T
hey may look like mirrored surfaces to your eye, but the bot-
tom of your heatsink and the top of your CPU look more like a
Thomas’ English Muffin when examined under a microscope.
Think nooks and crannies, folks.
Thermal paste is thus used to fill the microscopic gaps between
the CPU and heatsink to ensure a solid connection between the two
surfaces and promote efficient heat transfer. It’s the peanut butter
between two slices of Wonder Bread.
This is CPU Cooling 101—everyone knows you need thermal
paste between the heatsink and the CPU—but what kind is best?
There’s the silver kind, the white ceramique, and the cheap goop
that comes in a little plastic sleeve with your heatsink. For the uniniti-
ated, it can be very confusing. Hardcore enthusiasts seem to always
endorse the popular Arctic Silver
5, but why? Is premium thermalpaste any better than the stuff that comes with your CPU?
To find out, I had Frozencpu.com send us its four most popu-
lar thermal compounds: Arctic Silver 5, Arctic Ceramique, OCZ
Silver 5, and its in-house copper-based compound dubbed
ThermalGrease 4. To make things even more interesting I bought a
tube of generic paste labeled “Silicone heat transfer compound” at
Radio Shack and threw it into the mix. To test the effectiveness of
the pastes, I ran our Athlon 64 FX-55 zero-point system at idle and
under full-load for an hour.
As the benchmark results show, there’s no clear winner (although
the Frozen CPU copper was the clear loser). Arctic Ceramique
performed well, but it was only 1 C cooler than Arctic Silver, so it’s
hardly a decisive victory. Even the generic crap-compound in the 12
ounce tube did just as well as the fancy-pants Arctic Silver 5. The
conclusion is thus: It doesn’t matter what kind of thermal compound
you use, as long as you use something.I
f eight channels are good, 16 channels must be better, right?
That’s what nVidia wants you to believe about its new nForce4 SLI
X16. The SLI X16 is the first consumer chipset to feature a pair of
x16 graphics lanes instead of the two x8 slots on nVidia’s original
nForce4 SLI chipset.
To suss out the real-world truth, I fired up Asus’ new dual x16
slot-equipped A8N32-SLI with an Athlon 64 FX-55, 2GB of RAM,
and a pair of GeForce 7800 GTX cards. After running some tests,
I moved the cards over to an older Asus A8N-SLI with the dual x8
slots outfitted with another FX-55. The difference? Nada. At leastnot in any of the benchmarks I ran.
Using FEAR, Quake 4, and 3DMark05 at high resolutions and
high AA settings, both rigs performed exactly the same. This
isn’t really a big surprise, as few apps really need the monstrous
8.8GB/s of bandwidth that dual x16’s provide per graphics card.
nVidia is reporting numbers that show the dual x16 slots mak-
ing a much bigger difference in performance—increases range
from 1.6 percent all the way up to 5.6 percent. Of course, nVidia
also used its higher-clocked 512MB 7800 GTX card, an FX-57,
and beta drivers available only to nVidia. Regardless, I couldn’t
reproduce the same performance boost on my hardware using the
publicly available drivers.
Does that mean I think x16 is useless? No. I think anyone pur-
chasing an SLI board for the first time should buy one with dual
x16 slots. It just doesn’t make sense to trade up your dual x8 SLI
system for a dual x16. Newer applications and graphics cards will
certainly start to require the additional bandwidth down the road,
but you shouldn’t lose any sleep over it now.Gordon Mah Ung
EVALU ATES THE NEW
NFORCE 4 CHIPSET
How does high-end thermal paste stack up against
the crud that comes with your cooler?nVidia’s nForce4 SLI X16 offers the first dual x16
consumer chipset, but is it worth an upgrade?70 MA XIMUMPC JANUARY 2006
in the lab REAL-WORLD TESTING: RESULTS. ANALYSIS. RECOMMENDATIONS
Silver thermal paste takes on ceramique, copper, and generic goop.JOSH NOREM
Takes up the
Thermal Paste
Debate
BENCHMARKS
Best scores are bolded. We used an FX-55 CPU on an Asus A8N SLI Deluxe mobo with a Gigabyte Neon Cooler 8-BL heatsink/fan. Idle temps were recorded after 30 minutes of inactivity. Full load temps were
recorded after running CPU Burn-in for one hour. All temps were recorded with the Asus A.I. utility.ARCTICSILVER 5 ARCTICCERAMIQUE SILVER 5 CPU COPPER THERMAL PASTE OCZ FROZEN GENERIC
IDLE (C) 36 35 36 37 40
100% LOAD (C) 54 53 53 58 54