Bulldozer was not actually 8 core. Are people still trying to peddle this misinformation? Holy crap.
But yes, you're right that it actually did well in *certain* multi-threaded apps. That was its strength. In an age when multi-threading wasn't all that prominent. And by the time it was, Intel had progressed to a point where it left AMD in the dust.
IT was an 8 core regardless of if you believe it to be so. The real question is are people still peddling the crap that it isn't an 8 core when literally legal cases were brought against AMD because nutjobs believed it to not be an 8 core and were thrown out because it was proven by anyone with a brain in the industry to be an 8 core chip.
If it wasn't an 8 core, it could not have had much lower single threaded performance, but managed to actually compete with Intel chips with only 4 cores when more than 6 threads were used.
Give a single viable explanation for a FX8350 to be uncompetitive in single or 4 threads with a quad core Intel chip with/without HT, but does become competitive with 6-8 threads if it didn't have more actual cores?
The industry by and large deems an integer block as a 'core', Bulldozer had 2 per module and 8 integer cores in a FX8350. The entire industry deems it an 8 core chip, the legal system deemed it an 8 core chip. Intel never once ever anywhere stated the FX8350 wasn't a real 8 core chip.... but you're right, it's totally not an 8 core chip.
You also managed to ignore the actual point of my post, which is, you're comparing benchmarks with an 8 core vs 4 core chip, in which it was ONLY strong in multithreading and fell behind in single threading to the current Zen benchmark situation in which an 8 core chip is competing against another 8 core chip. If Zen had noticeably weaker single threaded performance(>10% difference), then it could NOT compete with an 8 core Intel chip.
Lets say Bulldozer had 60-70% of the single thread performance vs the 2500k, and HT boosted performance of that chip by 5-35%. In single thread it was left behind, in 4 thread it was left behind, in 8 thread you have 0.6x8= 480% performance of a single Sandy core the 2500k had 4x 1.0 = 400%. the 2600k has 400x1.05 to 400x1.35= 440% to 540% performance, which is what we saw in those early Bulldozer benchmarks. in 8 threaded situations it was usually ahead of the 2500k, sometimes ahead of the 2600k, sometimes level, sometimes behind by a relatively large margin.
For two chips to have the same performance using 8 cores with 16 threads in a 16threaded situation, if AMD had significantly less single threaded performance that would carry over. If you assume Zen has only 70% of the performance of a 6900k single core, then it would only have around 70% of the performance when both chips were running 16 threads. The math doesn't work. For two chips with an equal number of cores to have relatively equal performance using 16 threads, it can't be far behind in single thread. Yes I simplified and ignored efficiency/scaling performance from 1-16threads. In reality you don't gain 100% performance, but neither AMD nor Intel chips will. Same goes for HT, you won't gain 100%. AMD could be 10% down in single thread and make it up via scaling the number of threads, but that still puts it in the same ballpark performance wise rather than 40% down, which still makes it a great chip.
