[East 17]

I've started this topic to see if there is possible to steal and adapt one of the WinXP drivers for K6 & K7 CPUs from AMD and make it work with Win98 ...

While doin some benchmarking I've noticed that there was a 20 to 30 % performance difference between Win98's results and WinXP's results .... WinXP being faster in finishing the SUPER PI 8M benchmark and ... that could only be directly related to the K6 and K7 CPU drivers in WinXP bacause the result repeated itself althogh I've tried with several reinstalls of Win98 with different combinations of drivers and setting ...

Can this be done? ... Steal and adapt a K6 & K7 CPU driver from WinXp and adapt it for Win98SE?

What do you think ?

Will you try it ?

[eidenk]

Quote

I've started this topic to see if there is possible to steal...

Can this be done? ... Steal...

Will you try it ?

You have choosen the verb "steal" on purpose I think.

[miko]

Quote

AMD Athlon™ XP Processor PowerNow! Software for Windows 98SE, Windows ME, Windows NT 4.0 & Windows 2000 Version (exe) 3.0.4

http://forums.amd.co...showtopic=50561

don't know much about this myself but it looks like there might be 98SE ones already if thats what you mean, other than that no one around here will be stealing anything (well maybe mitchellovision & dirtwarrior, but no one else dammit)

[East 17]

eidenk, on Jul 22 2005, 02:23 PM, said:

Quote

I've started this topic to see if there is possible to steal...

Can this be done? ... Steal...

Will you try it ?

You have choosen the verb "steal" on purpose I think.
<{POST_SNAPBACK}>

I'm reading your post with a big smile on my face ... Although I hope I haven't ofended anyone.

Why "steal"? It was a choice of words ... random... spontaneous ... just that .

[East 17]

miko, on Jul 22 2005, 11:28 PM, said:

Quote

AMD Athlon™ XP Processor PowerNow! Software for Windows 98SE, Windows ME, Windows NT 4.0 & Windows 2000 Version (exe) 3.0.4

http://forums.amd.co...showtopic=50561

don't know much about this myself but it looks like there might be 98SE ones already if thats what you mean, other than that no one around here will be stealing anything (well maybe mitchellovision & dirtwarrior, but no one else dammit)
<{POST_SNAPBACK}>

Hmm ... you are talking about the PowerNow! technology ... but I was refering to the fact that the manifesting performace of my XP2000+ CPU differs by 25% from a CLEAN Win98SE install equiped with the best drivers and a CLEAN WinXPSP2 instal also with the best drivers in a simple benchmark that only uses the mathematical units in the CPU.

SuperPI calculates 8M in 12 minutes in WinXPSP2 and in 15minutes in Win98SE ... that's my riddle...

This post has been edited by East 17: 28 July 2005 - 05:24 PM

[zivan56]

You probably just have the specifc chipset drivers for your motherboard installed by default in XP, whereas in 98 it uses generic ones. This makes a dramatic impact on memory performance and bus througput.

This post has been edited by zivan56: 28 July 2005 - 05:28 PM

[Rhelic]

Disclaimer: I'm not 100% sure on any of this and please correct me if I'm not but...

There's no way to adopt NT5 based drivers (2000,XP,2003,Vista) for use in 9x (95,98,ME). IIRC things are VERY different under the hood (aren't 9x drivers 16bit and NT5 drivers 32bit?)

Plus, I believe 98 uses lots of generic drivers for motherboard chipsets and cpus, I don't think 98 uses specific drivers.

[miko]

WIN98SE shares the WDM (Windows Driver Model) with Windows 2000 (NT5)
that's about as far as it goes, in saying that it is quite signaificant,
afaik no other 9X OS (except maybe Me) does

some reading (pulled from Google) -
http://www.gdcl.co.uk/wdmdetails.htm
http://www.phdcc.com/WDMarticle.html

Windows 98SE drivers are not all 16bit,
remember it's a 16/32 bit OS not just 16bit,
but then in some ways thats what makes it such fun.

as for motherboard drivers, yes 98SE will use generic ones if you don't go and download the right ones (usually called motherboard 'inf' drivers) from the manufacturer, which is exactly the same as XP does. XP will however contain drivers for chipsets produced up until its realease 'out the box' as it were, 98SE is the same only it came 'out the box' quite a while ago so now doesn't recognise most of what it 'sees' without a downloaded driver (or two).

none of this is relavent to this thread i think
goes off to play some DOS games

This post has been edited by miko: 02 August 2005 - 03:10 PM

[eidenk]

Quote

WIN98SE shares the WDM (Windows Driver Model) with Windows 2000that's about as far as it goes, in saying that it is quite signaificant,
afaik no other 9X OS (except maybe Me) does

Yes ME does, I thought 98SE didn't. Anyway I have tried to install WDM drivers for a BT 878 card. They don't work. I need to install Brooktree's vxds for it to work. So I don't think the WDM architecture is fully implemented on 9x.

[miko]

well my soundblaster 5.1 digital driver is WDM,
and works fine, maybe its an implementation thing,
there are some previsos on WDM use in 98SE i just can't remember what they are
i think having a correct ACPI installation might have something to do with it.

This post has been edited by miko: 02 August 2005 - 02:55 PM

[East17]

Hello everybody,

I'm back with the same problem .

Sure I'm using the latest nVIDIA drivers on a clean Win98SE instal with Unoficial SP 2.1 still when I benchmark in WinXP (SP1 or 2 it doesn't matter) I get a score around 8 minutes for calculatin SuperPI 8M while in Win98SE I get around 13 minutes .

That's a huge 62% slower in Win98SE .

Why ? It can't be the drivers , I'm using the latest ones .

It can't be a patch that I've forgot to apply , I'm using Unoficial SP2.1 .

I'm sure it's about a specific procesor driver as only WinXP comes with K7.sys and K6.sys and Win98SE doesn't but the problem is no more if an Intel CPU is involved.

Yes, I've benchmarked a system with a Celeron 3 (a Coppermine derived Celeron) in Win98SE and WinXP and the results are quite close . Why ? Because there is a Pentium driver in Win98 .

Also , I was benchmarking on a KT266A before and I using an Abit NF7-S 2.0 now . So it can't be a particularity of the chipset or mainboard.

[ssmokee]

98se driver performance hasnt been a priority at nvidia for ages, whereas XP has been in the forefront for some time now. If you used a motherboard that was in its prime when 98se was (probably something with a VIA KT266a or SiS 735 chipset) I bet you would see much closer results as you did with the celeron system.

[East17]

ssmokee, on Jan 22 2006, 07:27 AM, said:

98se driver performance hasnt been a priority at nvidia for ages, whereas XP has been in the forefront for some time now. If you used a motherboard that was in its prime when 98se was (probably something with a VIA KT266a or SiS 735 chipset) I bet you would see much closer results as you did with the celeron system.

Just read my post ... I was using KT266A when I started testing this issue . Now I'm using NF2 .

Also , I was using , what VIA calls Win98SE optimised drivers , that's 4.43 version . Also I've tried with the latest Hyperion but still the same ...

[RJARRRPCGP]

eidenk, on Aug 2 2005, 03:44 PM, said:

Quote

WIN98SE shares the WDM (Windows Driver Model) with Windows 2000that's about as far as it goes, in saying that it is quite signaificant,
afaik no other 9X OS (except maybe Me) does

Yes ME does, I thought 98SE didn't. Anyway I have tried to install WDM drivers for a BT 878 card. They don't work. I need to install Brooktree's vxds for it to work. So I don't think the WDM architecture is fully implemented on 9x.

Windows 98 SE added WDM support. So Windows 98 SE still has high compatibility.

[LLXX]

I think this is a Microcode Update problem. A CPU doesn't have a "driver" in its proper sense; the K6.sys and K7.sys are pseudo-drivers that provide microcode updates for the CPU.

I know that on Intel boards, the BIOS loads microcode updates on startup and the updates are integrated into the BIOS. Maybe true for AMDs as well. Try updating your BIOS.

[RJARRRPCGP]

LLXX, on Jan 29 2006, 04:04 AM, said:

I think this is a Microcode Update problem. A CPU doesn't have a "driver" in its proper sense; the K6.sys and K7.sys are pseudo-drivers that provide microcode updates for the CPU.

I know that on Intel boards, the BIOS loads microcode updates on startup and the updates are integrated into the BIOS. Maybe true for AMDs as well. Try updating your BIOS.

Updating the BIOS probably won't fix that, because Asus isn't focused on that, they don't have anything saying that it adds a microcode update. It's only required if you have a Sempron, according to Asus. Also, the A7N8X motherboard family has a reputation of BIOS corruption when flashing the BIOS.

[Petr]

Juat an idea - what about disk drivers?

SuperPi extensively uses the hard disk, so maybe there could be the difference? Do you have DMA switched on?

Is the raw disk speed measured by HD Tach 2.70 the same in Windows 98 SE and in Windows XP?

Is the filesystem the same?

Is the caching the same?

What about to try it to run on RAMdisk?


Second idea - I don't know K7 processors, but K6 processors had some optimizations that was necessary to switch on in BIOS, and if BIOS did not supported these features, there was a separate utility called SetK6.

Some internal settings of both Intel and AMD processors can be found in Everest, just right-click on the bottom frame of the Everest windows, there are things like CPUID and MSR dump. The latest beta version has also additional item in Tools menu - cache and memory benchmark.

You can play with these tools and maybe you will be able to find the weak point.

Petr

[East17]

I using a WD 80GB ATA100 Drive w/ 8MB Cache

DMA is ON . HDD-Tach is showing a Transfer Rate Graph between 60MB/sec and 30MB/sec . The graph is clean ,no spikes , so the drive runs normaly . The Average Read Rate is 50MB/sec with 9% CPU utilisation.

So it's not the HDD .




Also ... I thought that "CPU Microcode" section in BIOS is only related to the functional specification of a new revision of the CPU like : the new name of the CPU , the multiplier and FSB setting and the voltage .

[RJARRRPCGP]

Petr, on Feb 1 2006, 02:47 PM, said:

Second idea - I don't know K7 processors, but K6 processors had some optimizations that was necessary to switch on in BIOS, and if BIOS did not supported these features, there was a separate utility called SetK6.

The FTP server appears to be down.

This post has been edited by RJARRRPCGP: 06 February 2006 - 02:25 PM

[Petr]

RJARRRPCGP, on Feb 6 2006, 09:25 PM, said:

SetK6.

The FTP server appears to be down.

It works fine now.

Petr