VOGONS


First post, by Warlord

User metadata
Rank l33t
Rank
l33t

Legit question, is it faster to disable the L3 Cache on a K6 III +?

Reply 1 of 6, by Lo-ResBros

User metadata
Rank Newbie
Rank
Newbie

I remember watching something about that on a phil’s computer lab video. He tested it but I think it was a myth, or maybe there was some benefit to it under windows I’m not sure. check out his K-2+ or K-III+ videos. I’ll try to find it for ya

*edit* here it is:
https://youtu.be/A2Oymnq5DEQ

AMD K6-2+/550MHz
Soltek SL-54U5 Super Socket 7
64MB SDRAM
3DFX Voodoo 3 3000 16MB
Sound Blaster AWE64
Roland SC-55mkII

Pentium MMX 233MHz
Tyan TitanVX S1470 Socket 7
128MB SDRAM
ATi Rage 128 pro 32MB
Audio Plus 320 v.1
Roland SC-88

Reply 2 of 6, by cyclone3d

User metadata
Rank l33t++
Rank
l33t++

I think it depends on the motherboard. If you have more RAM than the motherboard cache can cache, it might be faster to disable it.

It might also depend on the motherboard you are using. The later versions of the ASUS P5A do not like the III , II+, or III+ CPUs because of a hardware bug in the updated chipset. There is a hardware mod to get around that though.

Yamaha modified setupds and drivers
Yamaha XG repository
YMF7x4 Guide
Aopen AW744L II SB-LINK

Reply 3 of 6, by MrSmiley381

User metadata
Rank Member
Rank
Member

My Tyan S1590S with a modified BIOS refuses to load DOS if I have the L3 cache enable and at the very least my Gravis Ultrasound MAX installed. Not sure if it's a hardware conflict or not, but when I performed benchmarks with the conflicting hardware removed the machine was only ~1% faster with the L3 cache enabled. In this case it's technically faster but negligibly so. RAM count was 256 MB for both benchmarks, which I recall being more than the motherboard could normally cache, but well under the limit for the K6-III+. I'm thinking it partly depends on the chipset implementation. I see in Phil's video that his Turok benchmark gained a frame from disabling the cache, so it could also be how the software is written as well.

I spend my days fighting with clunky software so I can afford to spend my evenings fighting with clunky hardware.

Reply 4 of 6, by swaaye

User metadata
Rank l33t++
Rank
l33t++

One reason to disable the motherboard cache is it tends to limit the max FSB clock. A faster FSB means faster memory performance overall. That is quite beneficial for a K6+ that already has excellent L2 caching.

Memory latency is probably slightly higher with the motherboard cache enabled too. Cache levels add some latency.

Here's an old article with some good examination of the K6-III and L3 cache. It should typically be faster but yeah I imagine it varies by board and by cache characteristics.
https://web.archive.org/web/20040228094951/ht … /read.jsp?id=34

Reply 5 of 6, by Warlord

User metadata
Rank l33t
Rank
l33t
swaaye wrote:

One reason to disable the motherboard cache is it tends to limit the max FSB clock. A faster FSB means faster memory performance overall. That is quite beneficial for a K6+ that already has excellent L2 caching.

Memory latency is probably slightly higher with the motherboard cache enabled too. Cache levels add some latency.

I sort of question phils video bcuz how do I know that some of those performance hits had to do with high latency memory timings in his bios when he adds sticks in and out of that test rig.