Why didn't they just call it a quad core with clustered multi threading...

Why didn't they just call it a quad core with clustered multi threading? I gets more performance gains than intel hyperthreading because CMT because is basically physical hyperthreading, there would be no controversy and it would've done fine in the market.

Attached: A79-8350_sell01_mo_ec_4904561.jpg (725x415, 61K)

Because core count is easier to understand.

Because it objectively has eight physical cores, dumbdumb.

Attached: 22110348362l.jpg (720x341, 39K)

>there would be no controversy
This controversy wasn't big. Most people just knew the per-core performance was shit.
>it would've done fine in the market.
It wouldn't, because single thread was worse than the last generation. The architecture was a fuckup, plain and simple. How bad it was is still exaggerated, but there was no saving it.

the cripple cores is the whole reason these cpus were junked and dirt cheap.

i5s trashed them in synthetics and real world.

It's still four physical FPUs pham, that's why it works like a hyperthreaded chip.

Attached: CPU.png (371x138, 6K)

Because it doesn't use clustered multi-threading?

8 cores each with one thread of execution, no SMT.

>2 cores in a module sharing FPU and cache
>Not a cluster

except it doesn't have double the performance in single thread.
A single thread can all the pipelines available, but AMD physically separated the integer pipelines and then had a weak FPU to boot.
I'm glad Ryzen killed that trainwreck.

I remember having my old FX-8320 I bought at launch. Got some good binning on it and managed to clock it to 5.5GHz under custom water cooling. It was a fucking furnace that still got outperformed by stock Haswell i5's.

It failed because gaymers and YT reviewers (coincidentally sponsored by intel btw) running single threaded 640x480 graphical benchmarks, people who do modern multi threaded work got i7 performance for i5 money.

It is eight really shitty cores. Should VIA call their quad cores a .5 core CPU with 8 way SMT?

This is why it was slow.
For an application using a single thread, Bulldozer simply had less resources to throw at a problem.
5 ALUs to throw at integer math on SB, only 3 can be addressed by the USR at a time, but once they're cruching something for X cycles, they can be left and the port used for something else until the result is needed.
Bulldozer has 2 ALUs per core. (4 total) Less integer math capability out the door and it's locked behind two threads.

Intel's 256bit FMULT can be packed into 2x128 or 4x64 or 8x32bit. (one 32bit fp will take the same amount of cycles as a 256bit AVX, but you can pack 8 of them into a single 256bit instruction)
AMD's 128bit FMAC can't be packed, only padded - (one 32bit fp will take the same amount of cycles as a 128bit SSE and you can't pack multiple, so use 128bit SSE or you're just wasting cycles) however the 2 FMACs can be paired to do a 256bit AVX in the same cycles as a 128bit SSE.
Just the above already shows why Sandy murdered bulldozer, but then...
Sandy also has a 256bit FADD (fp addition only) which means certain fp ops can be done ever faster and not waste time on the fmult pipeline.

Bulldozer was a pretty terrible idea all round.
SMT is great because it allows you divide up a very wide processor, so you can squeeze more in when running light thread, but allow a single thread to eat the entire pipeline if it needs to. CMT offers none of that with no clear advantage of its own.

Attached: bulldozer.png (1306x819, 159K)

a shit ton of people on Jow Forums don't even understand the difference between CMT and SMT

oh wait you're a nigger

>a cpu core has to have an independant floating point unit to be a cpu core

each FPU had SMT strangely enough.

a Bulldozer module on the same process node would have been a smaller die size than Intels core; maximizing die size efficiency at the cost of all else

Source?

CMT had much better multi thread scaling because each thread was a physical core. For integer performance bulldozer was actually competitive against i7.

I'll get the slide when I get home, it's hard to find on a phone

It did trash the i5 in rendering and cost the same.

>Tfw bulldozer may have sold like hot cakes if it were released today with multithreaded games
>Tfw it was too intelligent to have good performance

It feels cheap but calling it 8 core is the most accurate. Some CPUs have no FPUs at all, much less 1 per 2 cores so we cant call it fundamental for coredness.

So glad I bought an fx6350 instead of i3 3220. They were going for the same price but the fx aged so much better...

It has aged well, and people seem to forget is that even when it was AMD's highest end chip, the FX-8350 still cost about $200, you can't compare it to a $350 i7 CPU, but when you compare it to an i5 that sits in the same price range, it's no comparison how much overall compute power you get on the 8350.

>For integer performance bulldozer was actually competitive against i7.
It was also twice the die size

Not 2600mhz ram with HT at 2800+

it was crap then and its crap now
t. ryzen

Remember when CPU's didn't have FPU's, and the mathematical coprocessors were either on the chipset or AIC?
Yeah, didn't think so.

It was also twice the value if you got it for the right reasons.
It obviously wasn't for everyone and suffered in games because of the abysmal single core performance, but I don't regret buying my 8320 at all, it still works great for me.

Attached: rendering.jpg (460x320, 28K)

No it wouldn't, the IPC is so low that it stutters in many games even at higher resolutions. You really can't defend FX, the 2500K hilariously has aged BETTER despite what people said in 2013.

>You'll be asking yourself, "What competition?" In no time.

Attached: latest.png (2250x2324, 110K)

>shared l2
its a quad core individual cores don't share l2

You don't decide what constitutes a core.

Neither can AMD, that's why there asses got sued, but they got out of it because it does have 8 processing thingies in it, they're just bottlenecked really bad by the caches and quad fpus

>make architecture for servers
>has ridiculously high power consumption
What did AMD mean by this?

Not sure what they were doing, but AM3+ is a good platform for a poorfag like me who couldn't Afford an X79 platform and has a need for a cut down server CPU on the desktop.

Attached: BD-Die.jpg (2000x1452, 1.18M)

Basically this, you buy a cheap cooler for the 8300 series and its fine.

>physical hyperthreading
is this Jow Forums?

It kinda is when you think about it, you're cramming 8 threads of stuff down the four FPUs, the thing with FX is your software better leverage them or else, because it's physical.

Attached: 4018MHz x4.jpg (2897x2605, 3.82M)

>10 FPS
That crappy CPU is the least of your worries.

> AMD FX
I remember when they launched the FX-9590 for $800. Shit was the absolute firehouse and got BTFO by the i7 4770K on almost everything.

I don't even know what those dumbasses were thinking when my FX8350 is hitting 9370 speeds and there's nothing special about my particular chip and board, even when the price did drop, the FX8350 was cheaper and overclocked fine, nowadays the FX9000 chips are only a few dollars more if you want the best binned FX 8 core for undervolting, but that's about it.

I had an FX 8350 and OCed it to 4.7Ghz it was a good CPU great upgrade from my old Phenom II X4. I wish I had gone Sandy for sure since it would still be a viable platform now adays. my FX 8350 and mobo were retired for an X58 board an an X5675 Oced to 4.6Ghz and it stomped the living SHIT out of the FX 8350 in IPC and multi core at that speed. I am now running a Ryzen 7 1700x and I have to say it is good to see AMD finally got their act together and made a CPU worth a damn. The FX 8350 will forever hold a spot in my heart since it served me well but it's power draw, performance, and heat output has it forever living in my Spare parts bin.

Shame a 2008 CPU has better ipc then the 8350, but these things are a great workhorse CPU, I'm a thermally limited to 4.5 on air and I'm happy with it

Attached: 4.5GHz 1866 bench.png (418x720, 50K)

Nice OC! And yeah they are great work horses I have wondered if there is any way to under clock and Undervolt it and see if I can make a small low power High Core Server out of it and throw like VMware on it or something.

I've had mine prime stable at 1.19v 3.2GHz 8320e speeds, I tried 2.1GHz at 1.000v, a p-state of the chip, but it wouldn't post.

what was the power draw like at that voltage?

I'm not sure but I was getting much lower temps on a crappier cooler I had at the time, the 8320e has a 95w TDP.

>1.440-0.25=1.19v
Also I'm sourcing the default vcore of an 8320e form a video of a board with terrible VRM defaulting to that so it should work.

Attached: stock.png (1634x903, 1.19M)