|
Page 1 of 2  This time we’ve compared two CPUs: Core2Quad Q9300 (this is the bottom model of current lineup) and Core2Extreme QX9650 (the top model of current lineup). We have experimented with QX9650, bringing down its multiplier so we got whole palette of Core2Quad processors from Q9x50 generation. To make evaluation of performances we have included in test even Core2Extreme QX6850 which is the top model Quad CPU from previous generation (65nm).
QX6850 has a same multiplier and FSB as new QX, so we can compare their performances directly. We also were interested in comparing differences between 6MB and 12 MB cache memory for Yorkfields (our impressions are presented in “L2 cache 6MB vs. 12MB”).
|
L2 cache 6MB vs. 12MB
We were very interested to find out is there any loss for Q9300 because
of shortage of half L2 cache memory. To find out more about that, we
have set clock at 2.5 GHz for QX9650, using multiplier 7.5 and FSB at
333 MHz. Results that we got are from comparing CPUs with identical
characteristics but with 6MB and 12 MB L2 cache memory. Picture always
tells more than thousand words so you can see performances presented in
table and all mysteries will be revealed
|
| |
|
4x4
For
all ones who are eager for megahertz’s and performances can look at an
interesting table column where is presented our QX test specimen
overclocked at 4 GHz. We accomplished this by raising multiplier at 12
so nothing, except CPU speed, didn’t affect on increased performances.
It’s clear that at this clock QX9650 is just outstanding and its
performances are the best in rendering software. Of course these
results would be even better if we overclocked FSB (Intel cores
communicate along external bus – FSB and on higher core clock FSB
becomes “narrow throat").
|
Basic improvement for Yorkfield against Kentsfield is made with finer
manufacturing process which effects on lower consumption (CPU is
working on lower voltage) and low thermal dissipation. Improvements
that make these new processors faster on same clock are certainly
greater amount L2 cache memory and SSE4.1 instructions. Also, we’ve
been occupied with overclock potential of new Quad generation and you
can read about our thoughts on that subject in section “4x4”.
Core2Quad Q9300
This model of Intel Quad CPU is manufactured in 45nm technology. Besides the fact that Q9300 has lowest multiplier (7.5x) it also has decreased amount of L2 cache memory (6MB). It shares with other models FSB speed at 333 MHz which is actually 1333 MHz QuadPumpedBus. It is clear that this CPU is not for demanding users but it will do the job for all users who use applications which can benefit from more than one core. Once more we remind you to read “L2 cache 6MB vs. 12MB” section and realize disadvantages of half amount of L2 cache. Of course, Core2Quad Q9300 has limited overclock range, caused mostly by low multiplier and it’s hard to pull out more than 450 MHz with stabile operation with Quads for FSB. This means that 3.3-3.5 GHz is a real limit, which is solid result. Also it is important to mention that Q9300 with 2 MB less cache memory than other Quads made in 65 nm is very fast and can keep up with rest of lineup. If you decide to upgrade to Core2Quad Q9300 we can only advise you to update BIOS to latest version so you can be certain that you have fully support for your CPU.
Core2Quad Q9450 & Q9550
We simulated these two processors by lowering multiplier on QX model. We did this because instructions and amounts of L2 cache memory are identical and they only have a lower multiplier.
Both CPUs have 12 MB L2 cache which made them full-blooded Yorkfields and as distinction to Extreme model you cannot change default multiplier value, so overclock is lead down only on lifting FSB. Q9450 has multiplier 8x and real overclock range about 3.5-3.7 GHz. On the other side Q9550 is a very serious overclocker with multiplier 8.5x and achieved 3.8 GHz, even 3.9 GHz with quality cooling. Results for default clocks can be seen in table.
Core2Extreme QX9650
This is, at the moment, the most powerful CPU from Intel and it works at identical clock as its predecessor QX6850 (3 GHz). Default multiplier is 9x but like any other member of Extreme edition Intel CPUs, multiplier is unlocked with possibility to set it to value up to 29. This actually leaves you incredible number of possibilities for overclock, regarding the fact that for Wolfdale and Yorkfield you can change multiplier in steps of 0.5. Regarding to all improvements and higher efficiency for same clock despite old Extreme edition, it wasn’t hard at all for QX9650 to beat QX6850 in tests. The interesting fact is that Core2Extreme QX9650 at 4GHz on daily basis, can be truly accomplished (section “4x4” and Test Results). As expected, this CPU doesn’t have a rival yet, neither by performances, nor by price. This CPU can be a choice for most demanding users, who wants to squeeze out every MHz from their system. As we expect from Intel to produce QX with FSB at 400 MHz we’ll hold ourselves from giving it InsideHW Choice Award and instead we are giving InsideHW Approved Award to Core2Extreme QX9650.
|
