Help: Fastest AMD for Via KT266A

[Kevin Lawton]

<snip>
| Sorry, but either I am misunderstanding what you are saying or you are
| confusing memory bus speed and FSB speed. DDR 333 is effectively 333
| M/sec on the memory bus (actually 2 x 166 MHz). 333 FSB is 333 MHz on
| the FSB to the processor. They don't have to be the same with several
| chipsets - including the VIA KT266, KT333 and KT400. The northbridge
| is able to run the front-side bus and the memory bus at different
| speeds. In the old days, both memory and process would run at the
| same speed - that of the slowest - or 'wait states' would be
| intriduced if a fast processor was attached to slower memory, but now
| in the twenty-first century we can run them at different speeds.
| My appologies for being either pedantic or pernikity, but the OP has
| a VIA KT266 chipset - 333 MHz FSB max to DDR 266 max.

Profuse appologies to all readers of this newsgroup for replying to my own
posting, but I got that last bit wrong - didn't I ? :-(
The VIA chipset does NOT support a 333 MHz FSB, but it DOES support DDR 266
maximum.
Kevin.

 

[Piotr Makley]

||| Then again a Thoroughbred or newer might work on your
||| motherboard, I just don't know. The faster the chip the
||| less likely it is to work from an amperage perspective too,
||| that board was designed around older CPUs that used less
||| current. It could run for awhile then fail prematurely.
||| If it were me at this point I'd be thinking about upgrading
||| the motherboard and memory, then whichever CPU fits the
||| budget, many people feel a Barton XP2500 is a good choice.
||
|| Doesn't the Barton core run on a 333 MHz FSB - putting it
|| outside of the OP's parameters ?
|| Thoroughbred core XP2400 runs 266 MHz FSB and the early
|| XP2600 did too - though newer ones are 333 MHz FSB.
|| Kevin.
|
| There are mobile Bartons that use DDR266 FSB, but they
| default at 6X multiplier, the motherboard must have the
| ability to manually set the multiplier higher. Even so,
| since they default to 6X (a multiplier under 13X), they're
| stuck in that sub-13X range since the board likley doesn't
| have full multiplier selections, so the maximum resulting
| speed without any hacks or bus overclocking is 12.5 mult X
| 133 = 1.67GHz.
|
| However, we still don't even know if the present motherboard
| will even support any Barton... the suggestion I made in the
| prior post was right after I'd written "I'd be thinking about
| upgrading the motherboard and memory", with the presumption
| being that the upgrade would be a fairly modern motherboard
| with Barton & DDR333 FSB support.

Sorry, but either I am misunderstanding what you are saying or
you are confusing memory bus speed and FSB speed. DDR 333 is
effectively 333 M/sec on the memory bus (actually 2 x 166
MHz). 333 FSB is 333 MHz on the FSB to the processor. They
don't have to be the same with several chipsets - including
the VIA KT266, KT333 and KT400. The northbridge is able to run
the front-side bus and the memory bus at different speeds. In
the old days, both memory and process would run at the same
speed - that of the slowest - or 'wait states' would be
intriduced if a fast processor was attached to slower memory,
but now in the twenty-first century we can run them at
different speeds.
My appologies for being either pedantic or pernikity, but the
OP has a VIA KT266 chipset - 333 MHz FSB max to DDR 266 max.
Kevin.

Hi Kevin and Kony, I am the OP. Here is a long posting with more
details. Just to recap this thread, I am trying to work out which
is the best processor to use on my mobo bearing in mind that I want
to mimimize any increase in power required.

I thought Kony had said the last word but Kev points out something
about bus speeds. Kev, just in case the 'A' makes a difference my
chipset is a Via KT266A not the KT266 you mentioned. (I am told
that the 'A' brings "Improved memory timing and deeper command
queues"). The Via specs are at:
http://www.via.com.tw/en/apollo/KT266A.jsp

My memory is 133 MHz SDRAM and the mobo is a Syntax SV266A (see
below for more info and links).

---

I'm far from being an expert on what you, Kev, wrote about cpu and
memory bus speeds so I consulted my copy of "Upgrading and
Repairing PCs" by Scott Mueller. http://tinyurl.com/yvz4j. He
says:

<QUOTE from page 319> The best solution is if the memory bus runs
at the same speed as the processor bus. Systems that use PC133
SDRAM have a memory bandwidth of 1066 MBps, which is the same as
the 133 MHz cpu bus. In another example, Athlon systems running a
266 MHz processor bus also run PC2100 DDR-SDRAM, which has a memory
bandwidth of 1066 MBps - exactly the same as the processor bus in
those systems. [-- snip Pentium stuff --] Running memory at the
same speed as the processor bus negates the need for having cache
memory on the motherboard. That is why when the L2 cache moved
into the processor, nobody added an L3 cache to the motherboard.
<UNQUOTE> He says elsewhere: "It is always best for performance
when the bandwidth of memory matches that of the processor".

As I said above, my mobo is a budget Syntax SV266A. I chose this
because I wanted to use the SDRAM from a previous mobo. The Syntax
mobo has a jumper for 100 MHz or 133 MHz cpu frequency but there
are no other jumpers for bus multipliers.

Syntax support (inc PDF manual) for SV266A is at:
http://www.syntaxusa.com/support/downloads.php

Syntax nowadays have a model SV266AD which I think only adds some
USB ports to the model SV266A. Specs for the current Syntax
SV266AD is at: http://tinyurl.com/yt2an

For what it is worth, Ebuyer customer feedback on my SV266A mobo is
at: http://tinyurl.com/3d44w. But note that this Syntax SV266A is
definitely *not* a PCchips board.

So, guys, with all this new info about my mobo what is the thinking
on the best way forward for a processor for me? Currently I have
two cpus on my shortlist ..

(a) Duron 1800
(b) one of late T'bred Athlons up to a 2400+ if PSU permits.

Any views?

PS: In case it is a factor, the BIOS has options (1) to regulate
the cpu voltage and (2) the CPU Host Clock as shown on the PDF
manual on page 47). I don't know if these help but maybe I can
underclock any replacement cpu if it gets to be too demanding for
my PSU. dont know about this sort of thing. Can I do that with
the BIOS options I have got?

 

[kony]

Hi Kevin and Kony, I am the OP. Here is a long posting with more
details. Just to recap this thread, I am trying to work out which
is the best processor to use on my mobo bearing in mind that I want
to mimimize any increase in power required.

I thought Kony had said the last word but Kev points out something
about bus speeds. Kev, just in case the 'A' makes a difference my
chipset is a Via KT266A not the KT266 you mentioned. (I am told
that the 'A' brings "Improved memory timing and deeper command
queues"). The Via specs are at:
http://www.via.com.tw/en/apollo/KT266A.jsp

Simply put, we can ignore the memory bus speed. It is not a deciding
factor on which CPU can be used, but rather it's derived from the FSB
speed, that speed need support the CPU used for it to run at proper AMD
spec'd speed. Being KT266A, it supports 100 or 133MHz Clock rate, at
double data rate it's "DDR266", sometimes confusingly called 266MHz.

This would allow the Duron or any Palomino, and T'Bred with DDR266 FSB,
not DDR33 FSB, and futher any Mobile Barton using DDR266 FSB, but not the
regular desktop Barton with DDR333 FSB. The primary issue is then which
CPU core (family) the motherboard supports. Again, I don't know this.
The most likley CPU to work would be a Palomino core but those are also
the hotter, most power hungry per MHz. To get higher performance with
similar power consumption (and heat) you generally have to use a smaller
process size CPU, like the Duron, T'Bred, or Barton.

As I said above, my mobo is a budget Syntax SV266A. I chose this
because I wanted to use the SDRAM from a previous mobo. The Syntax
mobo has a jumper for 100 MHz or 133 MHz cpu frequency but there
are no other jumpers for bus multipliers.

Syntax support (inc PDF manual) for SV266A is at:
http://www.syntaxusa.com/support/downloads.php

Syntax nowadays have a model SV266AD which I think only adds some
USB ports to the model SV266A. Specs for the current Syntax
SV266AD is at: http://tinyurl.com/yt2an

For what it is worth, Ebuyer customer feedback on my SV266A mobo is
at: http://tinyurl.com/3d44w. But note that this Syntax SV266A is
definitely *not* a PCchips board.

So, guys, with all this new info about my mobo what is the thinking
on the best way forward for a processor for me? Currently I have
two cpus on my shortlist ..

(a) Duron 1800
(b) one of late T'bred Athlons up to a 2400+ if PSU permits.

Any views?

Given that your power supply is a limit I'd get the Duron. An XP2400
might be faster, but would tend to consume more power too... performance
isn't "free".

 

[kony]

| 7VRXP was a intentionally crippled. Most KT333 boards could run
| DDR333
| FSB cpus, though it is technically out of spec, the chipset was
| actually stable past DDR333, nearer to DDR400, but Gigabyte in
| particular chose not to support the 1/5 PCI divider. I had one of
| those board and had planned
| to build myself a system out of it but when I learned of the FSB, PCI
| issue I swapped to an Asus KT333 board which I still have as an aux.
| system running an o'c Barton.

Interesting to hear of someone else with a GA-7VRXP. :-o
May I ask what CPU and op system you ran and if you had any problems ? I'm
finding mine a bitch to get going with Windows 2000 on an XP2400 !
Kevin.

I don't have the board anymore. I did with it what I do with most, when
it came in I plug-n-played a previously-existing Win98SE testbed
installation (simply because Win98SE will so easily plug-n-play install
with a testbed config with all software ready to go in under 10 minutes)
to test it for performance/defect/etc. At the time I put a (think it was
a T'Bred) XP1700 into it, only to find that it wouldn't even support
DDR333 FSB. There was silkscreening on the PCB for a switch header, IIRC,
but no switch or support circuit for it so the board was stuck with max
1/4 PCI divider. I don't recall the revision I had. Since it wouldn't do
DDR333 I thought about returning it but instead sold it, never had it in a
finished, "used" system.

There isn't anything I'm aware of that would interfere with a KT333
chipset board running Win2K, though of course you need the 4in1 chipset
driver. I thought I had heard someone report that theirs was instable
with higher speed CPUs but I don't have any better recollection of their
setup, it could've easily been the power supply, not the board. About the
only suggestion I can make is trying newest bios available and run
Memtest86 on it if you hadn't already.

 

[Kevin Lawton]

| On Fri, 16 Apr 2004 07:58:40 +0000 (UTC), "Kevin Lawton"
|
|
||| 7VRXP was a intentionally crippled. Most KT333 boards could run
||| DDR333
||| FSB cpus, though it is technically out of spec, the chipset was
||| actually stable past DDR333, nearer to DDR400, but Gigabyte in
||| particular chose not to support the 1/5 PCI divider. I had one of
||| those board and had planned
||| to build myself a system out of it but when I learned of the FSB,
||| PCI issue I swapped to an Asus KT333 board which I still have as an
||| aux. system running an o'c Barton.
||
|| Interesting to hear of someone else with a GA-7VRXP. :-o
|| May I ask what CPU and op system you ran and if you had any problems
|| ? I'm finding mine a bitch to get going with Windows 2000 on an
|| XP2400 !
|| Kevin.
|
| I don't have the board anymore. I did with it what I do with most,
| when
| it came in I plug-n-played a previously-existing Win98SE testbed
| installation (simply because Win98SE will so easily plug-n-play
| install
| with a testbed config with all software ready to go in under 10
| minutes)
| to test it for performance/defect/etc. At the time I put a (think
| it was
| a T'Bred) XP1700 into it, only to find that it wouldn't even support
| DDR333 FSB. There was silkscreening on the PCB for a switch header,
| IIRC, but no switch or support circuit for it so the board was stuck
| with max 1/4 PCI divider. I don't recall the revision I had. Since
| it wouldn't do DDR333 I thought about returning it but instead sold
| it, never had it in a finished, "used" system.
|
| There isn't anything I'm aware of that would interfere with a KT333
| chipset board running Win2K, though of course you need the 4in1
| chipset driver. I thought I had heard someone report that theirs was
| instable
| with higher speed CPUs but I don't have any better recollection of
| their setup, it could've easily been the power supply, not the board.
| About the only suggestion I can make is trying newest bios available
| and run
| Memtest86 on it if you hadn't already.

That's interesting, Kony, I didn't know you could get a Thoroughbred core
with 333 FSB - I thought they were all 266 FSB as my XP2400 certainly is. I
had believed you had to go to a Barton core or better to get 333 FSB. That's
why I got the XP2400 - fastest CPU with 266 FSB. I would have got an XP2600
with 333 FSB if I had thought the board would support it, but Gigabyte say
it doesn't.
I am successfully running DDR 333 memory, so the memory bus is running
effectively 333 MHz (hence the 'KT333' designation I understand). This is
even working fine with a Duron 650 CPU with the FSB switched to 200 FSB (100
MHz clock). The clever thing with the VIA KT266 and KT333 chipsets is that
they can run the front side bus and memory bus at different clock rates. On
this board - GA-7VRXP Rev1.1 - you set the CPU FSB clock and multiplier
using switches on the board, but set the memory bus speed in the BIOS. So
you can 'mix and match' CPUs and memory and have the chipset take care of
it. No worry, the board does DDR333 (PC2700), DDR266 (PC2100) or DDR200
(PC1600) just by setting the memory bus speed to 166 MHz, 133 MHz or 100 MHz
in the BIOS. If you were hoping to run the XP1700 with a 333 FSB (166 MHz
CPU clock) I can understand your disappointment - this board does not
support that FSB and I'm not too sure the CPU would like it either !
As regards my Windows 2000 problem (a little off-topic - I do hope the OP
doesn't mind) it has gone in and is running fine with the Duron 650 (or a
Duron 1200) in place, but stick in the XP2400 and it hangs at the splash
screen. Windows Me and Linux run fine on the XP2400 and this board, even
BeOS runs once the kernel has been patched to support SSE. It is just the
combination of XP2400 and Windows 2000 which screws up ! Very puzzling
indeed. I've swapped every system component in turn, even gone to a 550w PSU
and the problem remains - I even swapped the GA-7VRXP for another one. In
both cases, the BIOS version is the latest available on the website. A big
puzzler this one, it has kept me busy for a couple of weeks now. Someone
mentioned disabling CPU caches to see if the CPU was at fault, but I can't
find the right BIOS option for doing that (do you know any better than me
?). Woudn't mind running some CPU diagnostic software to fully test the
XP2400 before I go and buy another - do you know of anything ?
Kevin.

 

[kony]

That's interesting, Kony, I didn't know you could get a Thoroughbred core
with 333 FSB - I thought they were all 266 FSB as my XP2400 certainly is. I
had believed you had to go to a Barton core or better to get 333 FSB. That's
why I got the XP2400 - fastest CPU with 266 FSB. I would have got an XP2600
with 333 FSB if I had thought the board would support it, but Gigabyte say
it doesn't.

It's not that the CPU had a default DDR333 FSB, rather that that's how I
run 'em in any box with motherboard/memory that'll support DDR333. The
CPU will handle whichever FSB you like so long as you have multiplier
adjustements or "hack-it-out" to the multiplier you want. Then again,
some of the T'Bred I have/had would o'c to DDR333 FSB without fooling with
the multiplier, though I don't recall if that particular CPU would or not.
Point being, the most performance can be extracted from any given AMD cpu
by running at highest (stable) synchronous FSB & mem bus. Today that
would typically mean using an nForce2 Ultra board with PC3200 or higher
memory, shooting for a minimum of 200MHz FSB with any CPU installed. One
thing to consider though is that if the CPU can do more work running this
way, it may easily run hotter, not due to the FSB speed but greater,
closer to true "full load" potential.

I am successfully running DDR 333 memory, so the memory bus is running
effectively 333 MHz (hence the 'KT333' designation I understand). This is
even working fine with a Duron 650 CPU with the FSB switched to 200 FSB (100
MHz clock). The clever thing with the VIA KT266 and KT333 chipsets is that
they can run the front side bus and memory bus at different clock rates.

It's not all that great though, there is a performane hit involved with
running in async mode. The FSB throughput is a bottleneck so in some
cases performance goes up by reducing memory bus to sync mode.

On
this board - GA-7VRXP Rev1.1 - you set the CPU FSB clock and multiplier
using switches on the board, but set the memory bus speed in the BIOS. So
you can 'mix and match' CPUs and memory and have the chipset take care of
it. No worry, the board does DDR333 (PC2700), DDR266 (PC2100) or DDR200
(PC1600) just by setting the memory bus speed to 166 MHz, 133 MHz or 100 MHz
in the BIOS. If you were hoping to run the XP1700 with a 333 FSB (166 MHz
CPU clock) I can understand your disappointment - this board does not
support that FSB and I'm not too sure the CPU would like it either !

Yes the CPU, as well as Palomino, Thorton, Barton, Duron (at least the
newer cores), will do it fine, even 200Mhz FSB, providing the resultant
MHz speed of CPU itself is still within it's ability to run stabily. To
that end the board and memory are the determining factors.

As regards my Windows 2000 problem (a little off-topic - I do hope the OP
doesn't mind) it has gone in and is running fine with the Duron 650 (or a
Duron 1200) in place, but stick in the XP2400 and it hangs at the splash
screen. Windows Me and Linux run fine on the XP2400 and this board, even
BeOS runs once the kernel has been patched to support SSE. It is just the
combination of XP2400 and Windows 2000 which screws up ! Very puzzling
indeed. I've swapped every system component in turn, even gone to a 550w PSU
and the problem remains - I even swapped the GA-7VRXP for another one. In
both cases, the BIOS version is the latest available on the website. A big
puzzler this one, it has kept me busy for a couple of weeks now. Someone
mentioned disabling CPU caches to see if the CPU was at fault, but I can't
find the right BIOS option for doing that (do you know any better than me
?). Woudn't mind running some CPU diagnostic software to fully test the
XP2400 before I go and buy another - do you know of anything ?
Kevin.

How exactly does Win2K err? ALWAYS at the splash screen?
You wouldn't happen to have another Windows OS you could try, WinXP or
Win98SE?

This was an existing Win2K installation on that board with all same
components except you'd swapped out the Duron, replaced with the XP2400?

I'd try clearing CMOS to reset to defaults and leaving it at the defaults
except the following:
Does it exhibit same error if you reduce FSB speed to 100MHz, underclock?
I would definitely leave the memory in synchronous mode, 133MHz (or 100MHz
if you underclock the FSB to 100MHz), for any further testing, and try
with only one memory module.

What about the hardware monitor screen in the BIOS, or if you have the
ability to use a multimeter to measure CPU vCore on the board, does it
seem appropriate?

It's not a generic 550W PSU, is it?
I supose it's possible the CPU is defective, but in general and especially
considering that it ran other OS, seems unlikely.

 

[Kevin Lawton]

<snip>
| It's not that the CPU had a default DDR333 FSB, rather that that's
| how I run 'em in any box with motherboard/memory that'll support
| DDR333. The
| CPU will handle whichever FSB you like so long as you have multiplier
| adjustements or "hack-it-out" to the multiplier you want. Then again,
| some of the T'Bred I have/had would o'c to DDR333 FSB without fooling
| with the multiplier, though I don't recall if that particular CPU
| would or not. Point being, the most performance can be extracted from
| any given AMD cpu
| by running at highest (stable) synchronous FSB & mem bus. Today that
| would typically mean using an nForce2 Ultra board with PC3200 or
| higher memory, shooting for a minimum of 200MHz FSB with any CPU
| installed. One thing to consider though is that if the CPU can do
| more work running this way, it may easily run hotter, not due to the
| FSB speed but greater,
| closer to true "full load" potential.

To be honest, absolute top performance is 'too rich for my blood' - or to
put it another way, only a small gain for a lot more cash. I wanted to build
this machine for a reasonable price for video capture and editing. My
fastest machine so far was a 1.4 GHz Athlon Thunderbird with 133 MHz SDRAM.
I figured an XP2400 with DDR 333 memory would give sufficient performance
without shredding my wallet. To be honest, rather than overclock I'm happy
to just run a system as specced and hope for reliability. .
|
|| I am successfully running DDR 333 memory, so the memory bus is
|| running effectively 333 MHz (hence the 'KT333' designation I
|| understand). This is even working fine with a Duron 650 CPU with the
|| FSB switched to 200 FSB (100 MHz clock). The clever thing with the
|| VIA KT266 and KT333 chipsets is that they can run the front side bus
|| and memory bus at different clock rates.
|
| It's not all that great though, there is a performane hit involved
| with running in async mode. The FSB throughput is a bottleneck so in
| some
| cases performance goes up by reducing memory bus to sync mode.

I guess I'll not lose much throughput by running the memory bus at 266 MHz
the same as the front-side bus - and will probably get it back as it will be
synched.

|| On
|| this board - GA-7VRXP Rev1.1 - you set the CPU FSB clock and
|| multiplier using switches on the board, but set the memory bus speed
|| in the BIOS. So you can 'mix and match' CPUs and memory and have the
|| chipset take care of it. No worry, the board does DDR333 (PC2700),
|| DDR266 (PC2100) or DDR200 (PC1600) just by setting the memory bus
|| speed to 166 MHz, 133 MHz or 100 MHz in the BIOS. If you were hoping
|| to run the XP1700 with a 333 FSB (166 MHz CPU clock) I can
|| understand your disappointment - this board does not support that
|| FSB and I'm not too sure the CPU would like it either !
|
| Yes the CPU, as well as Palomino, Thorton, Barton, Duron (at least the
| newer cores), will do it fine, even 200Mhz FSB, providing the
| resultant
| MHz speed of CPU itself is still within it's ability to run stabily.
| To that end the board and memory are the determining factors.

So could I run the XP2400 with a 333 FSB (166 MHz x 2) and the correct
multiplier to sync with my DDR 333 memory on the right m/board ? The
GA-7VRXP won't let me set the FSB that high. :-(
|
|| As regards my Windows 2000 problem (a little off-topic - I do hope
|| the OP doesn't mind) it has gone in and is running fine with the
|| Duron 650 (or a Duron 1200) in place, but stick in the XP2400 and it
|| hangs at the splash screen. Windows Me and Linux run fine on the
|| XP2400 and this board, even BeOS runs once the kernel has been
|| patched to support SSE. It is just the combination of XP2400 and
|| Windows 2000 which screws up ! Very puzzling indeed. I've swapped
|| every system component in turn, even gone to a 550w PSU and the
|| problem remains - I even swapped the GA-7VRXP for another one. In
|| both cases, the BIOS version is the latest available on the website.
|| A big puzzler this one, it has kept me busy for a couple of weeks
|| now. Someone mentioned disabling CPU caches to see if the CPU was at
|| fault, but I can't find the right BIOS option for doing that (do you
|| know any better than me ?). Woudn't mind running some CPU diagnostic
|| software to fully test the XP2400 before I go and buy another - do
|| you know of anything ?
|| Kevin.
|
| How exactly does Win2K err? ALWAYS at the splash screen?
| You wouldn't happen to have another Windows OS you could try, WinXP or
| Win98SE?

Yes, it always hangs at the same place - either after the first re-boot on a
fresh install, or when booting a previous installation, the 'Starting
Windows' progress stripe completes and then it just hangs - overnight if
left that long. At that point it should bring up the Windows 2000 splash
screen, but it doesn't. If I choose 'Safe mode with Command Prompt' it shows
the last module successfully loaded as 'Mup.sys', so I think it is hanging
immediately after that.
DOS works fine, Windows 95 works fine, Windows Me works fine, Linux - Red
Hat 9 - works fine, BeOS 5 works after the 'SSE' patch has been installed.
It just seems to hang with Windows 2000.

| This was an existing Win2K installation on that board with all same
| components except you'd swapped out the Duron, replaced with the
| XP2400?

Yes, I've tried that with every system component in turn - one at a time of
course. Duron 650 - all fine, Duron 1200 - all fine, XP2400 - Windows 2000
hangs but any other op system seems to work okay. I was wondering if I could
try radically under-clocking the XP2400 - or might that harm it ?

| I'd try clearing CMOS to reset to defaults and leaving it at the
| defaults except the following:

Yes, I tried that - same results.

| Does it exhibit same error if you reduce FSB speed to 100MHz,
| underclock?

Yes, I have tried setting FSB clock to 100 MHz and the results were the
same.

| I would definitely leave the memory in synchronous mode, 133MHz (or
| 100MHz if you underclock the FSB to 100MHz), for any further testing,
| and try with only one memory module.

I have tried bringng the memory bus clock down to the same as the CPU clock,
and have tried several different memory strips - all PC2700 / DDR 333.
Various makes of memory - some 'Crucial' and others. All the same results
again.

| What about the hardware monitor screen in the BIOS, or if you have the
| ability to use a multimeter to measure CPU vCore on the board, does it
| seem appropriate?

I think so - I will get the specs off te AMD site and double-check. I have
tried two different (or identical) GA-7VRXP boards and two power supplies at
diffeerent times (only change one thing at a time) started with a 300w and
moved to a 550w. Would you suggest altering the vCore setting, or might that
damage the CPU ?

| It's not a generic 550W PSU, is it?

No, I use branded CPUs. I did also try a 450w Antec at one point, but that
was destined for a different machine.
Given that I'm only running a few components - one hard drive, etc, I would
have thought that the 300w would be ample and the 550w overkill.

I supose it's possible the CPU is defective, but in general and
| especially considering that it ran other OS, seems unlikely.

That's what I figured. If I could find any CPU diagnostic software to run
under dos, Windows Me or Linux, then I could be conclusive about the CPU
without having to buy another to swap with it. Any ideas here ?
TIA
Kevin.

 

[kony]

To be honest, absolute top performance is 'too rich for my blood' - or to
put it another way, only a small gain for a lot more cash. I wanted to build
this machine for a reasonable price for video capture and editing. My
fastest machine so far was a 1.4 GHz Athlon Thunderbird with 133 MHz SDRAM.
I figured an XP2400 with DDR 333 memory would give sufficient performance
without shredding my wallet. To be honest, rather than overclock I'm happy
to just run a system as specced and hope for reliability. .

I don't usually go for top performance either, rather the top performance
that can be attained _quietly_, without undue expense or excessive temp.
Running the FSB higher with reduced multiplier is only technically an o'c
but not really significant of the board and CPU are within core margins.

I guess I'll not lose much throughput by running the memory bus at 266 MHz
the same as the front-side bus - and will probably get it back as it will be
synched.

Often the greatest gain is from bumping the FSB a little bit if possible,
after setting tightest memory timings possible, though it takes a LOT of
testing to be sure there's no memory errors... better safe than sorry,
leave a good 5-10% margin for memory.

So could I run the XP2400 with a 333 FSB (166 MHz x 2) and the correct
multiplier to sync with my DDR 333 memory on the right m/board ? The
GA-7VRXP won't let me set the FSB that high. :-(

Yes, if your memory is PC2700 then any nForce board should do nicely.
IIRC the XP2400 is 2.0 GHz so that'd translate into 12 X 166. Some boards
don't allow reducing the 15X default multiplier below 13X though, and I
don't recall the easiest way to tell which boards do and don't allow it,
so research of a particular board is warranted, or just set it to 13X 166
= 2.17GHz... an XP2400 should be able to run @ 2.17GHz easily, even if you
had to up the vCore to 1.7V. Point being, since Athlons have multipler
adjustment except for the newer Thortons and desktop Bartons, there are a
lot of options for which multi & FSB combination to use.

| How exactly does Win2K err? ALWAYS at the splash screen?
| You wouldn't happen to have another Windows OS you could try, WinXP or
| Win98SE?

Yes, it always hangs at the same place - either after the first re-boot on a
fresh install, or when booting a previous installation, the 'Starting
Windows' progress stripe completes and then it just hangs - overnight if
left that long. At that point it should bring up the Windows 2000 splash
screen, but it doesn't. If I choose 'Safe mode with Command Prompt' it shows
the last module successfully loaded as 'Mup.sys', so I think it is hanging
immediately after that.

That's a really good place to start investigating.
http://www.google.com/search?q=Mup.sys+freeze+OR+halt
http://www.earthv.com/tips_detail.asp?TipID=63

DOS works fine, Windows 95 works fine, Windows Me works fine, Linux - Red
Hat 9 - works fine, BeOS 5 works after the 'SSE' patch has been installed.
It just seems to hang with Windows 2000.

It would seem quite unlikely to be the CPU. If it's just the USB2 as one
of the above links suggests, disable that and retry. You might also seek
a Win2K boot list to get an idea of what might be loading next after
mup.sys. The thing that makes no sense is that it'd happen with the
XP2400 but not the Duron, if it were the USB2.

| This was an existing Win2K installation on that board with all same
| components except you'd swapped out the Duron, replaced with the
| XP2400?

Yes, I've tried that with every system component in turn - one at a time of
course. Duron 650 - all fine, Duron 1200 - all fine, XP2400 - Windows 2000
hangs but any other op system seems to work okay. I was wondering if I could
try radically under-clocking the XP2400 - or might that harm it ?

Radically? Shouldn't harm it at all. Most motherboard chipsets are only
stable down to around 50MHz FSB, but I doubt your board even supports
under 100MHz FSB.

| I'd try clearing CMOS to reset to defaults and leaving it at the
| defaults except the following:

Yes, I tried that - same results.

| Does it exhibit same error if you reduce FSB speed to 100MHz,
| underclock?

Yes, I have tried setting FSB clock to 100 MHz and the results were the
same.

Try disabling all onboard features, USB, sound, etc, etc.

| I would definitely leave the memory in synchronous mode, 133MHz (or
| 100MHz if you underclock the FSB to 100MHz), for any further testing,
| and try with only one memory module.

I have tried bringng the memory bus clock down to the same as the CPU clock,
and have tried several different memory strips - all PC2700 / DDR 333.
Various makes of memory - some 'Crucial' and others. All the same results
again.

| What about the hardware monitor screen in the BIOS, or if you have the
| ability to use a multimeter to measure CPU vCore on the board, does it
| seem appropriate?

I think so - I will get the specs off te AMD site and double-check. I have
tried two different (or identical) GA-7VRXP boards and two power supplies at
diffeerent times (only change one thing at a time) started with a 300w and
moved to a 550w. Would you suggest altering the vCore setting, or might that
damage the CPU ?

| It's not a generic 550W PSU, is it?

No, I use branded CPUs. I did also try a 450w Antec at one point, but that
was destined for a different machine.
Given that I'm only running a few components - one hard drive, etc, I would
have thought that the 300w would be ample and the 550w overkill.

Yes, that's enough power but I was just trying to rule out overrated power
supplies... some generics marked as 550W are only worth 250... the Antec
450W should've been more than enough power.

I supose it's possible the CPU is defective, but in general and
| especially considering that it ran other OS, seems unlikely.

That's what I figured. If I could find any CPU diagnostic software to run
under dos, Windows Me or Linux, then I could be conclusive about the CPU
without having to buy another to swap with it. Any ideas here ?
TIA
Kevin.

Prime95? There are a number of programs claiming to do a diagnostic of a
whole system but nothing comes to mind that would stress test a cpu using
all possible instructions. Still, if the system runs flawlessly on WinME
(or relatively so, it IS WinME!) it would more likley be something onboard
the Gigabyte motherboard causing the problem, a feature, not the KT333
chipset since those are quite common and not reported in any large numbers
to be problematic with Win2K.

 

[iTsMeMa]

Just to throw in here,

I have an older chipset, KT133A, that is currently running an XP2400. Power
requirements for this proc are less than the 1.4 T-Bird I ran for 2 years.
My opinion is that your board will run the XP2400. It will, in fact run an
XP2500 333fsb. The limitation will depend on whether or not that barton is
locked, and how happy your board is being pushed. The mem (sdram) unless it
is high perf, will not like going past 143 - 147 fsb. If the proc is locked,
as most now are, your results are likely to be about 145 X 11.5 or 1667 mhz
on a barton.
The XP2400 will give you 2000 mhz, an obvious choice.
I have had a 1.4T-Bird, XP1800, XP2400, XP2500 Barton all running on mine,
the T-bird being the only officially supported.
Wes Newell and Michael Brown (amongst others) both offer great info on
boards and processors. Browse their sites.
Dive in and have fun.
And always watch temps!

Regards,

Garry.

 

[Kevin Lawton]

Garry,
I've got to thank you for giving me an idea as regards a problem I currently
have: trying to get an XP2400 to run Windows 2000 on a GA-7VRXP KT333 mobo.
I also have an Athlon Thunderbird 1.4 GHz running Windows 2000 in a GA-7IXEH
KT133 mobo. This system has been running and stable for two or three years.
I guess from what you've said below, though I didn't realise it before, that
I could try the XP2400 in this system just by adjusting the CPU FSB
multiplier ?
Gives me a better chance of solving the problem if I can try this too.
Thanks,
Kevin.

| Just to throw in here,
|
| I have an older chipset, KT133A, that is currently running an XP2400.
| Power requirements for this proc are less than the 1.4 T-Bird I ran
| for 2 years. My opinion is that your board will run the XP2400. It
| will, in fact run an XP2500 333fsb. The limitation will depend on
| whether or not that barton is locked, and how happy your board is
| being pushed. The mem (sdram) unless it is high perf, will not like
| going past 143 - 147 fsb. If the proc is locked, as most now are,
| your results are likely to be about 145 X 11.5 or 1667 mhz on a
| barton.
| The XP2400 will give you 2000 mhz, an obvious choice.
| I have had a 1.4T-Bird, XP1800, XP2400, XP2500 Barton all running on
| mine, the T-bird being the only officially supported.
| Wes Newell and Michael Brown (amongst others) both offer great info
| on boards and processors. Browse their sites.
| Dive in and have fun.
| And always watch temps!
|
| Regards,
|
| Garry.
|
|
|
| |||
|| Hi Kevin and Kony, I am the OP. Here is a long posting with more
|| details. Just to recap this thread, I am trying to work out which
|| is the best processor to use on my mobo bearing in mind that I want
|| to mimimize any increase in power required.
||
|| I thought Kony had said the last word but Kev points out something
|| about bus speeds. Kev, just in case the 'A' makes a difference my
|| chipset is a Via KT266A not the KT266 you mentioned. (I am told
|| that the 'A' brings "Improved memory timing and deeper command
|| queues"). The Via specs are at:
|| http://www.via.com.tw/en/apollo/KT266A.jsp
||
|| My memory is 133 MHz SDRAM and the mobo is a Syntax SV266A (see
|| below for more info and links).
||
|| ---

 

[Kevin Lawton]

Sorry to top-post, but this was getting really long.
I've tried disabling USB2 - in fact I've tried disabling all USB - and still
the problem persists.
I've also tried the various 'fixes' for the 'Mup.sys' problem - there are
several different ones - and still no avail. Basically, either disable
Mup.sys via the recovery console or remove hardware to not load some
drivers. Didn't work for me. :-(
Given the prevailing circumstances it appears that, unless the processor is
actually faulty, Windows 2000 is having the XP2400 do something faster than
the m/board can stand it, whereas a slower Duron or a different op system
doesn't push it so hard. Does this sound logical ?
I guess I have two lines of attack now:
1) Try moving the XP2400 into the system which is happily running Windows
2000 on the Duron 1200 in a GA-7VKMLS KM266 m/board - though that can only
use 133 MHz SDRAM.
2) Try underclocking the XP2400 in its present GA-7VRXP KT333 m/board
progressively to see if it will eventually work when slowed right down. I am
thinking of setting both the front side bus and memory bus clocks to 100 MHz
(therefore 200 due to DDR) and gradually taking the multiplier down as far
as I can. I'm happy to try this if it won't hurt the CPU.
Thanks for the help - at least I have something to try.
Kevin.

| On Fri, 16 Apr 2004 17:55:29 +0000 (UTC), "Kevin Lawton"
|
|| To be honest, absolute top performance is 'too rich for my blood' -
|| or to put it another way, only a small gain for a lot more cash. I
|| wanted to build this machine for a reasonable price for video
|| capture and editing. My fastest machine so far was a 1.4 GHz Athlon
|| Thunderbird with 133 MHz SDRAM. I figured an XP2400 with DDR 333
|| memory would give sufficient performance without shredding my
|| wallet. To be honest, rather than overclock I'm happy to just run a
|| system as specced and hope for reliability. .
|
| I don't usually go for top performance either, rather the top
| performance that can be attained _quietly_, without undue expense or
| excessive temp. Running the FSB higher with reduced multiplier is
| only technically an o'c but not really significant of the board and
| CPU are within core margins.
|
|
|| I guess I'll not lose much throughput by running the memory bus at
|| 266 MHz the same as the front-side bus - and will probably get it
|| back as it will be synched.
|
| Often the greatest gain is from bumping the FSB a little bit if
| possible, after setting tightest memory timings possible, though it
| takes a LOT of testing to be sure there's no memory errors... better
| safe than sorry,
| leave a good 5-10% margin for memory.
|
|
|| So could I run the XP2400 with a 333 FSB (166 MHz x 2) and the
|| correct multiplier to sync with my DDR 333 memory on the right
|| m/board ? The GA-7VRXP won't let me set the FSB that high. :-(
|
| Yes, if your memory is PC2700 then any nForce board should do nicely.
| IIRC the XP2400 is 2.0 GHz so that'd translate into 12 X 166. Some
| boards don't allow reducing the 15X default multiplier below 13X
| though, and I don't recall the easiest way to tell which boards do
| and don't allow it,
| so research of a particular board is warranted, or just set it to 13X
| 166 = 2.17GHz... an XP2400 should be able to run @ 2.17GHz easily,
| even if you had to up the vCore to 1.7V. Point being, since Athlons
| have multipler adjustment except for the newer Thortons and desktop
| Bartons, there are a lot of options for which multi & FSB combination
| to use.
|
|
||| How exactly does Win2K err? ALWAYS at the splash screen?
||| You wouldn't happen to have another Windows OS you could try, WinXP
||| or Win98SE?
||
|| Yes, it always hangs at the same place - either after the first
|| re-boot on a fresh install, or when booting a previous installation,
|| the 'Starting Windows' progress stripe completes and then it just
|| hangs - overnight if left that long. At that point it should bring
|| up the Windows 2000 splash screen, but it doesn't. If I choose 'Safe
|| mode with Command Prompt' it shows the last module successfully
|| loaded as 'Mup.sys', so I think it is hanging immediately after that.
|
| That's a really good place to start investigating.
| http://www.google.com/search?q=Mup.sys+freeze+OR+halt
| http://www.earthv.com/tips_detail.asp?TipID=63
|
|
|| DOS works fine, Windows 95 works fine, Windows Me works fine, Linux
|| - Red Hat 9 - works fine, BeOS 5 works after the 'SSE' patch has
|| been installed. It just seems to hang with Windows 2000.
|
| It would seem quite unlikely to be the CPU. If it's just the USB2 as
| one
| of the above links suggests, disable that and retry. You might also
| seek
| a Win2K boot list to get an idea of what might be loading next after
| mup.sys. The thing that makes no sense is that it'd happen with the
| XP2400 but not the Duron, if it were the USB2.
|
||
||| This was an existing Win2K installation on that board with all same
||| components except you'd swapped out the Duron, replaced with the
||| XP2400?
||
|| Yes, I've tried that with every system component in turn - one at a
|| time of course. Duron 650 - all fine, Duron 1200 - all fine, XP2400
|| - Windows 2000 hangs but any other op system seems to work okay. I
|| was wondering if I could try radically under-clocking the XP2400 -
|| or might that harm it ?
|
| Radically? Shouldn't harm it at all. Most motherboard chipsets are
| only stable down to around 50MHz FSB, but I doubt your board even
| supports
| under 100MHz FSB.
|
|
||
||| I'd try clearing CMOS to reset to defaults and leaving it at the
||| defaults except the following:
||
|| Yes, I tried that - same results.
||
||| Does it exhibit same error if you reduce FSB speed to 100MHz,
||| underclock?
||
|| Yes, I have tried setting FSB clock to 100 MHz and the results were
|| the same.
|
| Try disabling all onboard features, USB, sound, etc, etc.
|
||
||| I would definitely leave the memory in synchronous mode, 133MHz (or
||| 100MHz if you underclock the FSB to 100MHz), for any further
||| testing,
||| and try with only one memory module.
||
|| I have tried bringng the memory bus clock down to the same as the
|| CPU clock, and have tried several different memory strips - all
|| PC2700 / DDR 333. Various makes of memory - some 'Crucial' and
|| others. All the same results again.
||
||| What about the hardware monitor screen in the BIOS, or if you have
||| the ability to use a multimeter to measure CPU vCore on the board,
||| does it seem appropriate?
||
|| I think so - I will get the specs off te AMD site and double-check.
|| I have tried two different (or identical) GA-7VRXP boards and two
|| power supplies at diffeerent times (only change one thing at a time)
|| started with a 300w and moved to a 550w. Would you suggest altering
|| the vCore setting, or might that damage the CPU ?
||
||| It's not a generic 550W PSU, is it?
||
|| No, I use branded CPUs. I did also try a 450w Antec at one point,
|| but that was destined for a different machine.
|| Given that I'm only running a few components - one hard drive, etc,
|| I would have thought that the 300w would be ample and the 550w
|| overkill.
|
| Yes, that's enough power but I was just trying to rule out overrated
| power supplies... some generics marked as 550W are only worth 250...
| the Antec 450W should've been more than enough power.
|
|
||
|| I supose it's possible the CPU is defective, but in general and
||| especially considering that it ran other OS, seems unlikely.
||
|| That's what I figured. If I could find any CPU diagnostic software
|| to run under dos, Windows Me or Linux, then I could be conclusive
|| about the CPU without having to buy another to swap with it. Any
|| ideas here ?
|| TIA
|| Kevin.
|
| Prime95? There are a number of programs claiming to do a diagnostic
| of a whole system but nothing comes to mind that would stress test a
| cpu using all possible instructions. Still, if the system runs
| flawlessly on WinME (or relatively so, it IS WinME!) it would more
| likley be something onboard the Gigabyte motherboard causing the
| problem, a feature, not the KT333 chipset since those are quite
| common and not reported in any large numbers to be problematic with
| Win2K.

 

[CBFalconer]

Sorry to top-post, but this was getting really long.
I've tried disabling USB2 - in fact I've tried disabling all USB
- and still the problem persists.

.... snip ...

So snip the non-germane parts.

 

[kony]

Given the prevailing circumstances it appears that, unless the processor is
actually faulty, Windows 2000 is having the XP2400 do something faster than
the m/board can stand it, whereas a slower Duron or a different op system
doesn't push it so hard. Does this sound logical ?
I guess I have two lines of attack now:
1) Try moving the XP2400 into the system which is happily running Windows
2000 on the Duron 1200 in a GA-7VKMLS KM266 m/board - though that can only
use 133 MHz SDRAM.

That sounds reasonable, and also you could try the XP2400 in the VRXP
with FSB at 100MHz but higher multiplier so resultant CPU speed is
comparable, would come nearer isolating the chipset and memory bus
frequency, difference.

2) Try underclocking the XP2400 in its present GA-7VRXP KT333 m/board
progressively to see if it will eventually work when slowed right down. I am
thinking of setting both the front side bus and memory bus clocks to 100 MHz
(therefore 200 due to DDR) and gradually taking the multiplier down as far
as I can. I'm happy to try this if it won't hurt the CPU.

I'm surprised you haven't tried that yet. It's one of the first things to
try when there's a question of (is it the hardware or an OS/file/etc
[software or data] problem).

You might also try rasing the vCore in the bios by (think it's 7.5 %).

It is completely safe to underclock the CPU, even down to the lowest the
board supports which might be around 5 x 100 FSB = 500MHz, though that's
hardly necessary, reducing the FSB to 100MHz is typically enough given
that the system was already running another OS ok.

I'm not able to keep track of what you're tried (or I'm forgetful) if you
don't make mention of other things discussed, like whether you'd tested
the system with Memtest86.

There are some modified bios versions for that board, with some hidden
options available. I don't know if any of them would be helpful but if
so, here they are:
http://forums.amdmb.com/showthread.php?s=&threadid=132144

It seems that some revisions of that board might have some capacitor
problems too...
http://community.webshots.com/photo/81256449/120878185dbGvRR
If it's using GSC capacitors
http://community.webshots.com/photo/81256449/120878097oDJsMf

 

[Kevin Lawton]

| Kevin Lawton wrote:
||
|| Sorry to top-post, but this was getting really long.
|| I've tried disabling USB2 - in fact I've tried disabling all USB
|| - and still the problem persists.
| ... snip ...
|
| So snip the non-germane parts.

Aw, c'mon CBF - if I'd have snipped out everything which had gone before
then what I did write would have just stood alone without a context.
I know you could counter that by suggesting that an interested reader can
just read earlier messages in the thread - but people often don't do this.
They'll join in a discussion after it has been going for some time and throw
in a comment, or ask a question, which was covered earlier.
That's why I top-posted. Not my usual habbit, I know, but justified on this
occasion I think.
Anyone who had been following the thread could just read the top bit and
keep up to date, whereas a newcomer - or someone with a bad memory (like
me) - could quickly refer to what had gone before even if their news server
had already deleted some earlier messages. Anyway, I did snip some stuff out
from earlier on, and you'll often see posts from me with a few '<snip>' in
them.
Now, what ideas have you got for helping to fix my problem ?
Kevin.

 

[Piotr Makley]

... snip ...

So snip the non-germane parts.

Heads down everyone. This here could be the beginning of one of
those lively top posting threads.

 

[Piotr Makley]

| Kevin Lawton wrote:
||
|| Sorry to top-post, but this was getting really long.
|| I've tried disabling USB2 - in fact I've tried disabling all
|| USB - and still the problem persists.
| ... snip ...
|
| So snip the non-germane parts.

Aw, c'mon CBF - if I'd have snipped out everything which had
gone before then what I did write would have just stood alone
without a context. I know you could counter that by suggesting
that an interested reader can just read earlier messages in
the thread - but people often don't do this. They'll join in a
discussion after it has been going for some time and throw in
a comment, or ask a question, which was covered earlier.
That's why I top-posted. Not my usual habbit, I know, but
justified on this occasion I think.

In my book you did right, Kevin. It's always a bit of a balancing
act. Don't let it get to you.