near the CPU slot, near the MOSFETs, inductors and capacitors in that area. If you can read the part number on that chip (or take a photo) we might be able to look up what voltage range it supports.

I'm pretty sure it doesn't, this board is in an HP Vectra and they never released a coppermine-compatible BIOS for it. I thought maybe a 667 in a slocket with voltage jumpers would do the trick?

Voltage-wise, yes you can jumper it to 1.8V which will be understood by an older VRM, but the BIOS still needs to cooperate.

A lot of boards have Coppermine compatible regulators on them even if they weren't needed. I even found a Coppermine regulator on my Tyan S1590, which is pretty weird for a desktop socket-7 board.
This probably happened because once the newer regulators were available the board manufacturers ordered those and used them on everything. They're backward compatible so they had no reason to keep stocking the older parts.
The most frustrating example I ran into was a slot-1 IBM Intellistation. It had a Coppermine capable VRM, and it would go through most of the POST procedure with a Coppermine, only to then display an error message saying the CPU was unsupported. It would freeze at that point and refused to complete booting. So it was restricted to Katmai because the BIOS was actively choosing to be difficult.

If you try it and run into a BIOS problem, make sure you have the latest version because sometimes CPU support gets added without it being mentioned in the release notes, and even if the CPU was never officially supported. I think sometimes those updates happen unintentionally, because the same code gets dispersed/inherited between a million different systems and nobody pays much attention to which bits of CPU support are actually needed by each machine.

Reply 14 of 33, by precaud

User metadata
Rank Member
Rank
Member
shamino wrote:

If you can read the part number on that chip (or take a photo) we might be able to look up what voltage range it supports.

It uses an ADP3152, which is designed for PII cpus.

If you try it and run into a BIOS problem, make sure you have the latest version because sometimes CPU support gets added without it being mentioned in the release notes, and even if the CPU was never officially supported. I think sometimes those updates happen unintentionally, because the same code gets dispersed/inherited between a million different systems and nobody pays much attention to which bits of CPU support are actually needed by each machine.

Therein lies a problem. I have been unable to locate ANY documentation of any kind (let alone BIOS updates) for this board, or the Vectra system it came in. Internet searches have only uncovered long-dead links to HP's support website from 15+ years ago.
Sysytem: HP Vectra VE Series 8 DT D6546E
Board: FIC VB-609, HP part number D6550-60003

The best I can tell is; the VB-609 appears to be a skinnied-down version of the VB-601. Basic layout, switch settings, cpus supported, etc. are all the same. Some IC's, fan headers, and the like are unpopulated. Even the 601 BIOS was never updated with specific support for Coppermine.

I used this as my main business computer for a couple years in the early 2000's. It's a very solid and reliable, if unremarkable, board.

Reply 15 of 33, by PARKE

User metadata
Rank Oldbie
Rank
Oldbie

There is one BIOS dated february 2000 available on Driverguide (see below) but the accompagfnying textfile does not indicate which CPUID's are supported.

In case you find relevant CPUID information for the HP BIOS:
CPUID's 0672h and 0673h are assigned to Katmai 450-600 steppings kB0 and kC0
CPUID 0681h is assigned to Coppermine 500-800 stepping cA2
CPUID 0683h is assigned to Coppermine 500-800 stepping cB0
CPUID 0686h is assigned to Coppermine 600-1Ghz stepping cC0
CPUID 068Ah is assigned to Coppermine 600-1.13Ghz stepping cD0

oooooooooooooooooooo
Vectra VE 6/xxx Series 8
HT.01.08 Bios
February 4th 2000

https://www.driverguide.com/driver/detail.php … Pmtiv69GCIYreVj

Reply 16 of 33, by precaud

User metadata
Rank Member
Rank
Member

Good find, PARKE. Dunno why google didn't put that one up for me.

I'll check what BIOS is installed on this one, I have it apart for a thorough cleaning at the moment.

Reply 17 of 33, by shamino

User metadata
Rank l33t
Rank
l33t
precaud wrote:
shamino wrote:

If you can read the part number on that chip (or take a photo) we might be able to look up what voltage range it supports.

It uses an ADP3152, which is designed for PII cpus.

The VID->Voltage chart on the ADP3152 datasheet is interesting. It looks like it will stay at 1.80V for any Vcore value that is below that level. 1/3 of the chart is 1.80V entries, all matching different VIDs that are in the 1.3-1.8V range (those VIDs can be seen on Coppermine regulator datasheets, like HIP6004B for an example).
Lots of regulators are made to disable Vcore when they encounter an unsupported VID, and some are reputed to output a totally inappropriate (and destructive) Vcore when they get confused. But the ADP3152 looks like it will saturate (maybe that's the wrong word) at the minimum 1.80V without complaining.

If your CPU is already S370 then this doesn't matter much, since you can just jumper the slocket to 1.80V anyway. But if you have a slot-1 Coppermine, then the way this regulator apparently behaves could be convenient.

If you already have a slocket with voltage jumpers, and also a multimeter, you could double check this behavior to be sure.
If you want to check that, try jumpering the empty slocket (no CPU) to something like 1.70V or whatever value is of interest and start the machine. Then measure the Vcore with a meter and see what it's putting out. The safest way to measure Vcore is usually at the back tab of the MOSFETs - half of them will have Vcore at that location. Ground can just come from the case.
This will confirm without any doubt what voltage will appear if you put a Slot-1 CPU in there with that VID.

But just trusting what the datasheet says, it looks like it will happily power a Coppermine at 1.80V, even if the VID requested is below that.

Reply 18 of 33, by feipoa

User metadata
Rank l33t++
Rank
l33t++

The K6-2+ 450 has about the 3D accelerated gaming score (Quake 2) of a PII-300. Whatever you finally decide upon, you'll probably be wanting more in the future. I recommend building, both, a socket 7 and slot 1 system, and eventually, any other platform you have interest in. The largest obstacle for me is usually the case.

Plan your life wisely, you'll be dead before you know it.

Reply 19 of 33, by precaud

User metadata
Rank Member
Rank
Member
shamino wrote:

The VID->Voltage chart on the ADP3152 datasheet is interesting. It looks like it will stay at 1.80V for any Vcore value that is below that level. 1/3 of the chart is 1.80V entries, all matching different VIDs that are in the 1.3-1.8V range.
{snip}
If your CPU is already S370 then this doesn't matter much, since you can just jumper the slocket to 1.80V anyway. But if you have a slot-1 Coppermine, then the way this regulator apparently behaves could be convenient.

Very interesting sleuthing, shamino. I wasn't aware that BIOSes set the vreg based on stepping codes. The only slot-1 Coppermine I have is a 1000/133 which I won't even bother trying.

If the 667-at-500 boots, I will definitely check the voltage at the slocket with a meter. Thanks again.

feipoa wrote:

Whatever you finally decide upon, you'll probably be wanting more in the future.

I understand why you'd say that, but in this case, I won't. Once I transfer the 4-track analog masters and remix them, this unit will probably go back into the closet it came from...