good point, then again if that would've been the case, i wonder why'd they even released it in the desktop segment, it runs awkwardly horrible on usual apps, and compared with the older models it looks more "meh".
BD is an architecture designed for future workloads as well as current and future capabilities of the foundri(es) (GF) they're using to manufacture the design. The problem is they forgot present/past workloads. The drawbacks of the compromises/decisions they have to make in the design were too much.
Why they even released it at that state? It's simple. Delaying it more to improve it will only incur more costs. They can release it later, improved but costs them more R&D money and no income from it for now. Or they can release it now, steadily improve the design (more R&D money), get some money from selling it but suffer from less desirable performance. Which do you think will a cash-strapped company like AMD will choose? Which option is economically more feasible? They do not have the money to put out a new architecture like Intel every couple of years. The last "real" new architecture from AMD was used for 12 years.
Then there's the question: why not just shrink stars architecture to 32nm and add more cores? After all, BD seems to perform similar or worse than the older, mature arch. Doing that will incur some costs, and would be harder to integrate to AMD's general plan on heterogenous computing. Stars probably already reached it's full potential.
if what you'd say that the newer benches are more efficient that would mean the competition's score would increase as well, or are you saying the newer version is more bulldozer biased? or theres simply a hidden trick on running bulldozer that the older version lacks thats hindering bulldozer's performance?
Will newer benches improve intel's scores? Yes. But BD has some new instruction sets (AVX, FMA4, CV16, XOP). I have already posted earlier in this thread how some of these instructions can impact performance (which you probably just ignored because you're already on the bandwagon of immediately bashing AMD). There are also instructions that AMD doesn't optimize anymore. Take superpi for example. It uses x87 that's no longer supported (improved) by AMD. x87 has been superseded by newer instructions. Can you claim that the benchmark results represents its performance accurately?
In this thread, I'm probably the only fanboi left whose somewhat fine with BD's performance. Then again I have had lower expectations from it, especially in the last couple of months. I do not blame people for having high expectations though. I blame AMD's PR department for that.
You people seem to forget that AMD is a company trying to compete against a giant. Intel's
marketing budget is around the same as AMD's
market capitalization. R&D in this business costs a lot (and it's rising all the time) that simply getting close to the Intel's performance despite being tight in money on intended workloads is a nice thing. Give them the same R&D budget as intel does and you'll probably get something amazing from them.
