Slashdot Mirror
AMD Overclocks New Phenom II X4 To 7 GHz
CWmike writes "Advanced Micro Devices on Thursday introduced the latest member of its Phenom II X4 family of high-performance quad-core CPUs, which the No. 2 chip maker said it had run as fast as 7 GHz in extreme overclocking tests. Out of the box, the new X4 955 Black Edition, which is aimed at gamers and hobbyists, runs at 3.2 GHz, giving it similar performance to Intel's fastest desktop chips at lower cost, AMD says. The company was able to more than double the CPU's speed during its tests using extreme cooling technology that is not safe at home, said Brent Barry, an AMD product manager. The Web site Ripping.org notes that hobbyists with early access to the X4 955 chip have been able to clock it at up to 6.7 GHz. AMD said the chip was safe with fan cooling at up to 3.8 GHz."
AMD has been going belly up for so long now it was easy to write them off for dead. Yet, I'm tempted to pick up their stock. Has to do better than my NBFAQ.
I think there's still some brand loyalty in Opteron - I love mine and I still think an Opteron will be my next pick of CPU.
And, the newest go around of Ubuntu Linux has some new drivers for ATI cards that should improve those matters.
A 7ghz chip is a very healthy prize for AMD. I wouldn't expect them to advertise the power usage on such a thing, but hey, its engineerings, you can't have everything at once.
I like AMD a lot, and I just hope they succeed. I know that Nehalem from Intel is a strong series of parts, and AMD has a lot of work to do, but the capital costs are so high in chipmaking that it is doubtful we would see another competitor to Intel emerge in a generation if AMD goes out.
This is my sig.
To me the takeaway is just how little progress chipmakers are making.
Compare to the 1990s. x86 processors started the decade with the 80486 @ 33MHz and ended with the Athlon @ 1GHz mark and was doing more per clock for even more improvement than pure clock ratings would indicate.. Now in the decade we are about to close out we have managed to push that to around 3.5GHz and by the end of '10 we might hit 4GHz and eight cores (for those willing to spend serious coin) but work per clock doesn't seem to have improved at all and if anything have even slid back a bit.
RAM improvement have slowed down as well, probably because of Windows inability to get large deployment of 64bit editions limiting demand. The 1990s saw average ram go from 1-4MB to 64-128MB. It has only been recently that 2GB sticks went from exotic server stuff to mainstream.
Speed also isn't getting faster as fast as capacity is growing. Compare how many seconds it would take a 1990 vintage 486 to write to every memory location vs a modern machine. Same goes for disk access. Hibernation on a modern laptop is pretty much a dead issue since the time to write the whole memory load to disc is unworkable.
Democrat delenda est
... how long till we're able to capture the heat from processors and use them to cut power requirements for computers exponentially?
Look up the second law of thermodynamics.
Power goes in on the "work" side of the Carnot Cycle and comes out on the "heat" side. You can salvage a small percentage by running the heat through a heat engine on the way to the heat sink - more if you let the chip get hotter. But not a lot.
Further, the current technology can't stand being allowed to heat up - and its power consumption per unit of computation goes UP when it gets hotter. So even if you COULD put a bottleneck in the cooling (where you're normally spending more power to pump the heat away faster) to try to salvage some of the energy, you'll be running at a net loss.
Now if somebody wants to use ceramic, high temperature metal alloys, and low work-function oxides to build integrated circuits based on vacuum-tube technology they might be able to get away with it. But electrons tend to be even larger and fuzzier in vacuum than in condensed matter so you might not be able to get your scale down to that of even current integrated circuits, limiting your speed due to signal propagation time.
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
The bigger issue, here, is that cycles are getting cranked out faster than it's useful (or are getting to the point where an increase in speed is useless). Here's a little equation for you:
(speed of light)*(1/(7 GHz))
That solves to 4.282 cm. That's 1.6 in for people who don't speak metric. In the time that the processor does a single clock cycle, light in a vacuum can only go 4.282 cm. Electrons on a circuit can't propagate a voltage any further/faster than that.
This was illustrated by a Saturn V rocket test. After the test the space below the rocket was flush with Nitrogen to remove/dilute the harmful gasses, which could still explode in oxygen. Unfortunately they didn't flush the area with Air(79% N2 21% O2) before allowing 12 - 18 technicians walk into the area. They died before they even knew what was wrong.
I'm amazed at how many people work around the stuff and don't realize how dangerous it is. See parent for discussion of CO2.
That 21% Oxygen is really important.