VOGONS


Finally got my K6-2+ 500, did some Benchmarks.

Topic actions

First post, by maddmaxstar

User metadata
Rank Member
Rank
Member

So, a couple weeks ago I ordered a K6-2+ 500 off of eBay for my Super7 system "Titania". It originally had a K6 266MHz in 1998 but it's had a K6-2 500 in it since early 2001. I've always been curious what real world performance that the K6-III and K6-2+/III+ chips offered over the original K6's, but never got around to actually getting one until recently because they're hard to find. Most here likely know but for those who don't, the K6-2+ is a cut down K6-III+ with 128k on-die L2 Cache instead of 256k. From what I've heard there isn't a huge difference between the II+ and III+ so I wasn't worried about getting the cheaper chip, though I'd love to have a faster one for my collection in the future.

Anyway, while I had the SS7 system apart, I decided to do a bit of Benchmarking using Quake II and 3DMark 2001se.

So, here are the results, what you're about to see is the result of hours of possibly pointless work and testing:

Test system specs:

6699003811_e805bffa62.jpg
FIC VA-503+ BabyAT Super7, VIA MVP3 Chipset, 1MB L2 Cache, JE4333 BIOS (support for K6-2+/III+)
256mb PC100 RAM (2x128)
6.4GB Fujitsu MPC Hard Drive
3DFX Voodoo3 2000 16mb AGP (3DFX v1.07 Driver)
15" Dell LCD Monitor (1024x768@75Hz)
Windows 98 Second Edition.

Software Setup:
MadOnion 3DMark2001 Second Edition, full version.
Textures converted to 16bit during install, as the Voodoo3 supports maximum 16bit colour.

Quake II v3.20 with 3DNow! Patch.
Added "+set s_initsound 0" to shortcut to disable sound.
Tests run on the included demo tests, Demo1.dm2 and Demo2.dm2, 3 runs on each.

All tests run at 1024x768.

= K6-2 500MHz =
Quake II:
Demo1.dm2: 54.7 / 54.9 / 54.9
Demo2.dm2: 55.4 / 55.7 / 55.8

I ran 2 runs of 3DMark 2001 SE to be sure. The tests are excruciatingly slow, so I ran only 1 run each from here on.
6739901243_513f8ff1d3_z.jpg

= K6 300MHz (3x100) =
This chip is actually my original K6 266MHz 2.2v chip from 98. I knew I could overclock it to 350 at normal voltage/cooling on 100MHz FSB, so instead of screwing with the gawdawful amount of jumpers on the VA-503+, I went lazy and just set it to 3x100 and ran the test.

Quake II:
Demo1.dm2: 32.0 / 32.2 / 32.2
Demo2.dm2: 31.9 / 32.1 / 32.1

This one was fragging slow, eeesh!
6739901447_d8efb7c714.jpg

= K6-2 300MHz (3x100) =
For fun, to draw a comparison of how much difference there was between the K6 and K6-2, I dropped in my K6-2 333 chip from my CPU collection and left the 3x100@2.2v the same. I found the performance difference surprising!

Quake II:
Demo1.dm2: 44.0 / 44.5 / 44.3
Demo2.dm2: 44.1 / 44.3 / 44.5

Unfortunately I forgot to grab a screenshot of the 3DMark2001 results, but I did write them down.
SCORE: 414

Now for the moment I'd been waiting for, the K6-2+!
= K6-2+ 500MHz =

Quake II:
Demo1.dm2: 67.0 / 67.0 / 66.9
Demo2.dm2: 64.9 / 65.4 / 65.4

I also ran the Quake II tests with the 3DNow! enhancements off.
Demo1.dm2: 59.6 / 59.9 / 59.9
Forgot to run Demo2.dm2, but you get the idea. The K6-2+ results with 3DNow off are faster than the K6-2 with 3DNow on.

6739900733_19c2f9431a.jpg

These results I thought were fantastic. about a 10FPS increase in Quake II and a big boost on 3DMark just by adding a little bit of L2 cache. In the end, I'd have to say that anyone who's considering building a K6-2 based system, that upgrading it to a cheap K6-2+ (I got mine for 12$) is a worthwhile upgrade for that extra boost in performance, especially when there are games that'll put much more of a strain on the system than Quake II. Just be sure your board can supply 2.0 volts and has a patched BIOS available for K6-2+/III+ support.

And for sh*ts and giggles, I threw my Slot A Athlon "Titania-II" up on the bench.

Specs:
FIC SD11 Slot A Motherboard (AMD 751+VIA 686A)
512mb PC100 RAM (2x126+1x256)
20GB WD200BB 7200RPM HDD
3DFX Voodoo3 3000 16mb
Windows XP SP2

Athlon 550MHz (0.25um Argon)
Quake II:
Demo1.dm2: 83.6 / 84.2 / 84.2
Demo2.dm2: 84.7 / 84.8 / 84.6
Then, I installed the 3DNow patch, ran it again.
Demo1.dm2: 84.2 / 84.2 / 84.2
Demo2.dm2: 84.9 / 85.0 / 84.9

Originally I had 3Dhq 1.09beta drivers installed, however, 3DMark crashed when running, probably a buggy D3D Driver. So I uninstalled them and installed AmigaSport v3.0f for Voodoo3 cards. I reran the Quake II tests, but this had the unfortunate side effect of locking out the V-sync, so all tests pretty much came at around 75fps, due to the 75Hz refresh rate. However, the 3DMark test did run.

6740759959_7f28188f5e.jpg

Athlon 700MHz (0.18um Orion)
I popped in another Slot A chip I had lying around for testing, I had to blow on it like an NES cart, but it worked. Unfortunately the QuakeII tests were eschewed due to the refresh rate, but the 3DMark01 test ran fine.

6740759093_de187f5461.jpg

And just to put everything in perspective, I ran the tests on my Main PC.

Specs:
AMD Phenom II X6 1090T (6 Core, 3.2GHz with 3.6GHz Turbo)
Asus M5A88-E AM3+ AMD880G mATX Board
16GB (4x4) DDR3-1600 Kingston HyperX Genesis RAM
500GB Seagate 7200.11 SATA2 HDD
Sapphire 512mb ATI Radeon 4850 PCIe

Quake II:
Demo1.dm2: 804.9 / 793.8 / 903.0
Demo2.dm2: 682.9 / 886.7 / 886.7

6740965853_b026586c05.jpg

How far we've come. Thanks for reading!

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =

Reply 1 of 67, by ratfink

User metadata
Rank Oldbie
Rank
Oldbie

Interesting to see this, I've just put together a slot A box and I'm wondering about the comparison to my k6/3 [very similar to k6/2+ if not the same]. Your results show quite a performance jump to slot A for the same clockspeed - about double the 3dmark2001 score for the same clock speed.

OT: Makes me wonder whether there's any need for my k6/3 system as it stands; I think I'd better switch to a + model so I can slow it down or even a k6/2 400 as I think the only use compared to the slot A is to give 386/486-level performance. Seems like as it is, disabling L1 gives 386 speed but disabling L2 makes little difference to speed-sensitive dos games. The slot A behaves the same but maybe a different k6 chip will be better - I don't need the top end speed on the k6 anyhow.

Reply 2 of 67, by noshutdown

User metadata
Rank Oldbie
Rank
Oldbie

there's another important benchmark to do: superpi!
i had a k6-2+500 and ali5 mainboard(gigabyte 5ax rev4.1), and i remember i got around 5:50 or so, how does it perform on your mvp3?
superpi-1.5mod can't run in windows98, so i used an old version without decimal. i have seen pics of superpi version with digits that works in windows98 but couldn't find it, can you find it?

Reply 3 of 67, by ProfessorProfessorson

User metadata
Rank Member
Rank
Member

I told my Phenom II about what your Phenom II did, and just to put everything in perspective it said to post this :

3dmark2k1.png

😜

On a serious note, the test results of the K6-2+ really were a big improvement over the K6-2. 5 to 10 fps leaps may not seem like much now to some people since hardware is cheaper and almost everything runs at 60 fps or above constantly now, but back then it could really make or break the gameplay of a title that was pushing the system hard.

Reply 4 of 67, by keropi

User metadata
Rank l33t++
Rank
l33t++

just did the 2001se benchmark on my 2600k/560ti and it scored ~63000 🤣

🎵 🎧 PCMIDI MPU , OrpheusII , Action Rewind , Megacard and 🎶GoldLib soundcard website

Reply 5 of 67, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
noshutdown wrote:

there's another important benchmark to do: superpi!
i had a k6-2+500 and ali5 mainboard(gigabyte 5ax rev4.1), and i remember i got around 5:50 or so, how does it perform on your mvp3?
superpi-1.5mod can't run in windows98, so i used an old version without decimal. i have seen pics of superpi version with digits that works in windows98 but couldn't find it, can you find it?

There is one version of superpi that has decimals and works in 9x. It has 2 bugs though:
1)It crashes the system after it's first run, but works perfectly fine after that (it will remember that score though)
2)It sometimes displays seconds weird (hard to explain).

But despite these bugs, I still use this superpi. I'll see if I can upload it somewhere or find it, the bugs aren't serious enough for me to not use it

Edit:Btw, do we have a 3dmark2001se thread or something? 2001se is one of my standard benchmark utils along with superpi 😜

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 6 of 67, by ratfink

User metadata
Rank Oldbie
Rank
Oldbie
Tetrium wrote:

Edit:Btw, do we have a 3dmark2001se thread or something? 2001se is one of my standard benchmark utils along with superpi 😜

Might be fun. I just did my athlon 700 with a 3dlabs vp870 and got 4384, but my phenom 9350e with hd4850 only got about 19600. Pretty sure my p4 640 + fx5950u did 15-16000 so that phenom/radeon combo seems a bit underwhelming.

Reply 7 of 67, by maddmaxstar

User metadata
Rank Member
Rank
Member
ratfink wrote:

Might be fun. I just did my athlon 700 with a 3dlabs vp870 and got 4384, but my phenom 9350e with hd4850 only got about 19600. Pretty sure my p4 640 + fx5950u did 15-16000 so that phenom/radeon combo seems a bit underwhelming.

Well, that could be just from the much higher clock speed of the P4 940 (3.2GHz) than of the 9350e (2.0Ghz). I doubt 3DMark01 was written for anything more than 1 or 2 cores. (and maybe the Radeon isn't as suited for DX7/8 graphics 😜)

I figured there'd be a few people out there with faster systems to post higher scores, but for what it's worth, I'm impressed with the longetivity of my Radeon HD 4850, I've had it since Summer '08 (just after release) and the thing still runs everything I throw at it, Dirt 3 and Portal 2 run beautifully at max settings. Now if I could only become a better Rally driver (staying on the road would help).

I've never used SuperPi, but now I'll need to find it. What's the version number of SuperPi that'll work on Win98, even if buggy? Another benchmark I'll have to try for processor analysis is SiSoft Sandra, I have an older version somewhere, I'll have to find it and do more tests.

To be fair the Slot A Athlon and K6-2+ results aren't clock for clock, but they're close, K7-550 vs K6-2+ 500. The other difference is the Video Card is a Voodoo3 2000 vs 3000. Still, they can't be THAT different, and still shows a huge jump just from going from K6 to K7 archetecture.

So I think I'm gonna have to grab my collection of K6 chips and find SiSoft Sandra, and a SuperPi that works on Win98 and try them again. A detailing of Processor specific tests rather than Graphical tests (like QuakeII and 3DMark) would better show what these chips are capable of.

I've also got a Cyrix MII 366 (2.5x100) and a 6x86L-PR200 (if it runs@1.5x100) I could try with the K6's, and a Voodoo3 2000 PCI I can toss in my Socket 370(100FSB) system I can test a Celeron 466, Pentium III 650 and a Celeron 800 to see how they compare. Plus if I feel adventurous enough, I might try throwing some tests at my Socket 754 (Sempron 2600 & 2800, A64-2800 & 3000) and 939 (3500+) boards.

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =

Reply 8 of 67, by maddmaxstar

User metadata
Rank Member
Rank
Member

I found this link: http://files.extremeoverclocking.com/file.php?f=36
SuperPi 1.1e, says it works with Win9x/NT/2k/XP.

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =

Reply 9 of 67, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
maddmaxstar wrote:

I found this link: http://files.extremeoverclocking.com/file.php?f=36
SuperPi 1.1e, says it works with Win9x/NT/2k/XP.

Downloaded. I'm gonna give it a try 😀

Say, how about we create 2 new topics for our benchmark results? One for your version of SuperPi and one for 3dmark2001se? 😉

Edit:This one doesn't have decimals though

Last edited by Tetrium on 2012-01-22, 21:33. Edited 1 time in total.

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 10 of 67, by maddmaxstar

User metadata
Rank Member
Rank
Member

Sounds fun to me, I'll Bite 😀

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =

Reply 12 of 67, by maddmaxstar

User metadata
Rank Member
Rank
Member
Tetrium wrote:

Edit:This one doesn't have decimals though

Whats the version that runs under Win9x and does support decimals? (And where can I get it? 😜)

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =

Reply 13 of 67, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

Just finished first calculation:
1M ran in 3m 9s (=189s) on this rig:
Pentium 3 933Mhz
VC820 motherboard
256MB RDRAM PC-800

Which calculation for SuperPi will we use anyway?
I reckon the 1M will take too short on new rigs and anything over 1M will start to get boring on anything real slow

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 14 of 67, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
maddmaxstar wrote:
Tetrium wrote:

Edit:This one doesn't have decimals though

Whats the version that runs under Win9x and does support decimals? (And where can I get it? 😜)

I'll see if I can upload it...

Edit:Uploaded! Check the download link below 😉

Attachments

  • Filename
    Super Pi Modded.zip
    File size
    82.18 KiB
    Downloads
    210 downloads
    File license
    Fair use/fair dealing exception

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 15 of 67, by maddmaxstar

User metadata
Rank Member
Rank
Member

I'm good with 1M, that timing appears to be appropiate with older PI/II/III class systems

I just ran SuperPi 1.1e on the K6-2+ 500 at 1M, clear time of 6min 26s.

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =

Reply 16 of 67, by Tetrium

User metadata
Rank l33t++
Rank
l33t++

Just remember, this SuperPi always crashes the system after the first run. After a reboot, it still gives you the result and any calculation done after that runs without crashing anymore.
Also this runs on XP as well. I don't mind the crashing all that much though and theres a trick so it doesn't crash anymore.
The trick is to copy the result page along with the exe so when you go to benchmark your next rig, it'll have the result of the other machine already (I use 16k as I ignore that benchmark result anyway)

Edit:I uploaded a fix for the crashing.
I also included a readme this time. All you have to do is copy 1 extra file to the same directory as SuperPi and 2 things will happen if you open SuperPi:
1)It will display a result for 64k (it's the one from my P3 933Mhz)
2)No crashing 😀

Attachments

  • Filename
    Super Pi Modded2.zip
    File size
    82.45 KiB
    Downloads
    237 downloads
    File license
    Fair use/fair dealing exception
Last edited by Tetrium on 2012-01-22, 21:55. Edited 1 time in total.

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 17 of 67, by maddmaxstar

User metadata
Rank Member
Rank
Member

There, SuperPi 1.1Mod @ 1M:
K6-2+ 500MHz = 6m 20.973s

EDIT: Yeah, I'm not overly worried about the crash, as long as it gives results.

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =

Reply 18 of 67, by ProfessorProfessorson

User metadata
Rank Member
Rank
Member
maddmaxstar wrote:
ratfink wrote:

but for what it's worth, I'm impressed with the longetivity of my Radeon HD 4850, I've had it since Summer '08 (just after release) and the thing still runs everything I throw at it, Dirt 3 and Portal 2 run beautifully at max settings.

Yeah, I have to agree with you there 100 percent. My sons system uses a HD 4830, and that thing still runs stuff great for the most part. Even my two 8800 GTS 320 cards still hang tough even today in my SLI rig, granted I dont run anything at a res higher then 1440x900 on them, since that is the monitors max res that the SLI rig is connected to. All in all, the medium to high end DX 10 generation of cards really have held up well.

Reply 19 of 67, by maddmaxstar

User metadata
Rank Member
Rank
Member

Next result.
Mobile Pentium MMX 233MHz (Tillamook on MMC-1) = 12m 24.011s
System: IBM Thinkpad 380XD, 96mb EDO RAM, Win98se

= Phenom II X6 1090T(HD4850) =
= K7-550(V3-3000) =
= K6-2+ 500(V3-2000) =
= Pentium 75 Gold(Voodoo1) =
= Am486DX4-120(3DXpression+) =
= TI486DLC-40(T8900D) =
= i386sx-16+i387(T8900D) =