Is there any reason to use FX CPUs right now?

[Abwx]

Right, because you don't cherry pick results? Like the double pass h.264 encoding? Interesting you don't like the first pass in which the 3770k is clearly much faster.

I provided the link so it s not cherry picking, besides you know very well that the second pass is much more time consuming, just look at the FPS, most of the work is done on this pass.

And now you link to Blender 2.5 alpha results? 2.5 alpha came out in 2009. No one uses that anymore. We're now on 2.73. The results I posted are all custom compiled to the processor. That's as good as it's going to get.

That s not the point, obviously Blender is even less optimised than Cinema 4D for AMD CPUs, as said you are cherry picking softs that are obviously not optimised for AMD, once softs are correctly optimised the picture is different, another prove?

Let s see 3DS Max when the FX was released :

http://www.hardware.fr/articles/880-7/rendu-3d-mental-ray-v-ray.html

Averaging the two renderings times the FX is slightly better than the 3770k and noticeably better than the 2600K, and 3DSMax is ultra optimised for Intel, just that it s also optimised for AMD to some extent.

 

[Yuriman]

A lot of this seems to be beside the point. What software do people use frequetly, today, which the FX performs well in? I understand it's difficult to find reviews using new software on older hardware.

 

[Abwx]

A lot of this seems to be beside the point. What software do people use frequetly, today, which the FX performs well in? I understand it's difficult to find reviews using new software on older hardware.

The FX performs well in any generic task but if more is needed it has the necessary grunt for HPC, the softs that typicaly require this kind of CPU are the kind of ones used by Hardware.fr, they selected excellent softs to check a CPU capabilities, these are only real world softs and they updated it with recent versions.

http://www.hardware.fr/focus/99/amd-fx-8370e-fx-8-coeurs-95-watts-test.html

 

[Abwx]

I didnt pick the benchmarks, this site did. Of course any benchmark that AMD doesnt win in your universe is not valid.

You are relying to straws, i didnt say this, neither did you comment on the validity of the benches i questioned, do you think that using all subscores of Agistsoft as well as the total score to get 7 results is a valuable methodology.?

I think that it is not, do you support the contrary.?

Also i estimated that using 3 times three Single Threaded tests of 3 different versions of Cinebench is not relevant, do you hold as relevant thoses three scores.?.

Why not using 7 Zip or Winrar in single thread also why you are at it, and every multithreaded soft as well, with a single thread run...?

 

[Yuriman]

The FX performs well in any generic task but if more is needed it has the necessary grunt for HPC, the softs that typicaly require this kind of CPU are the kind of ones used by Hardware.fr, they selected excellent softs to check a CPU capabilities, these are only real world softs and they updated it with recent versions.

http://www.hardware.fr/focus/99/amd-fx-8370e-fx-8-coeurs-95-watts-test.html

Thanks. Pretty strong showing in the non-gaming benches, seems to be that on average, 8370E < i5 4670K < 8350 < i7 4770K < 9590 < i7 4790K, with the AMD chips being rather cheaper than their performance bracket suggests, though anyone using one would need to buy a video card.

Curious, how many of those are GPU accelerated?

 

[Ramses]

A lot of this seems to be beside the point. What software do people use frequetly, today, which the FX performs well in? I understand it's difficult to find reviews using new software on older hardware.

That's a good question.

I seem to use the same software the same way, more or less on both counts, as I did ten years ago or more. Today I have more internet bandwidth, multi monitors, I have one streaming netflix or running Google Music or Music Bee, I have two browsers, one work one mine, with 30 tabs each open, two dozen programs running in the background/sys tray, IM software of some sort, a VM for XP to run some ancient work software, and a remote desktop session open to work. It's more of it, but it's the same stuff I've always done. When I game, these days, I don't even close these things down. I can't speak to anything other than an FX with 16gb of ram, but with very few misbehaving software exceptions they game just as well with this stuff open as they do with it closed. Again I can only speak as to the 8350 and 9590, but anything not-gaming happens just next to instantly. I'm not sure how much faster one can get than instant. If I close Chrome with those 30 tabs open and re-open, it all opens in two or three seconds, tops. Takes a bit to refresh the pages but that's internet bandwidth. You can hear the CPU fans spin up a bit but it's a minor load. Other than gaming and multitasking, that's about all I regularly do. I rip some CD's and convert them to FLAC now and again, run BT, etc, etc, etc. Either I'm just a lightweight user or a lot of folks exaggerate their computing habits/requirements. None of this stuff requires even an 8350, but it offered a nice bit of headroom and it was cheap. I'm running the last few weeks a laptop with an i7-4510u, 16gb, 500gb ssd, via a DisplayLink usb hub for multimonitor, and while it's perceptibly slower than the FX box, it does OK for desktop use. If computing wasn't a hobby, id'd leave it be.

As to gaming, which if anything is supposed to be the achilles heal of the FX line, I gave my 8350 and 9590 both a pair of 280x's and played anything I wanted at either maxed out or very close to it on my plain 1080p60hz screen. I think Crysis3 and BF4 and Far Cry 3 are probably the newest and most demanding games I have, and they all ran great. (this cat has a bunch of good casual benches here http://www.overclock.net/t/1534128/vishera-vs-devils-canyon-a-casual-comparison-by-an-average-user) No crazy boggy slowdows, and no solid 100fps either, they just ran well, very playable. I'd have got around to FC4 and a few other newer ones eventually but Steam's wishlist "game is on sale" emails have totally ruined me from paying much for games, so that never happened. I expected C3 to drag but it ran great, they seem to have done a nice job on that engine. I also don't buy ill written buggy games, and especially not bad console ports or games that can't use a modern cpu/gpu.

My "benchmarks" don't have cool graphs and mostly involve my senses, but they all tell me the FX is still plenty viable if one is on a budget or not some sort of crazy power user in one aspect or another. There are even cases where it excels but I am no authority on those since I don't seem to do any of them either. All the test results people post over and over and over, I read them all before I got a 1090T to replace my aging 955BE years ago, I read them again before I swapped the 1090T for an 8350 once I discovered how little it'd cost after selling the 1090t, I read them yet again when I got the email price notification for what I'd told Newegg I was willing to pay for a 9590, and they all proved to be solid, peasant, relatively economical choices and I'm happy I made them. Hell they even sold for good money used, and quickly. I couldn't ask for much more out of the FX's I've had myself. I'm genuinely perplexed why they are continually compared to i7's that they were never intended to perform on the level of, and pleased that I run across so many people that are still getting good service from them. These threads keep happening because people are still asking and interested in AMD and the FX line because they have some good things to offer the right person.
It's not for everyone and that's perfectly OK, but to dismiss them as worthless is some sort of weird intentional ignorance.

I'm actually sort of disappointed leaving the 9590. It was a fun and interesting and entertaining chip. Observing it's heat production and power usage, which wasn't all that bad once I picked a smart case and cooler but I went through a couple before till I got there. It was interesting seeing how it performed in games that were by popular opinion out of it's league, how it dealt with crossfire(and how crossfire behaved), just seeing how well it did with so many (theoretical) strikes against it in general was pleasing every day.
I ordered this (http://pcpartpicker.com/p/DZ49vK) to replace it, isn't even here yet and it's anticlimactic. I'm sure it's very fast and it has some nice options on the board for I/O and it should be quieter and cooler, but now I've got the same junk as everyone else. Great, now what. Yawn.
I suspect as spring rolls around I'm going to move away from computers and back to cars for awhile again till something interesting happens. There is just too little improvement and everything is too fast to maintain this hobby the way I used to I think.

 

[Abwx]

Thanks. Pretty strong showing in the non-gaming benches, seems to be that on average, 8370E < i5 4670K < 8350 < i7 4770K < 9590 < i7 4790K, with the AMD chips being rather cheaper than their performance bracket suggests, though anyone using one would need to buy a video card.

Curious, how many of those are GPU accelerated?

You mean the softwares..?

These are purely CPU related, there s no GPU acceleration, besides i wouldnt put the 4670K above the 8370E in applications, not only the difference is a meaningless 0.2% but according to the reviewer the 4C/4T are maxed out on theses benches while the 4C/8T and 8C/8T are not.

Overall, as you point it, the perf/$ is clearly on AMD side, others will eventualy point the difference in power comsumption in HFR tests, but one has to consider the testbeds, on one side they used the most efficient 1150 MB they could find, wich has 10W lower comsumption than average such MBs, but they also picked the most possible power hungry AM3+ MB, surely that it help when doing overclocking tests but such a board is pointeless when testing CPUs like the 8370E, or almost any FX given the cost of this Asus Sabertooth.

 

[schmuckley]

All these massive Tom's-type bar graphs.
0 with Westmere-EP which scores higher than Intel 4-core or AMD 8-core.
Except in SuperPi
Board + chip = Around $300

 

[escrow4]

The FX performs well in any generic task but if more is needed it has the necessary grunt for HPC, the softs that typicaly require this kind of CPU are the kind of ones used by Hardware.fr, they selected excellent softs to check a CPU capabilities, these are only real world softs and they updated it with recent versions.

http://www.hardware.fr/focus/99/amd-fx-8370e-fx-8-coeurs-95-watts-test.html

Still using that tired old site? You never mention alternative graphs/sites. FX will fall over in anything that requires single threaded prowess, never mind the horrific power consumption:

http://www.techspot.com/review/943-best-value-desktop-cpu/

"When clocked at 4.6GHz, the FX-8320E used 63% more power on average in our application tests, 55% more when encoding and 27% more when gaming."

And you are tied to that ancient AM3+ socket (and 900 chipsets are basically tweaked/updated 700 series chipsets) that can't come close to 1150. Broadwell/Skylake will just put the final nails in the FX coffin.

 

[Ramses]

You say that last as if it was a hard won victory lol...

 

[TurtleCrusher]

Still using that tired old site? You never mention alternative graphs/sites. FX will fall over in anything that requires single threaded prowess, never mind the horrific power consumption:

http://www.techspot.com/review/943-best-value-desktop-cpu/

"When clocked at 4.6GHz, the FX-8320E used 63% more power on average in our application tests, 55% more when encoding and 27% more when gaming."

And you are tied to that ancient AM3+ socket (and 900 chipsets are basically tweaked/updated 700 series chipsets) that can't come close to 1150. Broadwell/Skylake will just put the final nails in the FX coffin.

A 3.2Ghz chip oveclocked to 4.6Ghz has high power consumption? That's news to me!

Any chip that's overclocked 1.4Ghz higher than it's base clock is going to have high power consumption. Whatever you're getting at, at least bring in some logic here.

 

[jhu]

I provided the link so it s not cherry picking, besides you know very well that the second pass is much more time consuming, just look at the FPS, most of the work is done on this pass.

The first pass is still 16% faster for the 3770k whereas the second pass is maybe 7% faster for the 8350. At best it's a wash.

That s not the point, obviously Blender is even less optimised than Cinema 4D for AMD CPUs, as said you are cherry picking softs that are obviously not optimised for AMD, once softs are correctly optimised the picture is different, another prove?

I'm not quite sure you understand what "-mtune=bdver2" means. Blender isn't optimized at all for any particular CPU. The most it does is detect SSE2. That makes sense because x86 is the most common platform. But it still compiles just fine for non-SSE2 x86, ARM, Power, MIPS, etc. They're not going to go out of their way to hand code Piledriver assembly.

BTW, please find me some software that is "correctly optimized" for Piledriver, whatever that means.

Let s see 3DS Max when the FX was released :

http://www.hardware.fr/articles/880-7/rendu-3d-mental-ray-v-ray.html

Averaging the two renderings times the FX is slightly better than the 3770k and noticeably better than the 2600K, and 3DSMax is ultra optimised for Intel, just that it s also optimised for AMD to some extent.

Wait, how exactly can you claim that 3DSMax is "ultra optimised" for Intel, yet the 8350 barely edges out the 3770k in Mental Ray? It's highly unlikely that the people responsible for Mental Ray and Vray performed any Piledriver specific optimizations.

But all of this is rather moot. AMD has pretty much given up on FX (and Opteron for that matter). FX is at best on par with Ivy Bridge and at worst, slightly slower than Phenom II. While FX languishes, Intel has moved on to Haswell already (and soon Broadwell). FX compares even worse to Haswell.

 

[soccerballtux]

The FX-8310 has dropped to as low as $89.99 after rebate:

http://forums.anandtech.com/showthread.php?t=2417172&highlight=8310

I am hoping we see that price drop further still.

when I got mine it was $99 - $25.

$100 for an 8 core is a steal. If I had wanted to spend $250 on the CPU then Intel would have been a choice, but they weren't. Not to mention, I can stream twitch no trouble at all.

 

[Abwx]

The first pass is still 16% faster for the 3770k whereas the second pass is maybe 7% faster for the 8350. At best it's a wash.

You could have done the maths, the first pass run at 163 and 140 FPs for the 3770 and 8350 respectively while the second passes run at 41 and 44 FPs, for a one hour film with 86400 frames the 3770K take 2637 seconds while the FX take 2580 seconds, a wash if you want, but we are far from your claim that the 8350 couldnt beat the 3770k in renderings and encoding, it s quite the contrary actualy and as such you are spreading falsehood.

I'm not quite sure you understand what "-mtune=bdver2" means. Blender isn't optimized at all for any particular CPU. The most it does is detect SSE2. That makes sense because x86 is the most common platform. But it still compiles just fine for non-SSE2 x86, ARM, Power, MIPS, etc. They're not going to go out of their way to hand code Piledriver assembly.

BTW, please find me some software that is "correctly optimized" for Piledriver, whatever that means.

Dont know for blender but this is the rendering soft that yield the worst results on the FX, while Povray, 3DS Max and even Cinema 4D show good performances, but since it goes in the direction of your bias i guess that we must hold the three others as negligible quantities, isnt it..

Wait, how exactly can you claim that 3DSMax is "ultra optimised" for Intel, yet the 8350 barely edges out the 3770k in Mental Ray? It's highly unlikely that the people responsible for Mental Ray and Vray performed any Piledriver specific optimizations.

But all of this is rather moot. AMD has pretty much given up on FX (and Opteron for that matter). FX is at best on par with Ivy Bridge and at worst, slightly slower than Phenom II. While FX languishes, Intel has moved on to Haswell already (and soon Broadwell). FX compares even worse to Haswell.

It is ultra optimised because they use all last uarch instructions of Intel CPU, check the difference between Haswell and the 3770K in this bench..

As for FX against Haswell, i prefer to check numbers than relying to urban legends like the one you argumented about :

http://www.hardware.fr/focus/99/amd-fx-8370e-fx-8-coeurs-95-watts-test.html

The 4770K 9% better than the 8350 in the application tests, and they are not all perfectly threaded according to the reviewer, is that what you call comparing even worse.?

 

[Idontcare]

Is there any reason to look at FX processors right now? Thanks.

Yes there are still reasons. (from this thread) But it all depends on the usage scenario.

For being a 32nm chip, it is quite capable of performing well. The power consumption at a platform level is a bit of a deal killer for folks who are TCO sensitive. But the upfront costs are quite reasonable as well.

The only thing that bothers me about my FX8350 is the same thing that bothered me about my Q6600's - there are much faster and lower power-consuming alternatives on the market.

It isn't enough of a concern to justify outright replacing an already purchased FX chip, but IMO it is enough of a concern to justify not buying one if you are in the market to buy another CPU at this time.

Go with the lowest TCO/performance option, not the lowest up-front cost/performance option.

 

[2is]

Good thing the super bowl was not played on your field. Nobody would know which way to go because you keep shifting the goalposts. You were the one who wanted to limit the discussion to the top ten Steam games. Now you want to include the top 100, as well as plan for two years into the future.

And I see you are using the classic AMD argument that cost matters when you buy the cpu, but not when you are paying for electricity. Whatever else you use power for is irrelevant when examining the cpus. The only difference that matters is the difference in cost between the cpus and the amount of cost savings from lower electricity use. If you take your own estimate of 60 watts, 4 hours per day 15 cents per KWH (pretty conservative for a lot of areas, including taxes and fees), that comes out to 13 dollars per year. If you keep the cpu for 3 years, that totals up at least half of the cost difference. Whatever else you spend money on is irrelevant to this particular comparison.

Multiply those electricity costs by 3 if your utility company has tiered pricing, and you're hitting the top tier like I am most months.

 

[DrMrLordX]

Total cost of operation is more of an enterprise or "power user" concern. If you are a self-builder who has one main rig and you run it maybe 3-4 hours a day, tops, then power consumption is less of an issue.

If your house is mega-computerized with machines in every room, performing work that is considered to be important or mandatory with run times in excess of 12+ hours a day, then TCO is more of an issue.

If you're a business . . . TCO is huge. But I digress.

Many single-machine users can tweak their budget accordingly, like relying less on AC/heat for personal comfort, turning out lights, and doing other things to mitigate the costs of running a CPU for a few hours a day with significant load. Only nitwits (like me) who peg their CPU at maximum clockspeed and vcore all day long and almost never turn off the box see any kind of real cost from electrical usage on a single machine.

FX becomes a problem when that high TDP mandates expensive cooling. Buying a $200+ FX-9590 + $60-$100+ cooling solution that may need a $100+ motherboard seems a bit goofy when you can put a $322 4790k with the stock HSF on a ~$110 Z97 motherboard. Heck you don't even need a video card for the 4790k if you are okay with HD4600 (4 the lulz).

FX shines when it's a dirt-cheap 8310 on a 970A-UD3p, or maybe even some cheaper 4+1 boards depending on how ambitious you are. You'll still have to spring for a cooler, but there ar a lot of options if your target is performance/watt. The sweet spot is probably 4 ghz - 4.5 ghz, depending on your needs and how well the chip responds to low voltage.

 

[TheELF]

when I got mine it was $99 - $25.

$100 for an 8 core is a steal. If I had wanted to spend $250 on the CPU then Intel would have been a choice, but they weren't. Not to mention, I can stream twitch no trouble at all.

I could stream anywhere no-trouble-at-all if I had a decent internet connection,and I only have a celeron...

Youtube video FarCry4 on the g1820 now thats what I call an APU, it actually has accelerates something (quick sync) that people do and will use.

 

[Abwx]

Yes there are still reasons. (from this thread) But it all depends on the usage scenario.

For being a 32nm chip, it is quite capable of performing well. The power consumption at a platform level is a bit of a deal killer for folks who are TCO sensitive. But the upfront costs are quite reasonable as well.

After reading through your testbed i realized exactly where you did a mistake that inflated the TDP, you ll notice that to this day you didnt correct the error in the protocol and still are using thoses irrelevant numbers as if it was a scientifical prove, true that in the meantime i didnt correct it myself in said thread.

Perhaps some help would be appreciated.?.

But before doing so i ll let you try to figure where the protocol went wrong, a hint, it s very usefull to read datasheets or any paper or article with accurate technical specifications...

 

[Idontcare]

Total cost of operation is more of an enterprise or "power user" concern. If you are a self-builder who has one main rig and you run it maybe 3-4 hours a day, tops, then power consumption is less of an issue.

If your house is mega-computerized with machines in every room, performing work that is considered to be important or mandatory with run times in excess of 12+ hours a day, then TCO is more of an issue.

If you're a business . . . TCO is huge. But I digress.

Many single-machine users can tweak their budget accordingly, like relying less on AC/heat for personal comfort, turning out lights, and doing other things to mitigate the costs of running a CPU for a few hours a day with significant load. Only nitwits (like me) who peg their CPU at maximum clockspeed and vcore all day long and almost never turn off the box see any kind of real cost from electrical usage on a single machine.

FX becomes a problem when that high TDP mandates expensive cooling. Buying a $200+ FX-9590 + $60-$100+ cooling solution that may need a $100+ motherboard seems a bit goofy when you can put a $322 4790k with the stock HSF on a ~$110 Z97 motherboard. Heck you don't even need a video card for the 4790k if you are okay with HD4600 (4 the lulz).

FX shines when it's a dirt-cheap 8310 on a 970A-UD3p, or maybe even some cheaper 4+1 boards depending on how ambitious you are. You'll still have to spring for a cooler, but there ar a lot of options if your target is performance/watt. The sweet spot is probably 4 ghz - 4.5 ghz, depending on your needs and how well the chip responds to low voltage.

I completely agree.

 

[Yuriman]

DrMrLordX, I'd like to add to the "if you are okay with HD4600" thought:

Where the FX shines is not in games, but in heavily threaded apps that are not GPU accelerated, as Abwx as suggested. I'm assuming a system made for useful productivity will not be overclocked. Because of that you're looking at $240 FX-9590 + cooling solution ($30 Evo should suffice?) + motherboard ($100 should be fine?) + video card ($45 minimum for something bottom-of-the-barrel from the last 2 generations), compared to a $322 4790K + $70 H97 motherboard.

Comes out to $415 vs $392, because of the need for a video card.

OTOH, if you're buying a high-end video card already, you're using the FX-9 outside of its optimal-use scenario.

I'm sold that the FX chips do have a place, but my use-case (personally) would not be well suited to one.

 

[frozentundra123456]

Yes, use case is important. Before we got into page after page of arguing about benchmarks, the original post was asked in regard to a mid level gaming rig. I would for sure not pick FX in that scenario.

 

[lamedude]

People still do 2 pass encoding? I thought we all moved to x264's CRF mode.

 

[svarog19]

Dying Light performs badly on AMD hardware, not because AMD's hardware is bad or weak, its because programmers/developers put more effort on Intel and Nvidia hardware which h ave more market share, if same amount of effort was put on AMD products then gap between would be small.

AMD CPU's are great for price for performance and FX 8350 is great for multitasking, just play your games and render videos in background as soon as you're done recording the footage of the game and then upload it to youtube if you have fast enough internet connection to play online while also upload videos.

 

[svarog19]

People still do 2 pass encoding? I thought we all moved to x264's CRF mode.

Should't x/h265 be the standard now?