The King is Dead . . . Long Live the King!

[Hulk]

I'm going to post both scores. They are both valid and the overclocked configuration provides information regarding the benefits of increasing the bus.

I don't think you're card could be accelerating MPEG-2 encoding but one of the beta testers around here would be better qualified to answer that one.

-Mark

[Dr Mordrid]

Originally posted by Hulk
Doc,

Is the new test going to be the same result as the old one, just using features in 7 that may expedite it? Or is the new test going to be completely new, thus making a more up to date and comprehensive test?

I'm thinking it's option number two. I don't think we need any backward compatibility.

-Mark

Same sequence, but edited/saved in MSPro7 just to make sure it's making the most of things.

Glad I wrote all that crap down

Dr. Mordrid

[Hulk]

I contacted Dave E. for some further clarification on his system setup and verification of his scores. He verified his scores and here is is system data:

1) AMD Athlon XP 2000+ CPU
2) 2 x 256MB Sansung DDR 333 PC2700 in banks 0&4 running synchronously
with FSB at 2 x 133 (no difference worth noting when running
asynchronously at 2 x 166) (Using banks 0 & 4 bring into play Dual
channel memory support)
3) BIOS Version 1001G (later BIOS have less aggressive optimal settings
for CPU Memory timings)
4) Windows XP Pro (set for best performance and page file distributed
over C & D drives)
5) Msp 6.52 all patches installed
6) Various S/Ware installed including MS Office XP Pro, Nero, Photo
Impact, DirectX 9, Windows Media 9 Series,
7) Drivers from distribution CD.

[davpen]

Perhaps this is down to faster memory. 2x256Mb gives the benefit of dual channel memory. The other XP 2000+ processors are all using slower memory sub systems. The exception is then the XP 2400+ on an nForce2 motherboard. It would be interesting to know if this system was taking advantage of dual channel DDR or not.

Another interesting thing is that the record holding XP2600+ has 768Mb of RAM. Is this 3x256 or 256+512 and in either case does this allow dual channel performance on an nForce2?


Cheers,

David.

[wkulecz]

Nforce2 can only do dualDDR on two identical modules, If you add a third I've heard claims that dual channel is disabled, and counter claims than you get dualDDR for part of RAM and single channel for the DIMM in the third slot. Not having one, I can't say which is true, or if it depends on the Motherboard/BIOS.


I'd sure like to see the MSP6.52 benchmark run on the same machine using MSP7 to get an idea of how much improved MSP7 really is for rendering.


I wasn't criticizing, I just wanted to point out some things that don't look right and stimulate discussion of what might be going on. I'm happy to conclude that the SiS 735 is a bottleneck for the faster Athelons. This is important as Fry's regurlay runs "specials" on SiS735 based ECS motherboard bundled with XP1800 thru XP220 ranging from $90 to $160. I was looking at these since I need to build another Linux box and was thinking of using my PIII-800 for Linux and buying one of these to get some improvement while waiting for MSP7 and a system specifically for it. Looks like the $90 special would be the most bang/buck for MSP rendering/encoding.


If memory bandwith is really the key then I may wait for the next rev of the P4 and "Granite Bay" chipset for dual channel DDR with 400MHz FSB.

I'm really being tight with spending on computer upgrades right now as I badly want one of the Kodak DCS14n 13.8 Mpixel full-frame 35mm DSLR still cameras and a Sony VX2000 or whatever its sucessor is. (I'm waiting to see if it offers improved low light, and if not I'll try and jump on a "closeout" VX2000).

--wally.