So are you telling me the only way to measure performance is doing it using a benchmark that test several applications and then spits a number based on some subjective weighting?
It seems to me it is possible to bypass the subjective weighting and just choose based on the performance of applications that are most important for each user.
Or do we need to use 3dmark to figure which graphic card is faster at game x? I would just measure performance in game x.
And even if AMD came with their own benchmark. So what?
Would it be more or less serious than a benchmark from Intel? Would a benchmark support by AMD and opposed by Intel have more recognition?
I doubt it.
The part I bolded, it speaks to the rest of your post.
We aren't dealing with a benchmark here that is geared towards any given user.
The benchmark is intentionally crafted to be usable by procurement offices who are trying to justify volume purchases of computing equipment for a whole spectrum of internal users.
They aren't looking to complicate their job by expanding the test matrix to be "excel heavy for the 112 folks in accounting down in Dallas, and MS Access heavy for the 345 database design engineers in Milwaukee, and browser heavy for the 1134 call center employees in New Delhi".
These guys want a one-size-fits-all benchmark that lets them easily rank-sort the productivity and cost-benefits of a set of computers that they are on contract to purchase from DELL or HP.
Small business and out-of-home businesses that have a much narrower range of apps in-house would obviously seek to find benches that best proxy their apps before attempting to rationalize the cost-benefits for any given purchase.
Its really not so different from the small-scale HPC guys who look at sub-scores in SPEC to find the apps that best proxy their own (I look for the hartree-fock based quant calc apps) and then zero in on rank-sorting those sub-scores for their performance assessments and estimations.
But sysmark is not intended to be the benchmark for the individual user.
That said, this is why AMD should have at the very least taken the time to cobble together some sort of coalition and beta-benchmark (heck even tesselation for the heaven benchmark was done on short notice and little leadtime) to coincide with their announcement yesterday.
It undermines them and their claims when they have nothing to present in terms of corroborating evidence of the sinister things in Sysmark2012 or the benefits of adopting their proposed alternative benchmark XYZ.
Instead we basically have AMD telling literally thousands of large businesses and government orgs that they should ignore the one benchmark that the industry has adopted as the de facto standard and use nothing in its place...uhm that is a non-starter for everyone, why even propose it if that is all you got?
They need a practical solution to the problem, right now they got nothing but Sysmark. (by "they" I mean the businesses and agencies that are required to rely on sysmark for procurement justifications)
Facebook
Twitter