MAX, Re: "Most people won't really read details about system configuration, they'll assume that what they see is what they'll get in Best Buy. Following your reasonning, we should in fact use the dual channel DDR board that Intel devellopped but we don't since its not really available."
For the first part, I see a very small intersection of people that both read benchmarks, and buy computers at Best Buy. I think this fact largely makes that point moot. Second, my reasoning does not suggest what you propose; rather, it is the opposite. I believe that "best of breed" and "available" should be the criteria on which to test a chip. Volume on PC1066 RDRAM systems is small, but it is available, and it's the best way to get the full potential from the Pentium 4. Springdale is not available, nor will it be for at least 6 months, so that does not comply with my criteria.
Re: "But another strong argument is that Intel abandonned RDRAM and is phasing it out. So it doesn't really inform us on what performance will be since it's not the future."
I'm not sure how long RDRAM will continue to be available, but my presumption is that Intel will make it an alternative until such time as they have a viable replacement. My previous post included dual DDR as an example for an additional technology that will allow someone to unlock the full potential in a Pentium 4 CPU. You presume that one standard will end before the other arrives, but I see no evidence to support that.
Re: "as for KT333, you know as well as me that assynchroneous designs doesn't help performance much... it's wasted bandwidth."
AMD should use the KT333, and as soon as it's *available*, the KT400. These would allow for the "best of breed" Athlon configurations, which have been my point all along.
Re: "I don't know how honnest you are when you say that you think those benches you skipped weren't interesting, but what I read in your post is a selection of benchmarks that, with 2 exceptions, have all in common that they favor P4 over Athlon so it makes me wonder what does really interest you?"
Let's not get emotional here. I am not being unreasonable. All I am saying is that Sandra and ViewPerf seem to offer scores that are simply not indicitive of real application performance. Note that I also prefer not to include the Sandra or PCMark memory bandwidth tests, even though those *really* favor the Pentium 4. Application performance simply does not match those results. I have given my reasons, yet you have not offered as good a reason why they *should* be included.
Also, like you have admitted, several applications in the Tom's Hardware suite favored the Athlon, and those also interested me. Unfortunately, the review did not choose additional applications that I have seen elsewhere that favor the Athlon, like the game Serious Sam. That "interests" me, even though it favors the Athlon, so I do not see the bias that you are clearly trying to indirectly connect to me.
Re: "I think the fact that AMD changed his formula is already a step in the right direction, and with larger cache and faster bus speed in the pipeline, they'll have even more occasions to keep the rating accurate."
You have actually made a mistake in this comment. The latter half seems to imply that the first half was not adaquate, which is what has been my point all along. In other words, AMD's future enhancements will be needed, simply so that they can "grow into" their already aggressive model numbering. Therefore, the implication is that their current numbering is inadaquate to describe the real performance difference between the Athlon and Pentium 4. I do not believe that AMD lives up to their model numbers any longer. I have proved under one set of applications that they have in the past lived up to their numbers. For the present, though, the faster FSB on the Pentium 4 chips give them an advantage over the Athlon, even with the model number adjustment.
As you said before, this is probably due to measuring with DDR as opposed to RDRAM, which is another thing that I disagree with. Especially in light of the SysMark fiasco, I hope you are not suggesting that it is "all right" for AMD to switch from testing with respect to RDRAM to testing with respect to DDR, without making that clear in their rating methodology. Once again, we are being reminded that benchmarking a PC is up to a lot of interpretation. You can analyze a number of scores, and come to a conclusion about your particular usage model, but applying any kind of general single number score is ambiguous in many ways. You seem to even agree with me on that point.
wbmw |
