Slashdot Mirror
AMD Announces 65-nm Chips, Touts Power Savings
Several readers wrote in about AMD's entry into the 65-nm manufacturing generation. The company introduced four chips to be manufactured with 65-nm process in the first quarter of 2007 to replace existing 90-nm chips in their lineup. AMD is playing up the power economy of its line, claiming that even its existing 90-nm parts consume less than 50% the power of Intel's Core 2 Duo, averaged over a typical day's usage, while the new 65-nm chips will be even stingier with power. Next stop, 45-nm. The article says that AMD has a goal of catching up within 18 months to Intel's lead on the way to 45-nm technology.
I went from a pair of 2.8ghz xeons to a pair of Opteron 250HEs and its a world of diference.
"Have you ever thought about just turning off the TV, sitting down with your kids, and hitting them?"
Extremely nice. Most people dont account for the integrated memory controller reducing the power consumption of the northbridge either. As a whole Turion notebooks should be extremely power stingy.
Keep in mind, though that 1 atom wide for Si is 111pm (that's 0.111nm for those who dont know their units), so we're talking about ~400 atoms wide at 45nm. We've still got a little breathing room. I'm still waiting for Carbon Nanotubes to replace the Si interconnects.
Not really screwed, especially if they fail to catch up but do narrow the gap. AMD has always been behind Intel in terms of process technology. AMD has gotten ahead on some specific process tech, like copper interconnect and SOI, but in terms of overall process shrinkage and quality of process, Intel has always been ahead. AMD has been more behind in the past, but also a lot closer, sometimes only months behind Intel. Basically, they are familiar with this situation.
65nm was a particularly bad node for AMD in terms of Intel's lead. Their plan for 45nm seems to be shaping up better with Fab 36, so I expect them to be closer though probably not caught up.
The enemies of Democracy are
i can't wait till the hardware sites test a laptop version....
i disable sigs
So Intel is shipping 45nm? According to Bit-tech, these 65nm AMD chips are shipping today.
I agree, take things with a grain of salt until we see reviews. But you sound a little too skeptical of AMD to not be working for Intel.
How are sites slashdotted when nobody reads TFAs?
AMD has consistently been a step behind Intel in the die-shrinking competition. The fact that AMD's chips run as cool as or cooler than their equal-performing Intel counterparts even at the larger process is a credit to good engineering.
Of course, since Intel will also be moving forward over the next 18 months, they might end up still in the lead. Making a huge turnaround like they did (from Netburst to Core) in such a short time is remarkable. Creating an architecture and setting up the process and designing a generation of chips takes a looooong time. Kudos to Intel for that. Now the ball is in AMD's court, and they have to respond.
At the high end, of course, Intel rules. What about processors that normal people buy?
I was recently looking at a Core2Duo review, and noticed something interesting. At each brand's bottom end (E6300 vs. X2 3800), Intel outperformed AMD. The problem in my mind, however, is that Intel's bottom-end starts at a higher price point than AMD's. Very smart marketing move by Intel. However, If you match the processors price-to-price, the E6300 matches up against the X2 4200 (both currently around $180), and there is relatively little performance difference. In other words, the price/performance metric really isn't in anyone's favor.
Another smart (but a little slimy) marketing move Intel has made is in the power dissipation numbers. AMD quotes their CPU's maximum dissipation, and Intel quotes a power figure for some arbitrary (under 100%) CPU load. Intel looks good here....until you actually measure a system's power draw at the outlet, and find that again, there's not that much difference. This may (and probably will) drastically change as AMD's 65nm parts get out, but we'll have to wait and see.
Help find a cure for cancer. Join the [H]orde
So the 35-Watt X2 consumes 1.53 times as much power as the 65-Watt X2? Something is wrong there...
A 65nm 65W X2 idles with lower power consumption than a 90nm 35W X2. At full CPU load the 35W X2 would still have the edge. Since your average desktop PC spends most of its time idling this is not insignificant.
Of course but some of AMDs changes in the last year have made a HUGE difference on the notebook front. DDR2 consumes much less power than DDR1, catching them up to intel as far as that goes. Then theres their new 65nm proccess with much less power drawing parts. After notebooks start coming with flashmedia for most hard drive duties we should start to see smaller screens and ultra low power notebooks that last as long as cell phones when in use.
http://www.tomshardware.com/2006/09/25/green_mach
That's actually not true... As Intel introduced new CPUs going into the 775 socket, they started using more and more of the pins that were originally "reserved" -- so, in order to support a new CPU, certain additional pins would have to be tied high, low, to calibration resistors, etc. What that means is that while *older* 775 CPUs will run fine on new motherboards, the new 775 CPUs will not run on old motherboards, even with a BIOS flash.
For example, my 775 board running a P4 3GHz will only take P4s up to 3.4GHz or so, since the faster ones were new 65nm cores with slight pin changes. Pentium D, and newer Cores are also in the excluded category..
Never underestimate the bandwidth of a 747 filled with CD-ROMs.
The chips aren't behind. AMD is behind. The semiconductor manufacturers keep scaling down based off the leader which has always been way ahead of AMD. (NOTE: we are speaking of manufacturing leadership based on the smallest feature in the chip (not necessarily the performance of the CPU (and it matters even if there isn't a performance imporvement (which there usually is) because you can make more for the same amount of money and if they are better, then you can sell them for more too))). AMD is behind because Intel sold more than 40 million microprocessors on the 65 nm manufacturing nade before AMD sold even one. So they are little behind.
1) That would take into account the efficiency of the transformer, which doesn't impact battery life.
2) Many laptops run in a high power/performance mode when plugged in.
3) At least be sure to take the battery out of the laptop so it's not charging while you're measuring!
Intel puts twice as much L2 cache on their dies as does AMD for their high-end parts. The high-end Intel part has 4MB, the high-end AMD part as 1MB per core for a total of 2M.
Back when Itanium was sporting both impressive SpecFP scores and also gigantic caches, I wondered what would happen if you put a cache that size on a mainstream x86 part -- specfp is really more about cache size and memory bandwidth than it is about the floating point execution units, so I never bought the specfp rating as being due to an inherent advantage of itanium. Well, now we know.
While not very exciting architecturally, big caches are certainly a good way to boost performance. Intel gets those huge caches because of two things: First, they have the fab tech and fab capacity to produce gigantic chips with good yield and still make money. Second, Intel has the smallest cache cell size in the industry. So they are uniquely positioned to put big caches on mainstream parts, and they are taking advantage of it.
Intel, with the C2D, introduced a more aggressive out-of-order memory architecture, basically allowing any memory op, even with an unresolved address, to execute out of order, fixing it up later if there was a problem.
Slight correction on a very fine post: Memory ops that have their address are allowed to go out-of-order even if there are older ops which do not have their address (meaning in the case of an out of order load and a store without its address a possible forwarding case that will have to be fixed once the store gets its address). It doesn't make any sense to have a memory op go without its address, because you couldn't really do anything with it (without something like an address predictor).
The enemies of Democracy are