Architecture question.

Hello, why would increasing the cache size decrease power consumption? I know Intel did this in their core 2 series to decrease power consumption compared to the Pentium D yet if I look at so called "efficient" processors today they have gone back to 512kb L1. What gives?

increasing L2 size does not decrease power consumption. it makes the cpu more efficent per clock with certian tasks, since it is able to hold more information. it doesnt have to keep going back to ram to get information. the core 2 arch is different then the Pentuim D you talk about. as it is a short pipe line that data going through vs the P-D which is a really long pipe. to put it this way P-D by the time it ended was at 38 stages if i recall right. the core 2 line started at 20 or 22 stages that is a rather big decrease, when you also consider they had the same clock speed or near it. now im not sure if anyone knows how many stages core 2 has when the series ended.

the new i series is just a modified core 2 arch, but that is prolly the more simple term. the reason the newer cpus have 512k per core is that they have a large L3 to store the data in. that way it can move large amounts between the cores, while core 2 did have the large(ish) L2 it was shared between cores. it was however not shared when they released quad cores, so to get data from one set of cores to the other. meant taking longer by having to send it back through the northbridge then back up. also 512k is a good inbetween for L2, that is if you look at gaming. more specificlly low res gaming where the fps depends on how much cpu power you have. as with core 2 every doubling of L2 increased performance in gaming by 6.7%, i thinks that the right % when i ran the numbers back then.

leaving from work, be back later to add more...

ok, so for reducing power, there are a number of other things they can do. like switching the material they use for the gates, going to a smaller manufacturing process. there is always something they can do but increasing L2 does not lower power consumption.

back to the L2 size and performance in gaming bit. with newer GPU's and larger monitors/higher resolutions, when you add AA/AF to mix with high resolutions. the load is soly on the GPU but you need a certian CPU speed to max sure your getting the most from the GPU. it has been some time since i have looked at this but i suspect it may still be on target. for every generation you need 200mhz more cpu speed to get the most from it. how ever i think this has changed a bit when you consider that every died shrink on a cpu increases efficiency per clock. while it isnt that big but it is always around the same percentage of a increase. i will still stick with 200mhz though since there are so many variables that need to be looked as a whole. i simply isolated one gen to the next and cpu speed for the same CPU(ie intel core 2, model number eludes me since its been a few years).

if your one to play at high resolutions then cpu speed and L2 size wont matter. since if the CPU is fast enough or more then fast enough to the gpu in that situation. the cpu will be waiting for the gpu to get done before it can send more data to be processed.

also one last thing, L2 runs at the same speed of the cpu. Having larger L2 sizes means that there is less a chance that it will be able to run at full speed. what intel would do is test the die to see what works and what speed it could/would work at. with just 512kb they can pretty much have a guaranteed full speed L2. the L3 runs at a difference speed, i have yet to see accurate numbers. however when intel last had a cpu released with L3 it was a P4(netburst based) extreme cpu. the L3 ran at half the cpu speed, wiether that is the case or not with LGA1366/LGA1156/1155 is something i cant find.

i think i covered every thing but sometimes you never know.

Wow! Thanks so much Evilsizer, That is everything I wanted to know plus wayyy more. Where did you learn all of this? Is there a book or a forum? Maybe you should consider writing a book haha.