Page 2 of 2

Posted: Sun Oct 29, 2006 10:17 am
by Stedman5040
Here are some more benchmarks using Hotcpu tester pro 4 lite (4.3)

Celeron 800 @ 8x100fsb
No tweaks

Total score 1597

With tweaks as prevoius post @ 800

Total score 1863

A near 17% boost in performance with this test.

Celeron 800 @ 1066 (8x133fsb)
No tweaks

Total score 2148

With the same tweaks as previous post

Total score 2467

a near 15% boost this time

Overall boost on this test from 800 no tweaks to 1066 with tweaks is 54%.

Not bad :)

I look forward to your results.

Stedman

Posted: Sun Oct 29, 2006 8:05 pm
by Jim
Not bad at all!! Have you checked yet to see if there are any other tweaks that may raise it further still? Seems to some extent that the reputation for lousy memory performance associated with VIA chipsets is really something brought on by the default settings chosen by the mobo manufacturers.

I am still working on SP3; but have got a new mobo on the way for SP2; and also a "Slot T" adapter for my FIC KA-6100 and a 1.4Ghz Tualatin P3 S for the FIC. Not sure yet which case I'll build it in,; but giving it the "Superpuppy" treatmeant is going to be a lot of work.

Posted: Wed Nov 22, 2006 12:05 am
by Uranium235
Just got the Epox EP-3VWB+ together to run some benchmarks. Here's the setup:
Epox EP-3VWB+, 9/21/2001 BIOS
Intel PIII Tualatin, 1400Mhz
512MB (2x256) Hynix PC133 CL2
ATI Radeon 9100 64MB PCI
Western Digital 80GB HDD, 7200rpm, 8MB cache

I did some minor BIOS mods using Modbin. First, the RAM could only be set to run at 2-3-3-6 so I unlocked the hidden "Turbo Mode" in order to set the RAM to 2-2-2-6. Second, the shared memory could not be disabled for the integrated video, so I unlocked the "N/A" setting for the Frame Buffer to kill the shared memory.

These are the results from Everest v1.51.195

@1400Mhz (10.5x133) no tweaks
Memory Read- 798 MB/s
Memory Write- 216 MB/s
Memory Latency- 128.2 ns

These are the results using Stedman's Wpcredit tweaks:

@1400Mhz tweaked
Memory Read- 1039 MB/s
Memory Write- 217 MB/s
Memory Latency- 98.3 ns

Here's some results overclocked:

@1470Mhz (10.5x140) no tweaks
Memory Read- 829 MB/s
Memory Write- 226 MB/s
Memory Latency- 122 ns

@1470Mhz tweaked
Memory Read- 1089 MB/s
Memory Write- 228 MB/s
Memory Latency- 93.9 ns

I'll run some different bench's later when I get more time......

Posted: Wed Nov 22, 2006 12:18 am
by Jim
I can see I am going to be bugging you for help once I get onto the SPZ project. Is that processor a P3S?

Posted: Wed Nov 22, 2006 12:42 am
by Uranium235
Jim, here's the markings on the CPU:
1.4 GHZ 512KB/133 1.45V SL5XL
I'm not sure exactly what a P3S is.

I'd be glad to help with SPZ if needed. :)

Posted: Wed Nov 22, 2006 2:40 am
by Stedman5040
That is a nice P III-s cpu you have got there. It is the 512k cache that makes up the difference. I believe that they were even able to outperform similar speed P-IV's and XP Athlons of that era.

Glad to see that the tweaks also worked for you.

I must see if I can get hold of a Slot-T adapter as I know that the VA6 boards can run Tualatin cpuz with one. All of the adapters on ebay in the UK seem to be onthe expensive side.

Has anybody tried piggybacking a socket 370 tualatin adapter in a slot 1/socket 370 adapter ?

Stedman

Posted: Wed Nov 22, 2006 9:19 pm
by Jim
@ U235 : Stedman is right. The "S" stands for "Server" and yes they have twice the on chip cache of other P III chips. Found out from TA152H. Needed to know cause I was E-Baying for the processor for SPZ. Your SL5XL is the earlier of the two types they made. Mine is the SL6BY, which oddly enough the same seller was selling for less than the SL5XLs. They have since revised their prices, reducing the price of the SL5XL to be the same as that of the SL6BY.

@ Stedman : Don't think I would try that unless you have a special adapter to mate the chip to the slocket. I don't think a regular slocket will give you the required 1.45v, though I could be wrong. I think the makers of the "Slot T" did make such adapters too. At least when I was surfing their site I saw something that apparently fits between the processor and the slocket to adapt a 1.4 Gig Tualatin to a slocket.

Posted: Thu Nov 23, 2006 3:45 am
by Stedman5040
Jim,

I was in fact talking about using a FCPGA to FCPGA2 socket370 adapter piggybacked or plugged into a conventional 370 to slot1 adapter. So you would have in effect an adapter plugged into another adapter.

I would think it should work but has anybody tried it in practice.

Regards,

Stedman.

Posted: Thu Nov 23, 2006 6:29 am
by KachiWachi
Stacking adapters is not a good idea in practice...even thought it *might* work just fine.

Posted: Fri Nov 24, 2006 12:02 am
by Uranium235
One thing that I find interesting is that although the P3S may be a very desirable CPU, the K6-III+ 550 CPU's are fetching almost double the money on Ebay.

Posted: Thu Dec 21, 2006 9:29 am
by Stedman5040
Update on the VA6. Currently running a Celeron 850 cD0 stepping at 1133MHz at 1.75V. I think that this is as far as it goes with the FCPGA cpu's. The set up runs smoothly on a day to day basis.

Have managed to get hold of a Tualatin Celeron 1000A (£3.40 off e-bay including postage. This cpu also came with a P III 700 and a celly 600. 3 cpu's for £3.40 all working) that I got running briefly on a PCChips SIS630 board at 1333MHz. This cpu is hopefully going to work on the VA6 as soon as the Tualatin adapter arrives. The performance of the SIS630 chipset is just as woeful as the SIS530 chipset. No AGP slot on the board and with onboard AGP? you just do not want to know. Do you?

Stedman

Posted: Thu Dec 21, 2006 9:15 pm
by Jim
"SIS" --- Doesn't that stand for "Sell It Stupid"? :P

Posted: Sat Dec 30, 2006 11:51 am
by Stedman5040
Have now installed the Tualatin Celeron 1000A into the Abit VA6 board and am running at 1333 MHz at 1.5v. The Tualatin adapter seems to work fine in the slot1/370 adapter.

Got a few bench results at 1333 as follows

Everest MR/MW/ML (1020/201/108) with same tweaks as before
Superpi 256k 27secs 1M 144secs
Sandra 2004 Memory bandwidth 962/953

Hotcputester

Total score 3226
Metamark to MHz ratio 2.41

The Celeron shows up in the bios as (222x6) 1332MHz but runs fine and shows up in CPUz as 133x10 Tualatin Celeron step tA1.

Stedman.

Posted: Sun Jan 14, 2007 3:10 am
by Stedman5040
Got hold of a P-III-s 1266MHz cpu and plugged it into the board. Tried some benches and found as follows

Using cpu with same wpcredit tweaks as stated earlier in this thread we have

Superpi 256k 22secs

This is 5 seconds faster than the Celeron 1000 @ 1333 (133x10)

Superpi 1M 130secs

this is 15 seconds faster than the Celeron 1000 @ 1333 (133x10)

Everest

MR 1015 (Celeron was at 1018)
MW 203 (Celeron was at 201)
ML 108 (same result as celeron)

Sandra 2004 memory bench 962/961

Here is the interesting part and this relates somewhat to Jim's experiences with his memory testing of the Superpuppies. Anyway I ran the memory and cache bench and then reran Everest and got the following

MR 1003
MW 241 (20% gain !!!)
ML 108

I ran Hotcputester as well and also found a boost in performance over then celeron which I will report on later.

The P-III-s really stomps on the Celeron in these benches and the only difference is an extra 256k cache which is working at a latency of 0 rather than at a latency of 1.

Stedman

Posted: Sun Jan 14, 2007 3:11 am
by Stedman5040
Got hold of a P-III-s 1266MHz cpu and plugged it into the board. Tried some benches and found as follows

Using cpu with same wpcredit tweaks as stated earlier in this thread we have

Superpi 256k 22secs

This is 5 seconds faster than the Celeron 1000 @ 1333 (133x10)

Superpi 1M 130secs

this is 15 seconds faster than the Celeron 1000 @ 1333 (133x10)

Everest

MR 1015 (Celeron was at 1018)
MW 203 (Celeron was at 201)
ML 108 (same result as celeron)

Sandra 2004 memory bench 962/961

Here is the interesting part and this relates somewhat to Jim's experiences with his memory testing of the Superpuppies. Anyway I ran the memory and cache bench and then reran Everest and got the following

MR 1003
MW 241 (20% gain !!!)
ML 108

I ran Hotcputester as well and also found a boost in performance over then celeron which I will report on later.

The P-III-s really stomps on the Celeron in these benches and the only difference is an extra 256k cache which is working at a latency of 0 rather than at a latency of 1.

Stedman