K6-3+ 450ACZ under / over clocking results.

Discussion relating to Socket 7 hardware.
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Hi Jim,

I basically ran both the cpu's at 450 as a head to head. I'll take a look through WPCREDIT settings to see if I can see any differences there that maybe can account for the difference in performance and I'll double check the Bios settings too and get back to you. At a pinch I can get 500 out the K6-III so I could also recheck the benchmark at 500 on both cpu's.

Stedman :)
User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

I would think there would be virtually no difference in a K6-III vs. a K6-III+. The only "real" difference is the 0.25-micron vs. the 0.18-micron process technology...which leads to lower power consumption...and "perhaps" a tiny bit of performance.

I did find it interesting the the machine "slowed down" writing when you added the motherboard (L3) cache. I had noted this on my DFI as well, but have yet to understand why.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

I don't think that there is a difference between the the two cpu's, but there seems to be a difference in the peformance between the two chipset revisions of the AladdinV. I think the 1.06 board has a later revision of the chipset and maybe this is what is giving the boost in performance. If this is the case then it is a real pity that the K6+ cpu's do not work on these boards. The only real thing I could pick up on WPCREDIT between the two boards was that on the 1.06 revision the Tag ram is set on for internal and external cache. I presume that this is the benefit gained from the updating of the chipset.

Has anyone any thoughts as to why the 1.06 board seems to give better Everest benchmarks compared to the 1.04 board given that they are both running with the same speed cpu (K6-III/450) and same bios settings.

I could try disabling the tag ram and running the bench again.
Jim
K6'er Elite
Posts: 1745
Joined: Wed Jan 21, 2004 7:10 pm
Location: Toronto

Post by Jim »

Never seen a 106 board, so not sure what it has. The 104 had defective internal tag ram in the chipset, so they set it disabled and put additional external tag ram on the mobo. Does the 106 have both internal and external tag ram? That would be a bonus if the internal is working; in that it would allow the cache to cache more ram. Though it is hard to see what benefit that would be w/ the cache disabled For what it is worth later versions of the chipset had larger (more bits) tag ram too. That is why ALi chipsets past a certain rev allow 512k cache to cache 512Meg of Ram.

One thing I can guarantee you. You will get your best Everest Memory Write performance numbers by using the 5.5 multiplier on the mobo.
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
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

well, I am not surprised with results going up after having the bios setup accordingly.

I am also not surprised to see the 1.06 board outperform the 1.04, no matter what combination is used: cache=on/enh.pg.count=disabled versus cache=off/enh.pg.count=16. (important note: always use one of these combinations! otherwise you loose performance. tested!)

in other words: with no further tweaking (WPCREDIT and alike) the 1.06 board performs very very well. no other board gets 340/140 (read/write) everest-results from the start! I can confirm these same results (tested with a k6-3-400).

what I am wondering about is that you did not run into troubles when using a k6-3-450. from that speed on and up I got very low mem-write results (65-70). @stedman: could you please o'c your k6-3-450 to 500 and see what happens? if you still get the same good results then please raise the fsb to 105 so that the cpu runs @ 525 (proper cooling adviced) and bench again. please post results, thank you!!!

@jim: yes, I also thought that the reason for good memory performance of the 1.06 board is the finally activated (because technically "working") tag-ram. but my understanding is that once the 3rd level cache is deactivated, the tag-ram is out of order, nothing to do. but we still get these quite impressive memory performance "out of the box" .....
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Jim,

looking at the host bridge registers in WPCREDIT for the 1.04 board register 4000 is set at a value of 13. Bit 6 of this register is the toggle for internal tag ram and in this instance it is off. The same goes for register 4001 (value 04)where bit 0 is the toggle for external tag ram. both of these bits are set to tag ram off (ie 0).

Going to the 1.06 board the same bits in the registers are set to 1, which enables both internal and external tag ram. So register 4000 has a value of 53 and register 4001 has a value 05.

Also in the 1.04 board bit 7 of register 4003 is set to enabled whereas in the 1.06 board it is set to off.
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

@ internal / external tag-ram.

it is my understanding, that the tag-ram is not built to administrate the cpu's on die 1st and 2nd level cache. it is the cpu that cares about proper on die caching.

it is also my understanding that internal / external tag-ram refers to whether the tag-ram unit is built into the chipset itself or is an external, added, extra chip on the mainboard. I know of times when the asus t2p4 mainboard offered an empty place on the mainboard for a possible add-on tag-ram chip.

but maybe someone here has a deeper technical insight. please post!
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Don Pedro,

First results of o/c the K6-III on the 1.06 board to 500MHz. Managed to clock at 2.2V which seems pretty good for this chip. results as follows

Memory read 338MB/s
Memory write 124MB/s
Memory latency 199.2ns

With ex/cache off and pg count 16

stedman
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

interesting, indeed.
the performance starts to drop.

I am looking forward to the numbers @ 105 fsb.

btw: what type is that cpu (ahx, acz, afq)? it is printed on the top of the cpu ....
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Tried posting @ 525 @ 2.4V but failed to get into windows so tried 105fsb at 4.5X (472MHz) and got the following results. the cpu is an AFX 2.2V.
Runs at 450 at 2.1V and 500 at 2.2V. I'll give 525 another shot at 2.5V

K6-III/450 @ 472.5 (105fsb)

Memory read 374MB/s
Memory write 136MB/s
Memory latency 182.7ns
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

thanks!

for comparison here are my results on a

P5A 1.06 board,
1011-05 bios,
cache=off, enh.p.c.=16,
K6-3-450 AFX, 2.2v,
fsb 105 -> 472,5mhz,
Win2k Svp4:

everest 201:

read 375
write 139
delay 184

so we are quite close (just as it should be) and that is a good sign.
Last edited by DonPedro on Sun Nov 20, 2005 12:28 pm, edited 1 time in total.
User avatar
KachiWachi
K6'er Elite
Posts: 507
Joined: Wed Sep 21, 2005 10:53 am
Location: Pennsylvania, USA

Post by KachiWachi »

I would expect a small performance increase with an internal TAG, since the distance between it and "the chipset" vs. an external TAG and "the chipset" means less time waiting to see if the needed information is in the motherboard cache or not.
Moderator - Wim's BIOS

PC #1 - DFI 586IPVG, K6-2/+ 450 (Cyrix MII 433), 128 MB EDO. BIOS patched by Jan Steunebrink.
PC #2 - Amptron PM-7900 (M520), i200 non-MMX, 128 MB EDO
PC #3 - HP8766C, PIII-667, 768 MB SDRAM
PC #4 - ASUS P3V4X, PIII-733, 256 MB SDRAM
PC #5 - Gateway 700X, P4-2.0 GHz, 768 MB PC800 RDRAM
PC #6 - COMPAQ Evo N1020v laptop, P4-2.4 GHz, 1 GB PC2700 DDR
PC #7 - Dell Dimension 4600i, P4-2.8 GHz, 512 MB PC2700 DDR
PC #8 - Acer EeePC netbook, Atom N270 @ 1.60 GHz, 1 GB RAM
PC #9 - ??? ;)
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Jim,

Tried out the 550MHz setting on the 1.04 board with the following results

External cache off / Pg count 16T

Memory read 314
Memory write 155
Memory latency 210

Extenal cache on / Pg count disabled

Memory read 308
Memory write 140
Memory latency 207

It certainly seems to give much better results for the memory write. how does that work

Stedman
DonPedro
K6'er Elite
Posts: 578
Joined: Wed Jul 27, 2005 2:11 pm

Post by DonPedro »

was it the +-version of the cpu you used?

I find it odd that your memory read values are down 10-12%. that is really strange ....
User avatar
Stedman5040
Veteran K6'er
Posts: 271
Joined: Mon Sep 05, 2005 4:22 pm

Post by Stedman5040 »

Don Pedro,

The board is the 1.04 and not the 1.06. The cpu is a K6-III+450 booted at 550.

The 1.04 board does give the poorer memory read when compared with the 1.06

with a couple of wpcredit tweaks the 550 gives

Memory read 333
Memory write 166
Memory latency 202

Anyway at 105fsb and multiplier at 5.5X I get the following results from the 1.04 board. Cache on and pg count disabled

Memory read 327
Memory write 154
memory latency 194.6
Post Reply