512kb => 1mb L2?

[CTho9305]

What sorts of applications benefit / don't benefit from doubling the L2 size on Athlon 64 CPUs?

 

[cmdrdredd]

Games don't benefit much, while video editing and photoshop do. That's the short version.

 

[cmrmrc]

Originally posted by: cmdrdredd
Games don't benefit much, while video editing and photoshop do. That's the short version.

actually its more the opposite, games benefit from the double L2 and editing don't. If you look at some reviews, you'll see the 4800+ ahead of the 5000+

 

[cmdrdredd]

Originally posted by: cmrmrc

Originally posted by: cmdrdredd
Games don't benefit much, while video editing and photoshop do. That's the short version.

actually its more the opposite, games benefit from the double L2 and editing don't. If you look at some reviews, you'll see the 4800+ ahead of the 5000+

personal experience > reviews that don't have real world benefits

Using an E6600 at 3.2Ghz is faster for transcoding DVDs and encoding divx than my E6400 at the same Mhz. Same memory speed and timings.

games are always reliant on the GPU so it's a poor CPU comparison.

I'm posting what I've found for my C2D.

BTW; the receint article shows the 5000+ behind the 4800+ because of the die shrink and increased latency, nothing to do with the cache. That's probably what you saw, unless you show a link of 90nm CPUs and a direct comparison I'm guessing that's what you were looking at.

 

[cmrmrc]

true, i did not say that cache didnt benefit while editing or encoding, i just said that in some games, the extra cache is much more valuable than on encoding.

 

[cmdrdredd]

Originally posted by: cmrmrc
true, i did not say that cache didnt benefit while editing or encoding, i just said that in some games, the extra cache is much more valuable than on encoding.

I haven't noticed any game benefit from any increase in cache. CPU speed does help, but the GPU is always the determining factor in game performance. Unless you're looking at very old games. I dunno what you're looking at so...I'm guessing

 

[cmrmrc]

true again, at high resolution there are no noticeable increase, but at med res or in really cpu dependant games like NWN2, i believe that the extra cache is worth it.

 

[cmdrdredd]

Originally posted by: cmrmrc
true again, at high resolution there are no noticeable increase, but at med res or in really cpu dependant games like NWN2, i believe that the extra cache is worth it.

I haven't played NwN2 but I ment to buy it, just got side tracked with my Xbox360 and Zelda for Wii. I'm suprised it's so CPU limited.

 

[myocardia]

Originally posted by: cmdrdredd

Originally posted by: cmrmrc
true, i did not say that cache didnt benefit while editing or encoding, i just said that in some games, the extra cache is much more valuable than on encoding.

I haven't noticed any game benefit from any increase in cache. CPU speed does help, but the GPU is always the determining factor in game performance. Unless you're looking at very old games. I dunno what you're looking at so...I'm guessing

Actually, you're both correct. Those are the three things that more cache helps. Cmdrdredd, the reason you don't/can't see any difference is because of your video card. Also, let me add that encoding and editing both benefit considerably more from the extra cache than does gaming, from the articles I've seen on it.

 

[applexxe]

seems you are from kahrizak !

 

[A554SS1N]

Originally posted by: cmdrdredd
BTW; the receint article shows the 5000+ behind the 4800+ because of the die shrink and increased latency, nothing to do with the cache. That's probably what you saw, unless you show a link of 90nm CPUs and a direct comparison I'm guessing that's what you were looking at.

The 5000+ is slower than or about equal to a 4800+ when on the same 90nm process if you look at the older reviews - so cache does matter.

 

[CTho9305]

Originally posted by: A554SS1N

Originally posted by: cmdrdredd
BTW; the receint article shows the 5000+ behind the 4800+ because of the die shrink and increased latency, nothing to do with the cache. That's probably what you saw, unless you show a link of 90nm CPUs and a direct comparison I'm guessing that's what you were looking at.

The 5000+ is slower than or about equal to a 4800+ when on the same 90nm process if you look at the older reviews - so cache does matter.

That only demonstrates that latency matters, not that size matters. If the working set of an app is larger than the L1, but fits in both L2 caches, then of course the faster cache will win. If the working set is larger than one of the CPU's L2, then the bigger cache will most likely win, even if the latency is a little higher.