Date: Thu, 19 Apr 2001 10:48:08 -0700 From: "Charles Burns" <burnscharlesn@hotmail.com> To: vince@oahu.WURLDLINK.NET Cc: lplist@closedsrc.org, kris@obsecurity.org, mwlist@lanfear.com, freebsd@sysmach.com?, questions@FreeBSD.ORG Subject: Re: the AMD factor in FreeBSD Message-ID: <F209gJKVApwOLZCHusM00002b18@hotmail.com>
next in thread | raw e-mail | index | archive | help
> > This depends on what you plan to do. The general consensus among the > > hardware reviewers is that the Athlon is overall faster than any other >x86 > > compatible CPU. > > Yep, that's what I read as well but are there any drawbacks to >being faster such as compatibilty and all that stuff? The last problem that AMD processors had that was in any way significant was when a few old K6 processors would become flaky with more than about 40MB of RAM. This was back when RAM was about $45/MB in the U.S. so most people never got that much anyway (and when they did they weren't using K6 chips) Oddly enough, as of late AMD's platform has been more reliable than Intel's. Intel released a chipset for the P3 that was designed exclusively for Rambus RAM. When Intel figured out that Rambus wasn't the greatest, they hurriedly implimented a "memory translator hub" that allowed the chipset to use SDRAM on top of the RAMBUS packet protocol. This obviously made the motherboard damn slow, but it also led to data corruption in certain cases which is a big nono. Intel recalled the chipset. They were in such a hurry to leapfrog AMD that they failed to properly test the chipset. This will long be a taint on Intel quality in the memory of many computer professionals. Intel also released a Pentium-3 at 1.13GHz. You may have noticed that you can't buy these--that is because Intel didn't seem to test them very much at all. Rather than learning from their mistakes, they rushed another product and had to recall it when almost none of them would even compile a Linux kernel. Their latest prize, the P4, has been the joke of the computer industry. "Why pay $1100 for a GHz Pentium three when you can pay $2500 for a lower performing 1.5GHz P4!?" In one article that I read, an Intel rep was quoted as saying "It's 1.5GHz. Of course it's faster." Ad nausium. Athlons have one current issue--heat. They are very stable at very high temperatures, but just in case, you should get a very good heatsink. Recently Anandtech (www.anandtech.com) tested several Socket-A heatsinks and crowned a king. I can't remember offhand what it was, though. With the 1.33GHz Athlon dissipating about 72.9 watts, you need some pretty decent cooling. This isn't to say that P3s and P4s aren't pretty toasty. > > > The only significant performance advantage that the Pentium 3 has over >the > > Athlon is that its l2 cache memory is _much_ faster than that of the >Athlon. > > If you are going to be running applications that for some reason depend > > almost exclusively on the bandwidth of the L2 cache (software with lots >of > > loops that are under 192K may be an example of this) than in some >situations > > a P3 at 1GHz will likely be faster than an Athlon at 1GHz. > > Hmmm, I guess that part is one I can't figure out since for >FreeBSD, would this really count as a typical server? It does help, but the Athlon seems to be slightly faster at the same clockspeed for the majority of tasks (and is available in higher clockspeeds) > Yeah, that's what I am concerned about. It seems that most things >are optimized for the Intel CPU's. While the FPU is faster on the Athlon >than the Intel, what about the non-FPU area? Theoretically the Athlon is 50% faster at non-FPU (integer) calculations. In practice, it is 5-25% faster, depending on the application. Don't worry too much about optimizations. The Athlon is designed to run P3 optimized software. The only optimizations that don't benefit the Athlon involve SSE optimizations (which make the P3 about 2% faster than the Athlon, depending on the application, in FPU intensive stuff at the same clock speed) and fixes for some of the P3's bugs, which the Athlon doesn't have. (Not to say that the Athlon is bug free, but all have been fixed in microcode because there aren't any known serious design flaws) > Speaking about DDR RAM, what kind of performance hits would there >be using DDR versus non-DDR RAM? None. DDR RAM is faster. Well, ok technically DDR ram has a higher latency, but if you use a chipset that is designed around low latencies like the AMD 760 chipset, this becomes a non-issue. (In fact, latencies are often lower than those of normal RAM) DDR stands for Double Data Rate. It's bandwidth is theoretically twice that of normal RAM. In practice the actual system performance is 1-10% higher, but this may improve as applications start to take more advantage of higher bandwidth and as SMP Athlons come out. Tom's Hardware recommends using ONLY the AMD 760 chipset for DDR Athlons. Tom's Hardware is usually right, so I would take that into consideration. I have an AMD 750 (the old AMD chipset) on my system now and it has been rock-solid even when overclocking my 1st generation Athlon from 500 to 800MHz. Stability is, of course, the main factor when considering server hardware. Asus generally makes the most stable motherboards. > > The Athlon is much, much cheaper. Motherboards, however, are more >expensive. > > The overall cost ends up lower with the Athlon, especially if you are > > considering the price/perormance ratio. > > Yeah, that's what I realized as well. It seems like the VIA and >AMD chipset based motherboards costs a lot more than the Intel variants. The EV6 bus is more complex and thus more expensive to implement. (FYI) > > The P4 is a different story entirely... I would avoid it like an old >Cyrix > > CPU if I were you. > > Even if it weren't slower than the P3 or Athlon in most software, the >socket > > is soon to be changed so you will be left without the ability to upgrade > > much in the future. The chip is terribly expensive (as is the rest of >the > > platform), has a short life, is amazingly inefficient with its >transistors > > and memory bandwidth, and is overall certainly something to steer clear >from > > until Intel fixes some of its unacceptable weaknesses. > > Atleast from the guys at Anantech, they are all anti-Intel and one >of the reasons is as you stated about the P4's socket. I guess the >choices were easier during the Pentium days since you can just pop in a >AMD K6 in place of a Pentium without a total reinvestment. I often hear "Such-and-such a site is anti [AMD/Intel/Cyrix] because of such-and-such." Usually Tom's Hardware is accused more than Anandtech because they are mor apt to be harsh on something that doesn't live up to the hype, but Anandtech isn't immune. I have followed many hardware sites for years and have even written a few reviews. With the exception of sites like AMDzone and Intelzone, they are anti-stuff-that-is-unwise-to-buy, not anti [insert brand]. Really, why should anyone care what brand a chip is? If a given platform is a good deal, great. If not, avoid it. I don't care who makes my computer stuff as long as they have a good rep, a good product, and didn't BS too much in their marketing. Intel just dropped P4 prices. If the chip somehow gets the best price/performance ratio, I will have to consider it in the future. > Thanks, I'm familiar with all of those. I guess I just wanted to >know how they do under FreeBSD since all the sites really benchmark it >under Windows. Ya... Isn't that irritating? I usually compile a Linux kernel with an optimized compiler and with the default compiler to test myself. Typically Windows software seems to perform similarly to similar Unix software, at least for me. _________________________________________________________________ Get your FREE download of MSN Explorer at http://explorer.msn.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?F209gJKVApwOLZCHusM00002b18>