Date: Tue, 17 Apr 2001 17:11:13 -1000 (HST) From: Vincent Poy <vince@oahu.WURLDLINK.NET> To: Linh Pham <lplist@closedsrc.org> Cc: Jeremiah Gowdy <jgowdy@home.com>, Kris Kennaway <kris@obsecurity.org>, Marc W <mwlist@lanfear.com>, Kyle <freebsd@sysmach.com>, <questions@FreeBSD.ORG> Subject: Re: the AMD factor in FreeBSD Message-ID: <Pine.BSF.4.31.0104171707140.4840-100000@oahu.WURLDLINK.NET> In-Reply-To: <Pine.BSF.4.33.0104171735110.45373-100000@q.closedsrc.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 17 Apr 2001, Linh Pham wrote: > On 2001-04-17, Jeremiah Gowdy scribbled: > > # This is not entirely true. Although the P4 has a 400mhz bus speed, actual > # testing shows that the Pentium 4 puts out almost the same throughput as it's > # 100/133mhz counterparts, it simply takes more clocks to do so. The > # applications in which the Pentium 4 wins are SSE/SSE2 optimized. These are > # not all that common. If you're referring to MPEG-4 encoding as tested in > # Tom's Hardware, remember that the Pentium 4 got the crap kicked out of it in > # the non-SSE2 non-3DNow! optimized version of the test, and since the > # application wouldn't even BE SSE2 optimized if it weren't for Intel not > # liking the results of the benchmark, I don't really care much for the Intel > # optimized benchmark. If Intel were going to go around and optimize > # EVERYONE's software, that would be another story. > > I forgot to include the note about SSE2... thanks for mentioning that :) > > The Pentium 4 does suffer from not only a crippled architecture but also > from the higher memory access latency, thanks to DRDRAM. The longer > pipeline affects the core speed, not the bus side (or, at least not > directly). The longer the pipeline, the clock rate must be higher to > run at the same speed as a processor with a lower clock rate and a > shorter pipeline. > > The reason why the Pentium 4 has a 100Mhz QDR (quad data rate) FSB is to > match the bandwidth of the dual Rambus memory channel architecture... > not because the pipeline is longer. > > # Not to mention the fact that the Pentium 4 clocks down as the processor > # heats up, causing a Pentium 4 1.5ghz to run at 750 mhz when the load is 100% > # (1.0) for an extended period of time. No wonder the Pentium 4 1.5 ghz loses > # head to head with an Athlon 1.2 ghz in non-Intel optimized testing. The > # processor speed steps like a laptop depending on the power/heat conditions. > # Even calling such a cpu a 1.5 ghz is a stretch in my opinion with that > # underclocking. The P4 will run at 1.5 ghz when idle, but then clock down to > # 750 mhz when the load is 1.0. What the heck is the point of that ?! It's > # more than happy to burn away power idling at 1.5 ghz, but when the processor > # is actually needed to get some work done, it has to underclock to avoid > # overheating. "1.5 ghz, except when you need it." > # > # http://slashdot.org/article.pl?sid=01/04/13/2041238&mode=thread > # http://www.inqst.com/articles/p4bandwidth/p4bandwidthmain.htm > > Kyle from HardOCP posted on Sunday 04.15.2001 refuting the article that > was written about the Pentium 4 throttling down. There isn't a direct > link to Kyle's note, but you can find it at http://hardocp.com. > > According to Kyle, the throttling option can be enabled/disabled from > within the BIOS setup and that the processor shouldn't be throttling at > all if there is proper cooling. > > # As for the AMD+SMP vs Intel+SMP, I can't say regarding the FreeBSD support, > # however, AMD's SMP is supposed to be far faster than Intel's because it has > # a Point-to-Point bus for the SMP connection, meaning _each_ CPU has a > # dedicated 200mhz (100mhz DDR) connection to the bus, when on an Intel SMP > # motherboard the two or more cpus will share the same 100mhz bus. That means > # on a 4 way SMP Intel system, each cpu will get an effective 25mhz access to > # the bus under full load in theory. By the same theory, AMD cpus would each > # have their own 200mhz dedicated connection to the bus, even in an 8 cpu > # setup. AMD cpus support up to 14 way SMP. The PIII can only support 2 > # cpus, and the Xeon can only support 4. http://www.sandpile.org > > The new Paliminos and the new C-series Athlons run on a 133Mhz DDR FSB, > not at 100Mhz, but I think the 760MP chipset does (officially or > unofficially, I don't know) support 100Mhz DDR FSB. Jeremiah has a very good point on the price/performance ratio since the AMD wins a few times over the Intel's in terms of performance. I know this might be offtopic but I was looking at the distributed.net rc5 and the AMD does 3.5 keys/sec per Mhz on the Athlon while the Intel PII/PIII/Celeron's all do 3.3 keys/sec per Mhz but the PowerPC G4 does like 8.1 keys/sec per Mhz which seems to smoke the Intel/AMD platforms. Now, since Mac OSX is out which is based on FreeBSD even though their kernel is Mach based, would a G4 400 like the Cube be more powerful than a 1Ghz AMD or Intel? Cheers, Vince - vince@WURLDLINK.NET - Vice President ________ __ ____ Unix Networking Operations - FreeBSD-Real Unix for Free / / / / | / |[__ ] WurldLink Corporation / / / / | / | __] ] San Francisco - Honolulu - Hong Kong / / / / / |/ / | __] ] HongKong Stars/Gravis UltraSound Mailing Lists Admin /_/_/_/_/|___/|_|[____] Almighty1@IRC - oahu.DAL.NET Hawaii's DALnet IRC Network Server Admin 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?Pine.BSF.4.31.0104171707140.4840-100000>