To test the 3GHz Core 2 Extreme QX9650,
we used the new X38-based Asus P5E3
Deluxe motherboard with Corsair Dominator
DDR3 RAM rated for operation at 1,800MHz
(but operating at DDR3/1333), a 10K WD
Raptor hard drive, and Windows XP SP2.
First, we ran our benchmarks on the cur-
rent speed champ: a 65nm, 3GHz Core 2
Extreme QX6850. Then, we swapped the
proc for the 45nm QX9650.
Not all die shrinks have gone well
for Intel. The shrink from the 130nm
Northwood Pentium 4 to the 90nm Prescott
Pentium 4 was a whoop-de-frickin’-do
affair. Our reaction to this shrink is the
exact opposite. To cut to the chase, we’re
damned impressed by the 45nm Core 2
Extreme QX9650. The new CPU is generally
faster than its immediate predecessor, and
when you factor in its overclocking capa-
bilities and eventual performance benniesfrom SSE4, Intel has a real winner.
Of course, the performance spread
depends on what benchmark or application
you run. In our tests, we saw gains rang-
ing from a modest 5 percent to a very good
15 percent when running at the exact same
clock speed.
Most of the QX9650’s big gains came in
media encoding, where it was between 9 and
15 percent faster than the QX6850, running
our standard benchmarks. We also saw good
performance gains in applications tests and
respectable 11 to 13 percent gains in gaming
tests. Oddly, one test that Intel touts as being
faster with the QX9650—the FEAR bench-
mark—showed no performance gains in our
Lab when compared clock for clock.
Most notable is how well the QX9650
runs at higher speeds. Moving the QX9650
to 3.6GHz on a 1,600MHz front-side bus and
DDR3/1600 speed RAM, we saw completelyreliable performance (not one crash) using
a stock Intel air cooler and stock CPU volt-
ages. With tinkering, we’re confi dent we
could have seen 4GHz reliably. That’s a
good sign, as it has taken water cooling
and voltage increases to get the 65nm part
to the 3.7GHz range reliably. At 3.6GHz,
we also saw the QX9650 cough up a 20
percent performance increase—right on par
with the 20 percent clock increase in nearly
every benchmark.
This amounts to a big win for Intel.
Although some folks might be unmoved by
seemingly “marginal” gains in a clock-for-
clock test, they’re missing the point. Penryn
is designed for higher speeds. That it runs
on air cooling with stock voltage at 3.6GHz
only backs up our contention that Intel is
intentionally holding back the new chip’s
performance because it doesn’t have any
competition right now.Putting the QX9650 to the Test
PENRYN
Up Close and Personal60 MAXIMUMPC DECEMBER 2007
BENCHMARKS
CPU Intel Core 2 Extreme QX6850 Intel Core 2 Extreme QX9650 Performance DifferenceIntel Core 2 Extreme
QX9650
Overclocked
CLOCK SPEED 3GHz 3GHz 3.6GHz
Synthetic Gaming
3DMARK06 OVERALL 12,772 12,888 0% 13,284
3DMARK06 CPU 4,775 5,049 5.7% 5,935
3DMARK05 CPU 18,012 17,765 -1.4% 19,830
VALVE PARTICLE TEST (FPS) 103 116 12.6% 139
Gaming
FEAR AT LOW RESOLUTION (FPS) 311 313 0% 410
QUAKE 4 AT LOW RESOLUTION (FPS) 204.9 227.4 11.2% 272.3
WORLD IN CONFLICT AT LOW RESOLUTION (FPS) 185 205 10.8% 240
Applications
PREMIERE PRO 2.0 HDV (SEC) 1,577 1,426 10.6% 1195
PHOTOSHOP CS2 (SEC) 183 178 2.8 % 159
PROSHOW PRODUCER MPC TEST (SEC) 869 821 5.8 % 687
CINEBENCH 10 10,884 11,798 8.4% 14,216
BIBBLE LABS & NOISE NINJA (SEC) 1,740 1,718 0% 1,459
Encoding
MAINCONCEPT REFERENCE AVC 1,661 1,522 9.1 % 1,271
AUTOGK XVID 418 382 9.4% 315
AUTOGK DIVX 6.7 416 362 14.9% 305
Synthetic application
SCIENCEMARK 2.0 1,719.61 1,803.14 4.9% 2,154.19
VALVE MAP COMPILE TEST (SEC) 132 136 -2.9% 113
PCMARK2005 OV 9,665 9,736 0.7% 10,909
CPU 9,658 9,643 -0.2 11,587
MEM 6,517 6,675 2.4% 7,964
GPU 13,253 13,495 1.8% 13,876
HDD 7,019 7,014 0.1% 7,039
Our test platform includes an Asus P5E3 Deluxe Wi-Fi mobo, a 150GB Western Digital Raptor hard drive, 2GB Corsair DDR3/1800 memory, an Nvidia GeForce 8800 GTX videocard, a PC Power and Cooling Turbo-Cool
1,000 PSU, and Windows XP Professional. Best stock-clock performances in bold.