32 nm vs 45 nm

[autoct]

:confused: :confused:

How big is the difference between a 32 nm and a 45 nm running at the same clock speed ?

If my program run mostly in only 1 core, could a 6 cores guives worst result than a 2 or quad core ?

[GamerBR]

There is no difference between 32nm and 45nm, what makes 32nm better is that you can build better stuff than with 45nm because of the smaller size, it they have the same clock speed and identical architecture there will be no difference besides power consumption.
And if the program is single threaded and you talking about identical CPUs that all the difference is the the number of cores, the perfomance will be the same, but generally as CPUs with less cores runs with faster clocks, it's likely that it will run better on them.

[Linuxfan]

No speed difference between for example a 920 45nm at 4ghz and a 980x at 4ghz except the number of cores. :)

[Sn0wm@n]

the 45nm vs 32nm thingy is only the size of the transistor inside that cpu die .....


at the same speed it wont do nothing .. except draw less power if they didnt add any logic etc.....

[Sushi Warrior]

No speed difference between for example a 920 45nm at 4ghz and a 980x at 4ghz except the number of cores. :)

Not really true, Intel has refined Nehalem. Just like how the i7 860 is a bit faster.

[Sn0wm@n]

Not really true, Intel has refined Nehalem. Just like how the i7 860 is a bit faster.


did intel add a new instruction set or something like that for the i7 860 ???

[Linuxfan]

Not really true, Intel has refined Nehalem. Just like how the i7 860 is a bit faster.
Wasn't aware of that. I'll look into it! :up:

[w4tch0]

The shrink from 45nm to 32nm by it self will not result in higher clock for clock performance. It will only result in lower power consuption and usually lower temps (higher possible max clocks). On the other hand, such technology upgrade is usually acompanied by other improvements as well, i.e. changes in logic, improved architecture, improved instruction set etc. Not sure if this is also the case of i7 965 vs. i7 980X though.

[Serpentarius]

:confused: :confused:

How big is the difference between a 32 nm and a 45 nm running at the same clock speed ?

If my program run mostly in only 1 core, could a 6 cores guives worst result than a 2 or quad core ?

32nm is smaller ... lesser heat ... generally assumed to be better o/c .............. but it's not always the same case ...

like example, the P4 Northwood 130nm vs P4 Prescott 90nm .... surprisingly, the Northwood o/c better than Prescott ... heck, even the heat output is lower than Prescott


If my program run mostly in only 1 core, could a 6 cores guives worst result than a 2 or quad core ?
yes. depends on the o/s ... but it's already rectified on WinXP SP2 ... if for somehow, you're using SP1, i recall the o/s will bluntly run on 1-core.

nowadays, the o/s (XP and above) would move out and try to run the program in multi-core environment ... if by somehow the programmer is a hard arse fella who strictly programmed it to run on single core, what else to say? i do not know about other o/s

hope this helps

[Sushi Warrior]

Wasn't aware of that. I'll look into it! :up:

Now, of course if you look at benches of i7 980X vs. i7 920 it won't be 50% better, as multithreading isn't perfect, but from i7 920 to 860 there is a jump and I presume a bit more to 980X if you have perfect multithreading. But either way, it is probably ~5-10% better core-for-core, clock-for-clock

[Movieman]

:confused: :confused:

How big is the difference between a 32 nm and a 45 nm running at the same clock speed ?

If my program run mostly in only 1 core, could a 6 cores guives worst result than a 2 or quad core ?

temps also
Gainestown(45nm qiads) at 3458mhz: 61C loaded
Westmere's(32nm 6 core) at 3458mhz: 51C loaded
same heatsinks,etc..same exact system just with different cpu's.
a 10C cut in temps..:up:

[breakfromyou]

temps also
Gainestown(45nm qiads) at 3458mhz: 61C loaded
Westmere's(32nm 6 core) at 3458mhz: 51C loaded
same heatsinks,etc..same exact system just with different cpu's.
a 10C cut in temps..:up:

So...even with those extra 2 cores, they still run cooler?! What about idle temps, and how does the temperature scale with overclocking on that specific heatsink. Not even 3.5 Ghz...come on now. Go for higher! 4 Ghz vs. 4 Ghz!

[Movieman]

So...even with those extra 2 cores, they still run cooler?! What about idle temps, and how does the temperature scale with overclocking on that specific heatsink. Not even 3.5 Ghz...come on now. Go for higher! 4 Ghz vs. 4 Ghz!

Idle temps are less but I didn't grab numbers and as to 4GHZ ahh, at 4200 temps are in the mid to high 60's with 1.4vcore..
Drop that to 4000 and mid 50's..
Shows that the TRUE's and fans have reached their limits.

[LarsMarkelson]

wrong thread whoops

[Sn0wm@n]

temps also
Gainestown(45nm qiads) at 3458mhz: 61C loaded
Westmere's(32nm 6 core) at 3458mhz: 51C loaded
same heatsinks,etc..same exact system just with different cpu's.
a 10C cut in temps..:up:


perfect post that OP could interpret the result!!!

[miahallen]

The only significant difference in clock for clock performance for a single threaded app would be the availability of 12MB of L3 cache on 32nm chips...45nm had only 8MB.
The downside of the larger cache pool is slightly slower access to said cache.

Overall, the performance clock for clock in single threaded applications will usually be within 1-2%...not enough to notice. ;)

[autoct]

So Im better waiting for next generation of chips in about a year since Im running my 965 at 4.27ghz and can expect arround 4.5ghz with the 980.

Thank you all for your reply