S939 3800+ at most 5% faster than S754 3400+ ?!?

Toggle sidebar Toggle sidebar
N

Nickel020

Senior member
Jun 26, 2002
753
0
0
A German site, Hardtecs4u has done a S939 mobo roundup and a S754 mobo roundup in the same review and they used a 3800+ CPU for the S939 and a 3400+ CPU for the S754. What's interesting though is that the fastest 3800+ CPU/mobo combo is at most 5% faster than the fastest S754 CPU/mobo combo in any given benchmark. The only thing that differed between the two test setups is the CPU and the mobos.

Test setup

UT2004/ Max Payne 2 Benchmarks

They only used a Fx 5900 video card, but even so, the 200MHz clock difference and dual channel RAM should make more differnce than that.

Use google or altavista's Babelfish to translate it.
 
M

Matthias99

Diamond Member
Oct 7, 2003
8,808
0
0
Originally posted by: Nickel020
A German site, Hardtecs4u has done a S939 mobo roundup and a S754 mobo roundup in the same review and they used a 3800+ CPU for the S939 and a 3400+ CPU for the S754. What's interesting though is that the fastest 3800+ CPU/mobo combo is at most 5% faster than the fastest S754 CPU/mobo combo in any given benchmark. The only thing that differed between the two test setups is the CPU and the mobos.

Test setup

UT2004/ Max Payne 2 Benchmarks

They only used a Fx 5900 video card, but even so, the 200MHz clock difference and dual channel RAM should make more differnce than that.

Use google or altavista's Babelfish to translate it.
Click to expand...

Even at perfect scaling, there should only be about an 11% (3800/3400) performance difference between a 3800 and a 3400 (assuming -- and this is a big assumption -- that their PR system is accurate). That said, the fastest 3800+ system is 7% faster in UT2K4 and 6% faster in MP2 at the 640x480 settings, where the video card is less of a factor. That's not bad scaling for a game engine, and the differences are larger in some of the more memory- and CPU-bound tests (like the encoding benchmarks they ran). Of course, it would help if I spoke German. :p

This does seem to show that dual-channel RAM is not really very useful on the A64 (at least not at these clockspeeds), however. Certainly the S754 CPUs are not being held back by memory bandwidth, at least not in games.

Edit: Their 3DMark01 scores are also interesting. Usually you'd expect a rather large performance jump when you up the memory bandwidth and clockspeed like this, but they're showing a pretty small increase. They might be topping out the video card on that bench, though.
 
N

Nickel020

Senior member
Jun 26, 2002
753
0
0
Just noticed as well that 5% isn't the highest difference in benchmarks, that's just what it says in the "Final Words". But still the difference in lots of benchmarks is very small. In Quake 3 the 3400+ are actually all faster in 640x480 and that is not GPU bound at all, in 1024x768 the fastest 3800+ is less than 2% faster. The games they run aren't really GPU limited but still the biggest performacen differnce is 6% in Max Payne 2, which makes you wonder what you paid those extra $350 for. I would like to see a 3400+ against 3500+ comparison on current mobos (not reference boards).
 
flexy

flexy

Diamond Member
Sep 28, 2001
8,464
155
106
>>> They only used a Fx 5900 video card <<<

doesnt make much sense to me :)
 
I

iwantanewcomputer

Diamond Member
Apr 4, 2004
5,045
0
0
it would be nice to see a comprehensive review of socket 939 90 and 130 nm and socket 754 performance sometime

i'm wondering if dual channel ram will help more as soon as the clock speeds are ramped up to the 3 ghz range. ie. will the performance difference between dual channel and single channel be greater if the processor is 2GHz or 3GHz?

on another note i bet dual channel is even less significant when you are overclocking because the hiigher fsb already provides more bandwidth, so the extra dual channel bandwidth is being used even less
 
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom