-
FX4320,FX6300 and FX8350 review thread
-
Re: FX4320,FX6300 and FX8350 review thread
ill still buy an old AMD 6 or 8 core just cause i know its 6 and 8 cores. the average person cant tell the difference between intel and AMD in a blind test. To add, the average person does no were near what a benchmark can do.
-
Re: FX4320,FX6300 and FX8350 review thread
That's a lot of reviews Cat. Thanks.
Which are your favourites?
-
Re: FX4320,FX6300 and FX8350 review thread
Quote:
Originally Posted by
KeyboardDemon
That's a lot of reviews Cat. Thanks.
Which are your favourites?
The ones which use up to date software. Toms Hardware and Hardware.fr tend to use newish test suites. The Tech Report does latency testing which is good. HT4U tends to take some time for reviews,but they tend to be quite detailed.
-
Re: FX4320,FX6300 and FX8350 review thread
Added some reviews - the FX6300 looks good.
-
Re: FX4320,FX6300 and FX8350 review thread
Are there any folding benches in those reviews?
-
Re: FX4320,FX6300 and FX8350 review thread
Yeah I've been wondering what's happened to all the folding benchmarks lately.
-
Re: FX4320,FX6300 and FX8350 review thread
-
Re: FX4320,FX6300 and FX8350 review thread
Well spotted. Seems about right, even fairly impressive, considering F@H uses almost exclusively FP ops AFAIK.
-
Re: FX4320,FX6300 and FX8350 review thread
Now to find F@H stats on an overclocked chip!
A bit of geekery on the side, assuming the FX system consumes an extra 40W more than its close performing rival, that's about £35 a year more in electricity. Meaning, if you intend to buy a dedicated system to fold for a minimum of one year, the FX system would be better value than the competitors similarly performing system if the latter costs at least £35 more.
So, buying an AMD FX dedicated folder isn't bad value, relatively... I say that because nothing can complete in terms raw crunching power, ppd/£ and ppd/Watt with multi-CPU systems that have a silly amount of threads. :(
... Now to find Opteron folding reviews! :D
-
Re: FX4320,FX6300 and FX8350 review thread
Quote:
Originally Posted by
DDY
Now to find F@H stats on an overclocked chip!
A bit of geekery on the side, assuming the FX system consumes an extra 40W more than its close performing rival, that's about £35 a year more in electricity. Meaning, if you intend to buy a dedicated system to fold for a minimum of one year, the FX system would be better value than the competitors similarly performing system if the latter costs at least £35 more.
So, buying an AMD FX dedicated folder isn't bad value, relatively... I say that because nothing can complete in terms raw crunching power, ppd/£ and ppd/Watt with multi-CPU systems that have a silly amount of threads. :(
... Now to find Opteron folding reviews! :D
You might want to consider this:
http://www.cpu-world.com/CPUs/Bulldo...20FX-8300.html
Also,remember Intel CPUs have more functionality on the CPU itself unlike an FX CPU. Hence,I expect motherboard choice might make more of a difference than with an Intel system.
-
Re: FX4320,FX6300 and FX8350 review thread
Another review which has the FX6300 and FX4300 has been added!
-
Re: FX4320,FX6300 and FX8350 review thread
Another FX4300 and FX6300 review added!! So it appears even in the worst case scenario,the FX6300 is only slightly slower than a Core i3 3225 in games but extends decent lead in multi-threaded games,and the overclocking results look promising!!
-
Re: FX4320,FX6300 and FX8350 review thread
Is it just me, or is it a little suspect that Intel's x87 performance is better than its SSE3 performance? I thought the idea of new instructions was to improve performance...
It's almost as if Intel looked at which benchmarks people were using in processor reviews, and worked out that by boosting x87 performance they could give themselves a significant benchmark advantage, even if it wouldn't pan out that significant in the real world ... or am I being too cynical? ;)
-
Re: FX4320,FX6300 and FX8350 review thread
Which benchmarks do you mean?
Intel's still have an inherited hardware x87 unit, AMD Bulldozer onwards don't AFAIK, which explains the performance gap, but as you say it's completely irrelevant to most any modern software. It seems like a strange thing to spend money on, but when you have reviewers still comparing CPUs with SuperPi...
-
Re: FX4320,FX6300 and FX8350 review thread
It appears to be a bit of a hyperthreading thing, actually. The hardware canucks review for 6300/4300 has a synthetic FPU test - VP8/SinJulia. The VP8 test uses SSE3, whilst SinJulia uses pure x87. All the processors except the i3s have higher scores for SSE3 than x87, but by varying degrees: the quad core AMD K10 and Piledriver parts have the biggest difference, with the K10 X6 and i5s having a fairly small difference - then there's the i3s, which are FASTER at multithreaded x87 than multithreaded SSE3 :O_o1:
SuperPi is the classic example, but I suspect there are others out there we don't know about. It'd be an interesting bit of research. I could quite fancy taking a variety of common benchmarking tasks, and re-writing/compiling them with different level of feature support... ;)
-
Re: FX4320,FX6300 and FX8350 review thread
Supposedly,Skyrim used some X87 stuff:
http://www.joystiq.com/2011/12/27/sk...e-performance/
This is probably why Skyboost seems to work well,and supposedly newer Skyrim patches have compiler optimisations:
http://www.rockpapershotgun.com/2012...of-skyrim-1-4/
-
Re: FX4320,FX6300 and FX8350 review thread
Quote:
Originally Posted by
scaryjim
It appears to be a bit of a hyperthreading thing, actually. The hardware canucks review for 6300/4300 has a synthetic FPU test - VP8/SinJulia. The VP8 test uses SSE3, whilst SinJulia uses pure x87. All the processors except the i3s have higher scores for SSE3 than x87, but by varying degrees: the quad core AMD K10 and Piledriver parts have the biggest difference, with the K10 X6 and i5s having a fairly small difference - then there's the i3s, which are FASTER at multithreaded x87 than multithreaded SSE3 :O_o1:
SuperPi is the classic example, but I suspect there are others out there we don't know about. It'd be an interesting bit of research. I could quite fancy taking a variety of common benchmarking tasks, and re-writing/compiling them with different level of feature support... ;)
From the description of the tests, it looks like they do different things so really shouldn't be on the same graph?
To me it just says that SinJulia runs the fpu flat out (so it's thrashing a lot on FX4).
The VP8 test seems to have enough slack that Intel threading or AMD shared FPU can schedule a second thread with ease.
Note that on FX4300 the SinJulia is roughly half that of the FX8350, it scales with cores.
VP8 test doesn't so I expect it is bottlenecking on something else like memory throughput.
-
Re: FX4320,FX6300 and FX8350 review thread
Bulldozer lacks X87 hardware.
-
Re: FX4320,FX6300 and FX8350 review thread
Quote:
Originally Posted by
CAT-THE-FIFTH
Supposedly,Skyrim used some X87 stuff:
I did wonder, given how much Skyrim favours Intel hardware. Wonder how many other popular games/applications use x87 - perhaps there's some justification in using SuperPi a a benchmark after all ;)
-
Re: FX4320,FX6300 and FX8350 review thread
Well still not SuperPi as it's horribly inefficient anyway. I think PiFast also uses x87 but it's far better optimised.
-
Re: FX4320,FX6300 and FX8350 review thread
Quote:
Originally Posted by
CAT-THE-FIFTH
Bulldozer lacks X87 hardware.
Don't know where you got that from, and it seems to run x87 code just fine.
Looking at that SinJulia bench:
http://images.hardwarecanucks.com/im...FX-4300-42.jpg
the 8350 has 4 FPUs in its 4 modules, so should have about as much peak performance as an 980BE. It has a bit more, which is in line with the higher clock speed.
-
Re: FX4320,FX6300 and FX8350 review thread
Yeah, the FPU throughput in Piledriver looks pretty good, particularly when using modern instruction sets (where the FX4300s 2 FPUs are a match for the i3s). The i3s clearly *can't* efficiently hyperthread modern (if you can call SSE3 "modern") instructions, but can efficiently hyperthread x87 - something Piledriver can't (despite the claims that the flexFPU could handle multiple smaller instructions simultaneously).
I wonder if this is actually a dispatch or decode problem? i.e. the FPU could handle the instructions but the single decoder can't decode enough instructions to push them through the pipe...? Perhaps the AMD decoder has to do more work to decode an x87 instruction, leading to a larger disparity in performance?
It also begs the question that if theoretical throughput is actually pretty good (i.e. the 8350 with 4 FPU beating the 3570k with 4 FPU in SSE3) why doesn't that translate to better lightly-threaded performance? Or is that down to lack of int throughput? Looking at the int performance in that review the 6 int cores of the FX6300 have roughly the same throughput as the 4 int cores in the 3570...
I wish I understood CPUs better... *sigh* ;)
-
Re: FX4320,FX6300 and FX8350 review thread
Quote:
Originally Posted by
DanceswithUnix
Don't know where you got that from, and it seems to run x87 code just fine.
Looking at that SinJulia bench:
http://images.hardwarecanucks.com/im...FX-4300-42.jpg
the 8350 has 4 FPUs in its 4 modules, so should have about as much peak performance as an 980BE. It has a bit more, which is in line with the higher clock speed.
AFAIK,X87 is emulated.
-
Re: FX4320,FX6300 and FX8350 review thread
I have added another review which contains the FX6300.
-
Re: FX4320,FX6300 and FX8350 review thread
Quote:
Originally Posted by
CAT-THE-FIFTH
AFAIK,X87 is emulated.
To be fair, integer x86 has been emulated for some time even in the Intel chips so that isn't a game changer :)
It has floating point hardware, it maps instructions to use that hardware.
-
Re: FX4320,FX6300 and FX8350 review thread
So good for multithreaded applications on a budget, bad for single theaded apps and gaming. The high power consumption doesn't exactly inspire trust either.
