Is there any reason to use FX CPUs right now?

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DrMrLordX

Lifer
Apr 27, 2000
22,931
13,014
136
Settles the power argument once and for all I'd say. There is no reason to go FX.

Wat.

What kind of vcore were they pushing? 262W? Seriously?

4M chips chew up power like hogs when you push high volts. Volt-cap those things to 1.33v-1.36v and you do not get power draw like that. Regardless, it won several benchmarks, so it's pretty obvious why you should get an FX and why you shouldn't.

Don't get one if you're a Handbrake or TMPGEnc fanatic, or if you're big into a lot of Adobe Illustrator stuff, I guess.
 

TeknoBug

Platinum Member
Oct 2, 2013
2,084
31
91
I find it good only for cheap VM box use right now: those AMD integer cores are way better than 8 hyper threaded intel threads which are more expensive: better/more resources for less $. Otherwise, an intel non hyper threaded haswell i5 is faster in pretty much everything for only a few $ more.

I have an i5 3550 (OC'd to 3.9GHz thanks to the ASRock board) and it makes one hell of a VM machine because of the VT-d letting VM have direct access to the hardware instead of the VM's virtual hardware, quite a nice thing to have.
 
Dec 30, 2004
12,553
2
76
AMD's FX line hasn't had a true refresh in a while. Recently some new models were released with an "e" suffix that indicates they're binned for better power usage characteristics, but AFAIK there were no design changes.

A brief history:

Late 2011, AMD released Bulldozer-based FX-81xx, 61xx and 41xx as their successor to Stars (Phenom II), which was largely a Core2 competitor. Bulldozer introduced "modules" which were basically core-pairs that shared resources. If both cores were in-use, they would suffer a performance penalty (something like 25% for each on BD?), but the advantage was that each module was significantly smaller than two cores would have been otherwise. It was up against Sandy Bridge, which had been on the market for around 9 months. Generally speaking, Bulldozer was slower than Stars and power consumption was also higher, despite a process shrink, and thus was pretty non-competitive with Intel's offerings, considering Stars was a Core2 competitor and Intel had their 2nd generation i7 out.


In mid-2012 Intel released Ivy Bridge, which was a die shrink and slight design change from Sandy Bridge. Performance was up 0-10%, while power consumption was down. Shortly after, AMD released Piledriver-based FX-83xx, 63xx and 43xx, which were a significant improvement over Bulldozer in single-threaded performance, clocked higher, had less of a penalty from module shared resources. Despite this, compared with Piledriver, an Ivy Bridge core was often more than 50% faster. AMD would sell you twice as many cores for the same price though, which resulted in FX-8 chips being a fair value at around $250. You could get performance that was competitive with an i7 (if your programs could saturate 8 threads) for $75 less, though single-threaded performance was still relatively bad, and power consumption (when all threads were saturated) was considerably higher too. Games such as Starcraft 2 and Guild Wars 2 performed badly on Piledriver, while Battlefield ran as well or better than on Intel's counterparts.

Since then, AMD has not released a new FX-based architecture on AM3, only a 2-module/4 core APU on FM2 called Steamroller, while Intel has released Haswell, which included another single-threaded performance improvement and efficiency bump, and is relatively close to releasing Broadwell, Haswell's successor. AMD has gradually lowered their prices, and it's arguable that at around $100, an FX-8310 would be a suitable chip for some. Multithreaded performance is equal or better than a Haswell i5, which starts at around $180, but power consumption when doing the same amount of work is much higher. Single-threaded performance on Haswell is 50-60% higher. AM3 is also a very old socket and lacks some of the features of AMD's more recent FM2 and Intel's 1150 sockets, which matters to some but often isn't a deal-breaker.

The FX-9xxx series are a recent release of chips binned to run at very high frequency, but also have a very high power consumption, so much so that AMD originally shipped them with watercoolers. They are also a lot more expensive than the FX-8xxx chips.

~

It's worth noting that most games, while bottlenecked by Piledriver's poor single-threaded performance, are still bottlenecked above 40fps minimums. This may improve when Directx12 is released too, but Piledriver will probably be more than 4 years old before we see the first games that take advantage of this.

Here is a benchmark from the Battlefield Hardline beta:

http--www.gamegpu.ru-images-stories-Test_GPU-Action-Battlefield_Hardline_Beta_2-test-proz.jpg


FarCry 4:

http--www.gamegpu.ru-images-stories-Test_GPU-Action-Far_Cry_4-nv-test-fc_proz.jpg


CoD Advanced Warfare:

http--www.gamegpu.ru-images-stories-Test_GPU-Action-Call_of_Duty_Advanced_Warfare-test-cod_proz_amd.jpg



It's worth noting that AMD's chips idle nearly as well as Intel's, and that multithreaded performance is significantly better than Intel's similarly priced i3:

65065.png


65067.png



I probably wouldn't put one in a gaming rig though.

we've already covered countless times the optimizations (and lack thereof) that are going on behind closed doors in these titles. For comparison, Cinebench and Handbrake.
 

Ramses

Platinum Member
Apr 26, 2000
2,871
4
81
Funny part to me, is even if those charts are biased, I still see the FX doing exactly what it should be. And pretty well for the money and it's age at that.
 

jhu

Lifer
Oct 10, 1999
11,918
9
81
we've already covered countless times the optimizations (and lack thereof) that are going on behind closed doors in these titles. For comparison, Cinebench and Handbrake.

my tests with Blender show similar results between FX and Intel's offerings.
 
Last edited:
Aug 11, 2008
10,451
642
126
Funny part to me, is even if those charts are biased, I still see the FX doing exactly what it should be. And pretty well for the money and it's age at that.

What exactly "should" it be doing? Running 30 to 50% slower in gaming? Because that is what the results show. And I hardly think you can claim bias against AMD from Hardline, considering it is
put out by Dice, who works very closely with AMD.

Now productivity benchmarks are another story. You can show pretty much any result you want, depending on which benchmarks you pick (cherrypick?).
 

HumblePie

Lifer
Oct 30, 2000
14,665
440
126
picked up a used FX6300 with 8gb of ram and a motherboard for $100 recently for a build for my wife. Matched that up with a 600w gold power supply I got for $30, a 280x for $150, $40 case, $10 dvd burner, $50 2TB drive, my old 120GB SSD, and a $10 4 heatpipe cooling solution. All in all a damn fine light gaming machine for my wife who love to also do video and photo stuff mostly with her comp and all for a cheap price. Basically a nice gaming comp for under $400 that is hard to beat for the performance. Is the power cost going to be higher than if I have gone Intel/Nvidia setup? Yes, but the premium I would pay for budget light gaming friendly parts that meet or exceed the performance of what I built would tack on another $200 or more to the build I have. $200 is going to take a lot of time to show up in the price difference in my electrical bills, if not years as she isn't on the comp more than an hour or two at night.

By then I'll probably due a new build with better power consumption and performing parts and sell her current comp out to someone for about 80% of what I paid.
 

Ramses

Platinum Member
Apr 26, 2000
2,871
4
81
What exactly "should" it be doing? Running 30 to 50% slower in gaming? Because that is what the results show. And I hardly think you can claim bias against AMD from Hardline, considering it is
put out by Dice, who works very closely with AMD.

Now productivity benchmarks are another story. You can show pretty much any result you want, depending on which benchmarks you pick (cherrypick?).


Pretty much, yeah. You don't pay less money and get less efficiency and not give up something most times. So yeah. They game "ok" and do more than OK in everything else for relatively little money while being amusingly different. That's been enough to make me buy and enjoy and build and sell several, and all mine have sold briskly for good money when I was done with them too. I really couldn't have asked for anything more of them without feeling greedy or unreasonable.

I don't blame anyone for buying Intel, I just ordered a 4790K since the resale was still so high on my AMD rig I let it go while it was still good economically and I had a few (quite a few) more bucks to throw at it. Anything less than that i7 would not have been enough of an upgrade to bother for my uses. I still wouldn't have bothered just for the CPU, but I got m.2/SATAe support, and a faster SSD and faster GPU in the process too, so overall it'll be fun and leave me a few things to upgrade later.
 

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Pretty much. FX-8xxx chips are priced near Intel's i3, and more often than not outperform an i5 which is twice as expensive in non-gaming tasks that scale with core count. The downside is an older platform (not important to most) and higher loaded power draw. And, "ok" game performance too. Right tool for the job; don't buy a CPU with poor single-threaded performance to run programs that depend heavily on single-threaded performance, and don't buy an i3 over an FX-8xxx if you're doing a lot of encoding and your time is valuable. (assuming you already have a video card to pair with it)

For a non-demanding user (e.g. my father), i5, i3, or FX-8 would all be overkill and I may as well throw in whatever is cheapest. A good deal on an AMD APU or Celeron would be equally relevant.
 

ultimatebob

Lifer
Jul 1, 2001
25,134
2,450
126
Heh. It's nice to know that my two year old Core i7-3770 can still blow the doors off of anything that AMD currently produces. :)
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
Heh. It's nice to know that my two year old Core i7-3770 can still blow the doors off of anything that AMD currently produces. :)

Not true, there s tasks where an FX8350 blow a 3770K, anyway this is telling that AMD CPUs are much underestimated while Intel lead is exagerated to the same extent...
 

Ramses

Platinum Member
Apr 26, 2000
2,871
4
81
I believe, or rather hope, it isn't that exaggerated since I just spent a bunch of money on an Intel setup. Unless it creates a time warp and allows me to go back and correct the numerous mistakes I've made in life however, I'll still think they are overpriced.
 

ctsoth

Member
Feb 6, 2011
148
0
0
Fow WoT single thread performance is much more important, as the game only uses 1-2 threads.
 

jhu

Lifer
Oct 10, 1999
11,918
9
81
Not true, there s tasks where an FX8350 blow a 3770K, anyway this is telling that AMD CPUs are much underestimated while Intel lead is exagerated to the same extent...

Like what? Rendering and transcoding are not it.
 
Aug 11, 2008
10,451
642
126
Not true, there s tasks where an FX8350 blow a 3770K, anyway this is telling that AMD CPUs are much underestimated while Intel lead is exagerated to the same extent...

Well, I suppose if you call a 10% lead in 3 or 4 benchmarks out of 30 or more "blowing it away", you could be correct. FX 8350 vs 3770k

Of course comparing to current generation Haswell would make it even worse.
 

Ramses

Platinum Member
Apr 26, 2000
2,871
4
81
WoT is all set for 2006 multicore lol
wonder how available/affordable modern CPU's are in belarus?
 

Ramses

Platinum Member
Apr 26, 2000
2,871
4
81
Of course comparing to current generation Haswell would make it even worse.

Could do something nutty like compare it to a like-priced contemporary from back when it was released. That'd put one around an i5-3xxx I believe, $200ish new, mid-late 2012. Way on out of the i7's price/performance realm in any case.

I guess it's flattering in a way that we keep comparing them to i7's and that they are still worth talking about.
 

jhu

Lifer
Oct 10, 1999
11,918
9
81
The 3770K does better in Cinebench 11.5 only due to this soft being Intel optimised and giving this latter a 15% advantage, same thing in Euler 3D wich is exclusively Intel optimised, (these are FPs benches) that is, only with exclusives optimisations Intel CPUs show superior perfs...

As much as I like Povray, let's compare with a renderer that has much more relevance to a lot more people: Blender. The FX 8350 is slower than an i7 2600.
 

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
Well, I suppose if you call a 10% lead in 3 or 4 benchmarks out of 30 or more "blowing it away", you could be correct. FX 8350 vs 3770k

Of course comparing to current generation Haswell would make it even worse.


I already debunked what you call 30 benchmarks as being few benchmarks that are relevant, let s count them and name them to see what are thoses 30 benches :

Handbrake 1

Handbrake 2

Thoses two one i wont discuss, then les see what follows :

Agitsoft sub task 1
Agitsoft sub task 2
Agitsoft sub task 3
Agitsoft sub task 4
Agitsoft sub task 5
Agitsoft sub task Total score
Agitsoft sub task CPU mapping

That s seven benches extracted out of a single one, but hey, it inflate the numbers in the good direction, but let see the rest :

3 versions of Cinebench in single thread, that is, of softs that are used only in multithread but it help to scores three wins, besides what is the relevancy of 3D particle Movement, a self penned bench that show a Baytrail being faster than a 4GHz Richland.?.

And sorry, the FX is also faster in Winrar despite what AnandTech is displaying, at HFR it s as fast as a 4770K, so what happened at At..?.

If you want to look at real benches take a look here, only professional and renowed softwares, not Sysmark and other 3D Particle rubbish :

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

And just imagine where the 3770k would stand on this list, and you can make the comparison with Haswell, see how the FX is aging more than well..
 
Last edited:

Abwx

Lifer
Apr 2, 2011
11,885
4,873
136
As much as I like Povray, let's compare with a renderer that has much more relevance to a lot more people: Blender. The FX 8350 is slower than an i7 2600.

I provided a rendering task that show the FX contradicting your sayings and your answer is to downplay the score under the pretense that Blender is more relevant..

Of course you are cherry picking, in Cinebench, wich is intel optimised the FX is slightly ahead of the 2600K, i guess that it took a badly AMD optimised soft to try making a point, as i said in the previous post Intel CPUs need exclusive optimisations to be better and Blender just show that it s the case.

Besides, about Blender :

22220.png


22221.png


http://www.anandtech.com/show/2978/amd-s-12-core-magny-cours-opteron-6174-vs-intel-s-6-core-xeon/7

LAWL...
 

jhu

Lifer
Oct 10, 1999
11,918
9
81
I provided a rendering task that show the FX contradicting your sayings and your answer is to downplay the score under the pretense that Blender is more relevant..

Of course you are cherry picking, in Cinebench, wich is intel optimised the FX is slightly ahead of the 2600K, i guess that it took a badly AMD optimised soft to try making a point, as i said in the previous post Intel CPUs need exclusive optimisations to be better and Blender just show that it s the case.

Besides, about Blender :

22220.png


22221.png


http://www.anandtech.com/show/2978/amd-s-12-core-magny-cours-opteron-6174-vs-intel-s-6-core-xeon/7

LAWL...

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.

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.
 
Aug 11, 2008
10,451
642
126
I already debunked what you call 30 benchmarks as being few benchmarks that are relevant, let s count them and name them to see what are thoses 30 benches :

Handbrake 1

Handbrake 2

Thoses two one i wont discuss, then les see what follows :

Agitsoft sub task 1
Agitsoft sub task 2
Agitsoft sub task 3
Agitsoft sub task 4
Agitsoft sub task 5
Agitsoft sub task Total score
Agitsoft sub task CPU mapping

That s seven benches extracted out of a single one, but hey, it inflate the numbers in the good direction, but let see the rest :

3 versions of Cinebench in single thread, that is, of softs that are used only in multithread but it help to scores three wins, besides what is the relevancy of 3D particle Movement, a self penned bench that show a Baytrail being faster than a 4GHz Richland.?.

And sorry, the FX is also faster in Winrar despite what AnandTech is displaying, at HFR it s as fast as a 4770K, so what happened at At..?.

If you want to look at real benches take a look here, only professional and renowed softwares, not Sysmark and other 3D Particle rubbish :

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

And just imagine where the 3770k would stand on this list, and you can make the comparison with Haswell, see how the FX is aging more than well..

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