How much better is Nehalem really?

H

Hiyruu

2[H]4U
Joined
Jul 28, 2001
Messages
3,611
Will there be a Nehalem that's twice as powerful as an E6600?
I currently have the E6600, and don't plan to upgrade until I can get double its performance.
 
Depending on how you look at it, there are already CPU's that are twice as powerful as a stock E6600.
 
nehalem will probably double the performance in most situations, and quadruple it in others.. all while fitting in the same thermal envelope for the higher clocked variants, and cutting the power to almost 1/4 in the mid-lower end....

overall i think it will be very impressive
 
Hiyruu said:
Will there be a Nehalem that's twice as powerful as an E6600?
I currently have the E6600, and don't plan to upgrade until I can get double its performance.
Click to expand...

Depends on how you quantify performance. In any case, I would just wait and find out. Still a lot of speculation at this point.
 
1. Speculation
2. Speculation
3. Speculation

Rumors are saying a bigger performance jump than going from p4 to 775 chips. Personally I would wait for the refresh probably around late 09.
 
Dunno ; I must say that I don't like its cache structure.

Very small L2 ( 256KB ) coupled with large , shared L3 ( 8MB ) on a multithreaded CPU..hmm.Looks it was designed for servers , like K10.
I wouldn't be surprised it to lose core/core in desktop apps to Conroe/Penryn.Basically it all depends on how fast that L3 is.
 
yes, I understand there is a benefit,but does the larger Die warrant the benefit in speed.

Say a C2D with 4mb cache, has 40% of its transistors in L2, and has 1B transistors (just go with it), now say you grab the same CPU an put on 2Mb Cache on it and is 10% slower.

C2D 4MB 1,000,000,000 = 100%
C2D 2MB 800,000,000 = 90% = +11% more efficient in power per transistor (die size)
 
Depends on the costs of said CPUs. Also there isn't anything necessarily wrong with a larger die.
 
cost more to make. and loosing efficiency on speed to die size matters.

Intel must have known this an decided to change there design to follow efficiency. -Pulling this out of my as on speculation
 
er, I was thinking the size of the cache and its level is not sole the determining factor, the caching algorithm, depth of pipeline etc., in short the entire pre-fetch architecture both hardware (cache size and levels) and software (cache management) determine the performance and it all needs to be matched to the rest of the CPU's architecture.

Besides I thought the Nehalem is all about, well mostly about and I guess it is related, the on-board memory controller reducing latency and getting the memory bandwidth up there so that a dirty cache is not such a big deal.

I am sure the engineers optimized all of it , but optimized for what ? Desktop, Server, both ? And then the bean counters showed up and said we want so and so many dies from a wafer and it all went out the window :eek: :mad:

We shall see but it will be fun no matter what, at least to me, everything since P965 intro and C2D launch has been soooooo boring.
 
Hiyruu said:
I currently have the E6600, and don't plan to upgrade until I can get double its performance.
Click to expand...

The Q6600 has twice the performance of the E6600 - I just did that upgrade.

Of course that is for some of the apps I run - YMMV. For some apps, you can get twice the performance with a minor CPU upgrade and a major HDD upgrade - if the apps are significantly I/O bound.

You have to define what performance means before someone can answer that question.
 
Will some Nehalem processor be able to do calculations twice as fast as an E6600? Entirely likely, although until someone actually gets one to test, it's all speculation.

Will Nehalem make your games or web browser or OS twice as fast? No, graphics cards and Internet connections and hard drives are the limiting factors in most situations. For real-world situations, the perceived benefit of Nehalem is likely to be reasonably small, unless you regularly do the few tasks which are CPU dependent (e.g. video encoding).

Lots of enthusiasts will rush out to replace their QX9650s with Nehalem chips because they are numerically faster, but the main thing most will be getting out of it is feeling good that they have the fastest thing on the market, not day-to-day performance gains. Stick with what you have until it's holding you back, then upgrade - don't upgrade purely because something faster exists.
 
You must log in or register to reply here.
Back
Top