K6-III+ works good on 133 FSB

Discussion relating to Socket 7 hardware.
User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

I couldn't get Windows to run with the L2 cache on above the 100FSB on this rig. I don't think it would have made much a difference anyway.

I'm not surprised how weak the performance is on these SiS530 boards, they have always fell behind MVP3 and ALI5 boards partially due to their small L2 cache size (normally 512k) which only caches 128MB and they were castrated for greater stability for Compaq and Hp machines.

This SiS530 board can keep up with Via and ALI boards at 133FSB with some tweaks which opens some possibilities for people who own these machines. :)
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

hi k6ers,
greetings from my vacation on rhodos island !

I could not help but stop by and follow your discussion. it seems to me that at the current point the bench numbers uranium gathered are lower when synthetic benches are run (like everest), but are GREAT when a real world bench is run, like superPi or 3dmark2001. I have not heard of any such good number gathered on a p5a or any via/chipset board.

Jim, could you please run the superPi bench (1mb) on your asus but also dfi boards? superpi stresses 2 components> the cpu/core logic and ram-subsystem. of course for better comparability the cpu-clock should be the same (600) and mb/cache set to off (as uranium set his system up) so that we are then finally be in the position to attribute any difference in this numbercrunching bench to the memory performance (100 vs 133 mhz)

also I would like to point out that uranium used lower memory settings (2/3/2) which may eat 10mb/s into the everest mem-read result. according to my investigations memory write scores of 157 is nothing to complain about. real world benches don"t change much whether you get 157 or 177 write score in everest. I wonder if the 2/3/2 setting uranium had to use to get a stable system is because of the ram-stick he uses or whether this is due to the 530 chipset ....
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

Something not clear to me is how much RAM is in Uranium235's machine, and what configuration. It has been my experience that "More is better". It could be that his results are limited by that factor.

Could also be that the advantage of the 5.5 multiplier is : A) Not applicable to the SIS chipset; or B) Over balanced by the larger FSB speed increase entailed by going from 112 up to 133 as opposed to from 100 to 105.

On my machines I get better memory write values @ 5.5 x 100 than I do @ 6 x 105 on an ALi chipset; and better memory write values @ 5.5 x 100 than I do @ 6 x 103 on a VIA chipset.

Also curious, not that I have one of those boards; but are you going to post the PCR File you created?
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

I have the Gigabyte GA-5SMM board and have tried it with hynix pc133 cl2 ram at 100fsb. The best memory settings I can get are the same as Uranium235 (2-3-2-5). The same memory on a via board or an ali board will get (2-2-2-5) without breaking out into a sweat. I think that Uranium's board is ASUS.

At 100fsb on my board running at 5.5X with the memory (3x128mb, 2-3-2-5), cache off, and wpcredit set up as Uranium's sytem I get for Everest

Memory read 237
Memory write 125
Memory latency 267

by the way I get exactly the same results with the cache on as well provided wpcredit is set up correctly.

I get similar results with the Aopen AX59PRO board (Via chipset) straight out of the box with no wpcredit tweaks. When I get some time I'll start to get some o'cked figures. You can set the GA-5SMM board up to get 100,105,112fsb with 3x divider and 124 and 133 with a 4x divider.

Stedman
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

What chipset is in that board Stedman? (Presumably SIS, but want to be sure). Also 768 Meg of RAM should give better numbers.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Jim

Yes the GA-5SMM board has the SIS530/5595 chipset combination. I can max the ram out and see if it gives any improvement.

In reference to your earlier query with regards to the 5.5x multiplier it does have the same effect on the SIS530 chipset as well. I can post some figures for you when I get home tonight and you can see if the percentage gain is equivalent to the via and ali chipsets.

Stedman.
User avatar
Uranium235
Senior K6'er
Posts: 183
Joined: Mon Sep 12, 2005 9:59 pm

Post by Uranium235 »

Those GA-5SMM boards also use the SiS530 chipset....

The config I used is unchanged from the start of this thread:
Asus P5S-VM
K6-3+ 400ATZ @600MHz (4.5x133) 2.3v
512MB(2x256) Crucial PC-133 2-3-2-6
ATI Radeon 9100 64MB PCI
Western Digital 40GB HDD 7200rpm 8MB cache
Windows98SE

I believe the memory timings are the tightest allowable by the SiS530 chipset. This Crucial memory is plenty capable of much faster speeds. I've ran them as high as the 150FSB on a KT133A board.
BTW, I was mistaken in my last post; the L2 cache only caches 64MB, not 128MB, which is why it hardly makes a dent in any benchmarks.
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

@ Stedman thanks; - would appreciate that. @ both of you : For what it is worth, XP will raise your numbers by about 10 points in Memory Read; at least it did on my MVP3 board. The ALi board never got tried w/ XP.
@ Uranium235 : Thanks for running those tests, goes to show that just when I think I've got things figured out, something else crops up. Pity your cache won't work over 100 Mhz. I have found that where RAM is uncached @ 3rd level the Sandra memory test results will rise w/ repeated running of the test because of an increased concentration of cache hits.
Superpuppy 3
K6-3+ 450 ACZ (6x100)
DFI K6BV3+/66 Rev B2 (2 Meg) w/ 2x28mm Chipset Fans
2x256 Meg PC 133 Hynix SDRAM
1x 20G Maxtor (7200)
2x 80G Maxtor (7200) Ducted w/ 2x486 Fans Mount
52/24/52/16 LG CDR/RW/DVD
8/4/3/12/24/16/32 LG Super Multi
ATI 9000 aiw Radeon AGP
SB Audigy 1 MP3 Sound
CMD 649 IDE Controller
NEC USB 2 Card
Post Reply