AMD K6/II -v- PII

[philtech]

We have AMD K6/2 400's running alongside Intel PII/333's on our NT Network. The Intel PII's seem to outperform the AMD systems in every respect - is this normal, or should we be looking at the set up of the AMD Motherboards?

They all have 128 Mbyte RAM, Win98se. They have varing HDD's, but this seems to make little, if any difference. The AMD's have onboard Video (SIS530), whereas the PII's have PCI Cards (4 Mb Sparkle). All have indentical PCI 10Mbps Combo Network Cards connected to identical hubs.

There is a slight chance that the video drivers on the AMD systems, although certainly SIS 530, have been loaded from incorrect CD, so could this have a bearing? Loaded drivers seem to works fine.

The AMD systems are on average twice as slow or worse.

[muno]

The PII seems to be running at 95mhz fsb, are you certain that the amd system is running at 100mhz fsb rather than 66mhz?
Twice as slow seems little odd...
Hmm.. Smaller virtual memory setting?
-M

[MaxVal]

The 333MHz PII's run on a 66Mhz clock, The AMD's could be running at 66 or 100MHz. Due to the performance difference you note I would suspect 66MHz.
The PII has a couple of advantages in that it has a large 512K L2 cache that runs at 1/2 core speed, also the floating point unit is more efficient than that of the K6-2. The K6's L2 cache is provided by the MB, and can be anywhere from 256kb to 2048kb (512kb most likely). The K6-2 cache runs at the clock frequency, much slower than the PII's.
Still, other variables such as chipset can have a big impact on performance. The AMD's may be using a TX chipset that won't cache more than the first 64Mb of memory. Let’s not forget the video either. SIS video solutions leave a lot to be desired!
MAX

[philtech]

Thanks for replies.

Chipset is SiS 530/5595 two chip AGPset with integrated graphics controller. We are pretty certain AMD's are set at 100 x 4. Is it likely that overclocking will help performance, if so what do we have to change. ie. is it just like changing jumper setting to 100 x 5 or are there other changes to make.

Are there any BIOS settings which affect performance of AMD systems?

Many thanks in anticipation.

Ian.

[LostBok]

Just to be clear - the K6-2's L2 cache is on the mobo and runnin ga the MOBO clock speed (66 or 100Mhz)...

The main difference by a LONG way will be the FPU unit on the P2's was so much more advanced than the K6-2's that in ANYTHING invloving maths computations (most benchmarks, nearly all games, etc) it will thrash the K6-2.

IF you REALLY want to see how much quicker the Deschutes (P2) core is... get hold of a Celeron 400A and compare it to a K6-2 400Mhz... keeping in mind that the Celeron has no motherboard cache to help it out and tha thte Celeron is clockable to 600Mhz if you've go the right RAM and motherboard...

Come to think of it, I think my friend worked out that when he clocked his K6-2 down to 200Mhz and his K5-166Mhz up to 200Mhz, the K5 was actually substantially faster at number crunching!!


[edit] TYPO's [/edit]

[MaxVal]

I suspect the SIS chipset may be impeding the K6-2's. The integrated graphics are great for reducing initial costs, but they are a poor solution for performance. Also, some of the K6's system memory is being utilized for graphics memory.
You would need to check the multiplier jumper on the K6-2's to determine the clock speed. It could be 4x (tooo bad!), or it might be 2x. The 2x translates to 6x on the K6-2. If it was 2x, you could try 4x multiplier (if it exists), with a change from a 66 to 100MHz clock. An upgrade to PC-100 RAM may be needed.
You could also look through the chipset setup options at memory timings. Try lowering latency (lower numbers). Be advised that this can cause instability.
As for the hard drives, they can vary greatly in their contribution as to a systems performance!

MAX

[BFlurie]

Also, if the mobos do not support write (back)-allocation for the L2 cache, this would be a huge performance hit.

The program ctcm17a.zip will determine what is supported. There are programs that can enable memory write-allocation & write-combining for AMD processors (I use them), but the old site is defunct. They are the programs mxk6opt.exe & setewb.exe.

Also, I agree, the on-board video is slowing things down. At least try keeping vid & color resolution down to what is minimally acceptable.

[DanU]

What benchmarks are you using to determine that the K6-2 400 is half as fast as the P2-333? The K6-2 400 should edge out the P2-333 in office apps. In gaming... it's a toss up but the P2-333 might have a slight advantage, especially in non-3dNow optimized games. Games with good 3dNow support should show the K6-2 400 besting the P2-333.

My guess is that the integrated video is slowing down everything. I've used an SIS5598 mobo with a K6-2+ 400. Using the integrated video, it was dog-slow! The same processor on an older 430VX board with a voodoo banshee PCI was much, MUCH faster.

Try putting one of the PCI video cards into the K6-2 system. It will free up your main memory from having to deal with the video. You should see everything speed up.

[MegalosSkylos]

I think an important point has come --perhaps unrealized--that nothing in a computer exists in a vacuum.

One CPU may have a chipset (for example) whose virtues are better bought out by another CPU with its chipsrt.
So its actually a chipset+ CPU vs. chipset + CPU.
Actually , there are other variables.

Thats why I take with a grain of salt the "roundup" comparisons you often see. In the real world "all other things being equal" is almost impossible to achieve.

Always interpret tests in context.

[cadetstimpy]

If the onboard video is sharing with the sysytem RAM that could cause the difference.