What went wrong?

>>105804070 (OP)
Nothing
Was great for poorfag builds
i3 from intel was 150 bucks in stores here and i5 were 200
This was around 100

>>105804101
this

>>105804070 (OP)
it was a bad design but intel could have made it work
AMD/gf's 32nm was bad and late so AMD was fucked when they had a speed demon uarch that couldn't clock high

>>105804101
>Nothing
It had worse performance clock/clock than Phenom II.

>>105804165
Yes, BUT, it really wasn't all bad. Piledriver was far more efficient than K10, had updated instructions, and could clock much higher, led led to it being faster overall.

>>105804165
>Phenom II.
Nobody was buying phenoms in the 2010s

>>105804070 (OP)
Nothing
Debian kernel compiled in under 5 minutes

I had an fx6300 because it was cheap and it was shit.

something about multiple cores sharing FPUs. IDK, I'm a retard myself.

>>105804070 (OP)
Cores sharing crucial parts of the CPU. It's essentially hardware SMT. A 4c Bulldozer processor is in practice a 2c/4t processor.

>>105804070 (OP)
attempted a reverse smt design (multiple cores acting as 1) ended up with half baked cmt with low ipc, attempted to push frequency way up to compensate, launched before performance bugs were fixed due to launch pressure.

>>105804070 (OP)
shared FPUs in the compute modules, and the arch was more suited for low power conditions. the last mobile only generation of this lineage was really good under 15 watts, and it was decent enough for cheap PCs. on the right loads (non FP tasks) it was great at multi threading, but it was power hungry at high clocks.

>>105804964
no they were bad, phenom 2 wiped the floor with the orignal bulldozer chips in some tests, it was awful frankly, when the bulldozer chips did win it by a small margin, clocking the an x6 phenom ii to 4ghz would destroy any lead the bulldozer chip had easily even in multicore workloads.

>>105804070 (OP)
The main issue was the crappy GloFo node. they had to cut the cpu down significantly.
The shared FPUs are not an issue, you can test it yourself.
The shared "front end" was an issue, the cpu slowed down when a single module ran 2 threads.

>>105804070 (OP)
I had an FX-6300 from about 2013-2017 because my i7 totalled and I needed access to my computer IMMEDIATELY...I bought it + a motherboard because it was cheap, but I still didn't upgrade for another 4 years because meh.

AMD CPU's were just not good for the decade between the Core 2 Duo and the first Ryzens. But Intel got too complacent and lazy in that decade, and got completely fucked once Ryzen launched. It's been almost a decade later, and Intel still hasn't meaningfully responded to Ryzen.

Shame about AMD graphics though; the RX 480 came out in 2016 and then...nothing transformative ever again. Who are AMD cards meant to be for nowadays?

>>105806145
>Who are AMD cards meant to be for nowadays?
linux users since nvidia drivers are terrible on linux

It didn't have Radeon cores for the FPU units. That is literally why they bought ATI, used "the future is fusion" as their marketing tagline for years and developed GCN and they STILL DIDN'T ACTUALLY DO IT.

>>105807580
there was no software that leveraged the radie shadies

>>105806793
nvidia drivers are ok on Linux. I mean by windows standards anyway, because Nvidia drivers have issues on both whereas AMD is only bad on Windows.

>>105804070 (OP)
a lot mostly due to microsoft not supporting FX properly on windows which fucked it and then AMD for lying about how many cores there cpu's had

I also wish AMD had also released that quad channel enthusiast board for AM4 that they originally had planned

>>105804070 (OP)
Hated my 4100
Loved my 8350

>>105804070 (OP)
Nothing.
FX series CPUs still hold records for lowest and highest clock speeds.

>>105804070 (OP)
They couldn't keep up with Intel in single core performance so they had to start the moar coars meme too early.
>>105804165
That's like comparing sports cars with econoboxes by gallon/gallon. At the end of the day, the FX were faster CPUs than the Phenoms.

>>105804995
I did though. 2011-ish. Good CPU.