Dual-Core Shoot Out - Intel vs. AMD
sebFlyte writes "The Intel vs AMD battle of the benchmarks continues. ZDNet is running its rather comprehensive-looking guide to a side-by-side test of Intel and AMD's dual-core desktop chips, the Athlon 64 X2 3200+ and the Pentium D 820. They look at pure performance, as well as the difference it makes to apps you might use on the desktop. In the end, AMD comes out as the winner. From the article: 'AMD currently offers the most attractive dual core option. The Athlon 64 X2 3800+ may cost $87 more than its Intel counterpart, the Pentium D 820, but the AMD chip is a much better performer. It also uses considerably less power.'"
Or, put another way, the bottom-of-the-line AMD 3800+ is less than 1/3rd the price of the top-of-the-line Pentium 840 EE ($328 vs. $999), yet it still beats it in most of the benchmarks.
Too bad they didn't compare the Pentium D 830 in the benchmarks - this is closer in price to the AMD 3800+
HIV Crosses Species Barrier... into Muppets
"A kernel compiled for a single CPU is faster than a kernel compiled for multipe CPU's, even when you only have one CPU. This is why OpenBSD has two kernels: 1) one cpu and 2) multpiple CPU's. The main developer of DragonBSD said that his preference is single CPU, performance wise (I'll leave that as a Google exercise)."
Dillon said he felt FreeBSD's focus on many CPUs to the exclusion of single-CPU performance was a mistake, not that single CPUs are preferable.
Also, most desktop workloads benefit from having two CPUs, it helps responsiveness quite a bit (even on OSes with good schedulers like Linux). There is overhead for the locking in the kernel, but the benefit almost always outweighs the cost.
I rarely criticize things I don't care about.
The ultimate multi-core processor technology is VLIW (or EPIC as INTEL calls it). The cores are broken up into lots of tiny pieces, instructions are distributed through various pipes and run through whatever is available in parallel. The Itanium processor is Intel's EPIC problem child. Too complex, too much heat. Maybe it is just a bit too early for this technology. I think Intel could try to start a "mobile" Itanium project. They were quite successful with their Pentium M. Maybe that will give Intel an advantage. ;-)
Or, Intel designs a dual-Alpha processor to beat AMD, but that sounds not like Intel, does it? Someone at AMD who might like the idea?
Your PC may have Intel inside, but did you know that Intel's fabs have VMS and Alpha inside?
Okay. According to this page, at full-tilt the Pentium D 820 consumes 130.6W, while this page says the Athlon 64 X2 3800+ consumes 89W. So, how long would the Opteron have to run at full blast to make up the difference in cost of $87? Last month I paid $0.078 per kilowatt-hour. This seems to be reasonably average for the United States. 130.6W - 89W = 41.6W difference between the two. Some back of the Google-calculator math reveals: (US$ 87) / (41.6 W * ((US$ 0.078) / (kW * Hr))) = 3.05871582 years. A not-insignifigant amount of time. If you're in an area where electricity is more expensive like New York or California, the amount of time is even less!
Feel free to correct my math!
It's also worth noting that Intel's marketing machine has been an enormous success. For many average-joe computer users, AMD is in the same category as Linux and Firefox -- something nerds get excited about, but (they think) will destroy their computer or break the intarnets.
Even some of the more tech-savvy people I've run into aren't aware that an AMD will be essentially transparent to the user, except in that it has historically been more bang for the buck, especially for the gamer. And even when you tell them of a personal preference for AMD, or certain AMD chips, for whatever reason, the vast majority will still assume that since everybody (even Apple, soon) has "Intel inside," Intel is the way to go. Even if you convince them that AMD is a processor and not some kind of nerd voodoo magic, they won't be compelled to switch unless a computer salesman tells them it's cheaper and better while they're at the store.
By way of analogy: Every time I see a Dodge commercial for some HEMI-powered truck, I think "Oooh, it's got a HEMI. That must mean it's better than my poor HEMI-less Chevy." Do I know what the hell HEMI even means? Not at all. Which is why I'll have to ask one of my car-nut friends if I ever (god forbid) truck shopping. It's pure marketing and name recognition. People trust Intel becuase they think it makes the world go round. AMD is some communist revolutionary nerd chip from hell. (OK, that's exagerrating. But you get the idea.)
Personally, I don't use dual cpus for servers because they are faster. As you say, they are not, kernel wise anyway. Dual CPUs do offer a higher availability, and the ability to crank the crud out of one cpu (compiling, etc.) while you can still get stuff done with the other.
Thats called scaling.
I would rather have dual 1gz than single 2.5ghz any day.
Me too. Personally and professionally I am simply able to do more when I have more processors available. In fact, a researcher friend of mine has a single CPU Intel machine with hyperthreading. His other researching buddies like to run CPU intensive programs wherever they can find a spare processor. It was irritating him that people were running programs on his box and it wasn't very responsive. He would renice the process, and that helped some. He then enabled hyperthreading, and then he didn't notice when people were running on his machine anymore.
I've done benchmarks with "normal" applications, and overall I get the best performance when doing X number of things in parallel where X == the number of processors, cores, or "hyper-whatevers". Its that simple. I'm saying "best performance" not a linear or superlinear performance, but oftentimes its at least I get 30% more out of enabling hyperthreading. It also just makes the machine more smooth and interactive.
I welcome the day when every computer has 30 or so processors. The more the better. Just so long as they go completely to sleep when not needed or in use. Someday.