VOGONS


First post, by Briggz

User metadata
Rank Newbie
Rank
Newbie

Hi there, first time poster long time lurker 😀

I have an old socket 7 motherboard that I've been working on for a while now, I just recently upgraded the cpu but it doesn't appear to have had a noticeable impact on system performance.
For years I had an AMD k6-2 500mhz cpu, now I've upgraded to a k6-2+ 500mhz cpu.
The FPS results in 3dmark99 are almost the same using the new k6-2+, there is still noticble stuttering in the 2nd 3D test and the 3rd synthetic cpu test appears to run slower than when using the old k6-2.

System specs.
Motherboard: Asus P55T2P4 rev 3.10
BIOS: patched bios to support k6-2+ from http://web.inter.nl.net/hcc/J.Steunebrink/k6plus.htm
Memory: 256mb (4 x 64mb 72 pin EDO modules)
Video card: Voodoo 3 2000 PCI using amigamerlin 2.9 drivers (no overclocking, stock 143mhz)
HDD: old Maxtor 6GB IDE, still works perfect
OS: Windows 98SE clean install
New atx power supply (cut & soldered old AT power plug onto new psu) (missing -5v wire) 5v rail supports high amps
Replaced all electrolytic caps with new low esr caps
Did the cut and grind trick on the cmos time clock to hold new 2032 battery
I have the COAST module to expand up to 512mb ram but I've left it out for now.

CPU clock: Multiplier is 6, FSB is 83mhz, core voltage is 2.2v (picked up from kalles page)
Using CTU.exe to enable write combining (got the correct addresses using setk6.exe)

Here are some benchmark results,

AMD k6-2/500
PCmark2002 scores
CPU score: 771
Memory score: 467
HDD score: 223

3Dmark99MAX
1874 3D marks
4601 CPU marks

AMD k6-2+/500 (500ACZM, 2.0V)
PCmark2002 scores
CPU score: 819
Memory score: 514
HDD score: 234

3Dmark99MAX
2244 3D marks
4934 CPU marks

I've actually had the old k6-2 cpu running higher scores in 3dmark before but I just couldn't replicate it this time.

Side note: the k6-2 cpu I use is in conjunction with a TAG ram IC so the system can cache the full 256mb (k6-2 has no L2 cache)
The k6-2+ I don't use the TAG ram IC as it is able to cache the system ram using its own on die L2 cache (I disabled onboard L2 cache in BIOS)

I have confirmed using CTCM the new k6-2+ is caching the whole 256mb of ram - no uncacheable areas found
Whats weird though is CTCM says- write strategy L1: Write back, write allocation, linear fill, unknown-LRU (on the k6-2 it was something else)

Another weird thing while using the k6-2+ is setk6.exe says write allocation is enabled but then says its disabled at the bottom of the screen.
Example, I can disable write allocation using the /off command
I can re-enable using /on:256 command, however after I press enter is reports write allocation disabled?
Running CTU.exe or k6speed.exe under windows both report write allocation being enabled even when setk6 says disabled?

So.... maybe I got a bad cpu or I just set my expectations too high for the k6-2+
After looking at the benchmark results from philscomputerlab I thought my new k6-2+ would boost performance significantly?

P.S cheers to philscomputerlab for providing alot of useful information to me and the vogons message board

Reply 1 of 16, by doaks80

User metadata
Rank Member
Rank
Member

It's amazing you researched the plus models of the K6 and came to the conclusion that they are valued because they boost performance.

k6-3+ 400 / s3 virge DX+voodoo1 / awe32(32mb)
via c3 866 / s3 savage4+voodoo2 sli / audigy1+awe64(8mb)
athlon xp 3200+ / voodoo5 5500 / diamond mx300
pentium4 3400 / geforce fx5950U / audigy2 ZS
core2duo E8500 / radeon HD5850 / x-fi titanium

Reply 2 of 16, by brostenen

User metadata
Rank l33t++
Rank
l33t++

Have you tried running 3D-Mark-99 ? I think you will get a totally different score. 😉

Don't eat stuff off a 15 year old never cleaned cpu cooler.
Those cakes make you sick....

My blog: http://to9xct.blogspot.dk
My YouTube: https://www.youtube.com/user/brostenen

001100 010010 011110 100001 101101 110011

Reply 3 of 16, by meljor

User metadata
Rank Oldbie
Rank
Oldbie

Running without L2 makes a ''+" cpu on a super socket 7 board about 5% slower. I guess it's the same on your Intel board so keep it enabled!

And even with that bit of performance loss it's STILL a good 5%-10% faster as your old scores. On a good day (right bench/game) the k6-2+ kan be around 20% quicker compared to the normal k6-2. The k6-3+ can hit 30% and maybe even a bit higher in some cases. On a super 7 the extra bonus is the overclocking headroom, making 550/600mhz a breeze.

So: You can't overclock it, you run it without L2, and it's the k6-2+ not the k6-3+..... I think your cpu is performing fine!

Also, thinking that a stuttering part of the benchmark will be fixed by a faster cpu is a tiny mistake: let's say it runs 12fps and you place a 20% faster cpu. It's max would still only be 14fps and since your ram, video etc. is not getting the 20% boost it will probably be only 13fps. Not really a jump in performance.

asus tx97-e, 233mmx, voodoo1, s3 virge ,sb16
asus p5a, k6-3+ @ 550mhz, voodoo2 12mb sli, gf2 gts, awe32
asus p3b-f, p3-700, voodoo3 3500TV agp, awe64
asus tusl2-c, p3-S 1,4ghz, voodoo5 5500, live!
asus a7n8x DL, barton cpu, 6800ultra, Voodoo3 pci, audigy1

Reply 4 of 16, by Intel486dx33

User metadata
Rank l33t
Rank
l33t

Phils computer lab has some programs and a CPU tools CD you can download and burn.
Has some good tools for benchmarking CPU, Cache, BUS, Memory, Harddrives, etc.

https://www.philscomputerlab.com/k6-2-2-3-resources.html

Burn to cdrom dosbench. Boot off dos diskette and then run these tools.
https://www.philscomputerlab.com/dos-benchmark-pack.html

Reply 5 of 16, by Briggz

User metadata
Rank Newbie
Rank
Newbie

Yeah I got exited when I saw a few benchmarks between the 2 different CPU's and since I'm stuck at 500mhz (6 x 83mhz is the max) I thought the k6plus would be a worth while upgrade.
The thing I'm still stuck on is the minor inconsistencies reported by ctcm and setk6?
I've tried setk6v2 and v3 and get the same result.

I don't think putting more ram in the system will boost fps much further, maybe a better video card but I love my voodoo 3.

BTW does anybody know what the 'm' in ACZM means? As far as I know its exactly the same as a k6 ACZ?

Thanks for the link to the dos benchmarks, I keep finding new info every time I go to Phil's website.

Reply 8 of 16, by dionb

User metadata
Rank l33t++
Rank
l33t++
Briggz wrote:

I'll test with the mobo cache enabled and see if it makes any difference.
I'm guessing I'll have to put the tag ram IC back in? Or will it conflict with the k6-2+?

Nope, it purely relates to the L2 (now L3) cache. The K6-2+ tags its own cache.

Reply 9 of 16, by Briggz

User metadata
Rank Newbie
Rank
Newbie

I've put the tag back in and re enabled external cache in the bios however cpu-z and crystalcpuid both have L3 cache greyed out? Ctcm doesn't show L3 either?
I know the onboard cache works because it was showing when I had the old k6-2 in?

I'll run some benchmarks now.

Reply 10 of 16, by appiah4

User metadata
Rank l33t++
Rank
l33t++

I believe your motherboard can already cache >64MB RAM, if that's indeed the case you really had no use for a K6-2+ to be honest, bad investment..

Retronautics: A digital gallery of my retro computers, hardware and projects.

Reply 11 of 16, by xjas

User metadata
Rank l33t
Rank
l33t
appiah4 wrote:

I believe your motherboard can already cache >64MB RAM, if that's indeed the case you really had no use for a K6-2+ to be honest, bad investment..

Not really, on-chip L2 cache is much faster than 83MHz motherboard L2 cache.

twitch.tv/oldskooljay - playing the obscure, forgotten & weird - most Tuesdays & Thursdays @ 6:30 PM PDT. Bonus streams elsewhen!

Reply 12 of 16, by dionb

User metadata
Rank l33t++
Rank
l33t++
xjas wrote:

Not really, on-chip L2 cache is much faster than 83MHz motherboard L2 cache.

Indeed, it's the other way round - the added value of external cache is much smaller with a CPU with on-die L2.

This topic inspired me to order a K6-2 500. I already have a K6-3+, and expect a K6-2+ any day by post. Even though a K6-2+ isn't a speed wonder, I'm puzzled by the minimal speed delta here. I have a nice P5A, so can try lots of FSB options...

Reply 13 of 16, by Briggz

User metadata
Rank Newbie
Rank
Newbie

I can confirm enabling the onboard cache does help performance wise, I've only run 1 benchmark so far...

3Dmark99max
2385 3D marks
5137 cpu marks

Already an improvement compared to no onboard cache.

I just downloaded a benchmark package from this site which contains a lot more programs to try out, so I'll give them a shot also.

I think there was no such thing as L3 cache on asus motherboards back in 1996? so maybe that's why cpu-z reports no L3 cache? just a guess

Reply 15 of 16, by meljor

User metadata
Rank Oldbie
Rank
Oldbie

Good to see your performance is up.

L2 on the motherboard becomes L3 when the cpu already has L2 of his own. So your Asus does have L3. I don't know why it isn't showing as enabled but as long as it works..... that's the main thing right?

And you talked about maybe a better graphics card: the voodoo3 is already too fast and is simply waiting for the motherboard/cpu to give it instructions so any card that's even faster is really a waste.
Even a single voodoo2 will not get fully up to speed on a k6 plaform. 3dfx cards do however work fine and are preferred on such platforms.

asus tx97-e, 233mmx, voodoo1, s3 virge ,sb16
asus p5a, k6-3+ @ 550mhz, voodoo2 12mb sli, gf2 gts, awe32
asus p3b-f, p3-700, voodoo3 3500TV agp, awe64
asus tusl2-c, p3-S 1,4ghz, voodoo5 5500, live!
asus a7n8x DL, barton cpu, 6800ultra, Voodoo3 pci, audigy1

Reply 16 of 16, by Briggz

User metadata
Rank Newbie
Rank
Newbie

Just a quick update,
I changed a few ram settings in my BIOS last night and have received quiet a big performance increase in 3dmark99.
All these years I've been using the auto 60ns settings in my award bios, I remember playing with these settings at one stage and getting some instability under windows 98 so I just reverted back to default auto settings and never really looked into it again.

RAM sticks: 4 x 64mb 72 pin EDO 60ns

Auto configuration: Disabled
DRAM read burst timing: x222
DRAM write burst timing: x222
RAS to CAS delay: 2T
DRAM R/W leadoff timing: 6T/5T
DRAM: turbo read leadoff: disabled
DRAM speculative leadoff: enabled
Turn around insertion: disabled
Turbo read pipelining: disabled

Before ram setting changes my best 3dmark scores were,
3Dmarks: 2388
CPUmarks: 5164

After with new ram settings 3dmark99 max scores,
3Dmarks: 2874
CPUmarks: 6070

A nice improvement 😀
I sat down last night and played need for speed 3 and shadows of the empire, both games perform incredibly well and I experienced no instability or crashing under the new ram settings.

Side note:
I e-mailed Jan Steunebrink who wrote all of the patched bios files to support K6-2+ & k6-3+,
I asked why L3 cache was not being shown in ctcm and cpu-z for example and his reply was those utilities only show L1,L2,L3 cache of the cpu and not whats on the motherboard. He said an older version of cpu-z might of shown motherboard cache but he's not sure.
I ran sandra sisoft 99 and it does indeed show my motherboard cache size and that it is present.