Just swapped my 1055T for a FX-8350 - underwhelmed!

[GoogalyMoogaly]

Hi,

Just wondering if anyone else is in a similar predicament to me.
I recently swapped my 1055T @ 3.8GHz for a FX-8350 @ 4GHz (i.e. stock).
Now I appreciate I've not overclocked the 8350 yet, but I'm waiting for 8 Pack's guide so that I do it right.

However it's still running 200MHz faster on each core and has 2 extra cores.
So I'm quite disappointed that given this the performance seems to be barely any better or in some cases possibly worse than the 1055T.

Video encoding using Handbrake has gone from ~190fps on the 1055T to ~225fps on the 8350. Now considering it has 33% more cores and is running slightly faster on each core I was expecting a bigger difference.
In games, such as Guild Wars 2 my framerates actually seem to have dropped (I'm sure where it was previously getting high 40s it's now getting low 30s).

So is Piledriver really that bad core-for-core and clock-for-clock compared to the Phenom IIs?

 

[ubersonic]

 

[Marine-RX179]

Video encoding using Handbrake has gone from ~190fps on the 1055T to ~225fps on the 8350. Now considering it has 33% more cores and is running slightly faster on each core I was expecting a bigger difference.
In games, such as Guild Wars 2 my framerates actually seem to have dropped (I'm sure where it was previously getting high 40s it's now getting low 30s).

So is Piledriver really that bad core-for-core and clock-for-clock compared to the Phenom IIs?

Clock for clock the Piledriver FX8 is still a bit behind the Phenom II...you probably need at least 400MHz higher clock in order to match it.

As for Guild Wars 2, I play the game as well, and I can tell you that the game only use up to 4 cores, with each of the 4 cores on average would only use up to 75-80% max. So basically the two extra cores on the FX8 doesn't help in this game, and you need to clock the FX8 to quite a bit higher than your Phenom II in order to see SOME improvement (but FX8 is quite power hungry and run quite hot once you clock beyond 4.5GHz, so your existing CPU cooler "might not" cut it). For Guild Wars 2, something like a highly clocked Intel i5 would be much better.

 

[Hagar]

I did much the same upgrade from a 1090T at 3.8 to an 8350 at 4.6 (went straight to OC). Overall I thought it a much smoother experience, and a cheapish upgrade as I sold when 1090T were fetching £80 here. I did lots of comparisons and benches in a thread started by Gibbo a while back which were generally positive. I have had this now since October last.

In terms of system power consumption I was only finding a 60W difference between the two systems at prime + heaven 2.5 which is not bad considering two extra cores.

My only bugbear is I changed the motherboard from an M5A99X to a Crosshair V Formula and in my opinion it is not nearly so good (Asus ROG again )

 

[GoogalyMoogaly]

I've got the M5A99X EVO (Rev. 1).
How do you go about overclocking it? FSB, multiplier or both?
Do you have CnQ and C1E (or whatever those power saving things are called) enabled? All the guides I see say to disable them, but to me it seems silly the CPU sitting at 4.6GHz and 1.45v when it's idling.

I tried lower my voltage with the speed at stock by 0.06v (or whatever the smallest increment is) and it crashed the system within seconds in Prime95.
I did up the multi to 22 (giving 4.4GHz) with auto volts but that took me up to ~1.4v and after 10 minutes of Prime my temps were 56ºC.

 

[Hagar]

I've got the M5A99X EVO (Rev. 1).
How do you go about overclocking it? FSB, multiplier or both?
Do you have CnQ and C1E (or whatever those power saving things are called) enabled? All the guides I see say to disable them, but to me it seems silly the CPU sitting at 4.6GHz and 1.45v when it's idling.

I tried lower my voltage with the speed at stock by 0.06v (or whatever the smallest increment is) and it crashed the system within seconds in Prime95.
I did up the multi to 22 (giving 4.4GHz) with auto volts but that took me up to ~1.4v and after 10 minutes of Prime my temps were 56ºC.

I have used both at times, FSB and multi. I have gone up to a max 302MHz bus and 16x to get 10 runs of IBT at 100Gflops just to test. I benched a lot at 233MHz x 21.5 to give 5GHz at 1.486V. This was all on the M5A99X.

My long term OC was 215 and 21.5x on the M5A99X which was 4.6GHz CPU and gave 2GHz memory frequency using 1866Mhz memory clock x 215/200. I used a vcore of 1.42V vmem of 1.62V.

My current OC on the crosshair is 200 x 22.5 = 4.5GHz at 1.38V (0.94V idle).

I do use CandQ and C1E as like you I fail to see the virtue of running flat out 24/7 unless benching.

 

[biuro74]

Mate, you need to accept two simple facts just not to be disappointed, like now.

First - FX is NOT 8-cores. It's 4-modules, one of which consist of one full core and one integer unit. Treat module as "1 core + AMD's version of HyperThreading", so 8350 means 4 modules and 8 threads, which is closer to truth & benchmarks, unlike marketing blahblah-selling "8-cores CPUs".

Second - FX in single thread performance is worse than old Phenom II. DOn't worry, people responsible for faildozer has been "punished" and company roadmap has been changed. It's for future reference anyway, but for today accept this simple fact that clock by clock Phenom II is not so bad. FX overclocks better as a pint of sugar in all that mess.

 

[GoogalyMoogaly]

How does that help!!!

 

[Martini1991]

It is 8 cores, they're EXACTLY the same as each other, just they share resources, the problem is they're just weak and the sharing resources, at least in Bulldozer put a dampener on the scaling.
I'm not sure how bad Piledriver is for the module scaling, but I believe they're trying to rectify it in Steamroller.

 

[kaku]

I don't understand why people are adamant in the '8 cores' argument?

End of the day, you get decent performance at the low end, but outside of very specific tasks you need to go Intel if you want performance.

 

[GoogalyMoogaly]

I don't understand why people are adamant in the '8 cores' argument?

End of the day, you get decent performance at the low end, but outside of very specific tasks you need to go Intel if you want performance.

At the £155 price point the best Intel I can see is the i5 3470.

Going by these benchmarks the Intel doesn't seem to run away with it. Especially considering the AMD can be overclocked (I'd guess by 10% or more).

But yes, if you're willing to spend a bit more (see my sig) AMD don't seem to have anything to touch the SB-E chips.

 

[DrBombcrater]

First - FX is NOT 8-cores. It's 4-modules, one of which consist of one full core and one integer unit. Treat module as "1 core + AMD's version of HyperThreading", so 8350 means 4 modules and 8 threads, which is closer to truth & benchmarks, unlike marketing blahblah-selling "8-cores CPUs".

Google die shots of the FX-8xxx processors. The 8 integer execution blocks are clearly visible, there is no quesiton of them being 4 core processors.

The one area where the cores really do share resources is the FP units, which are essentially 4 256-bit FPUs capable of working as 8 128-bit units when necessary. But that's a slender thread to hang the '4 core' argument on since historically x86 processors lacked any FPU hardware at all for many years.

Getting back to the original topic, I've just been in a similar position to Googaly. My main PC has been running an i5-2500K @ 4GHz for a while now. But last week I bought an FX-8320 to play with, with the intention of using it as an upgrade to my second system (currently using an old Q6600).

And it turns out the i5 is getting demoted instead. All the games I play (admittedly only 5 or 6, so not a huge sample by any means) run as well on the 8320 @ 4.5GHz as on the i5, while video encoding and compiling are noticeibly faster on the FX. The only negative is the increased heat output, but personally I find that's not a deal breaker.

 

[james_2k]

At the £155 price point the best Intel I can see is the i5 3470.

Going by these benchmarks the Intel doesn't seem to run away with it. Especially considering the AMD can be overclocked (I'd guess by 10% or more).

But yes, if you're willing to spend a bit more (see my sig) AMD don't seem to have anything to touch the SB-E chips.

doesnt run away with it?? the i5 wins a lot of tests by a long way, especially the gaming ones. also the i5 can be overclocked well, much more than 10%

amd are just for chartable types who for some reason have an attachment to the company from when they actually made good cpus.

 

[Marine-RX179]

And it turns out the i5 is getting demoted instead. All the games I play (admittedly only 5 or 6, so not a huge sample by any means) run as well on the 8320 @ 4.5GHz as on the i5, while video encoding and compiling are noticeibly faster on the FX. The only negative is the increased heat output, but personally I find that's not a deal breaker.

Em...either you should drop the overclock on the FX to 4.0GHz, or up the overclock the i5 2500K to 4.5GHz for a fair comparison.

"All the games you play" doesn't say a lot (it could well all be high-threaded EA FPS titles)...go try 90% of the other games other there that use 4 cores or less (for example as Guild Wars 2, Assasin Creed 3, Hitman Absolution, PlanetSide 2, Hawken, Sleep Dogs etc) , the i5 2500K would leave the FX-8320 behind in the ditch if both were on the same clock.

Also, your preception of your FX-8320 is "faster" than your i5 2500K on gaming could well be under the influence "it's new and shiny it feels faster than my old i5 2500K" sympdrom. Also if the graphic cards you ain't the same, that's not like for like comparison neither.

 

[cheesyboy]

Em...either you should drop the overclock on the FX to 4.0GHz, or up the overclock the i5 2500K to 4.5GHz for a fair comparison.

"All the games you play" doesn't say a lot (it could well all be high-threaded EA FPS titles)...go try 90% of the other games other there that use 4 cores or less (for example as Guild Wars 2, Assasin Creed 3, Hitman Absolution, PlanetSide 2, Hawken, Sleep Dogs etc) , the i5 2500K would leave the FX-8320 behind in the ditch if both were on the same clock.

Also, your preception of your FX-8320 is "faster" than your i5 2500K on gaming could well be under the influence "it's new and shiny it feels faster than my old i5 2500K" sympdrom. Also if the graphic cards you ain't the same, that's not like for like comparison neither.

Unless he's using a high-end card, it's likely that the 8320 will be just as good as the 2500k in many games.

Besides, he's just reporting his experience and made the appropriate disclaimer about range of titles. Your response seems a little emotional.

 

[joeyjojo]

doesnt run away with it?? the i5 wins a lot of tests by a long way, especially the gaming ones. also the i5 can be overclocked well, much more than 10%

http://www.anandtech.com/bench/Product/697?vs=701

Most of the tests are irrelevant though. The only useful ones are 1. the 2nd pass video encoding ones and maybe Cinebench multi-threaded (which indicate pure multi-threaded power) and 2. the games.

The 8350 is significantly faster in the former, and significantly slower in the latter. Horses for courses.

 

[james_2k]

yes and there are much better horses if you want to encode..

and the cinebench score for the amd is terrible. consider that it has 8 cores and my old 920 beats it by a decent margin..

and also consider that the i5 is running 400mhz slower than the amd.. silly

 

[DrBombcrater]

Em...either you should drop the overclock on the FX to 4.0GHz, or up the overclock the i5 2500K to 4.5GHz for a fair comparison.

The two chips are running at the maximum stable speed I can get out of them. The i5 can't make it 4.5GHz, the best it can do is 4.2GHz at 1.35v and even then there was the odd problem that suggested it wasn't 100% stable so I dropped it down to 4GHz. (I suspect this is more to do with the motherboard than the chip itself - there's a story behind this but suffice to say my i5 has never got beyond 4.3 on any motherboard, and my mobo couldn't push any of the three 2500Ks I've tried beyond 4.2. The board and CPU may both possibly be lemons, but I've got to work with what I've got)

The FX was a much easier overclock; it does 4.5GHz at 1.325v, stable through OCCT, Prime 95 and 6-7 hours of gaming. I can get 4.8GHz at 1.45v, but haven't been able to test that enough to have confidence in it yet.

"All the games you play" doesn't say a lot (it could well all be high-threaded EA FPS titles)...go try 90% of the other games other there that use 4 cores or less

To be specific, Eve Online, Minecraft + FTB, Supreme Commander/Forged Alliance, Sins of a Solar Empire / Rebellion, XCOM, WoT and a a few others. None, as far as I'm aware, use more than 4 cores.

Also, your preception of your FX-8320 is "faster" than your i5 2500K on gaming could well be under the influence "it's new and shiny it feels faster than my old i5 2500K" sympdrom. Also if the graphic cards you ain't the same, that's not like for like comparison neither.

I did not say the FX is faster, only that it matches the i5 in the particular games I play. Both systems are using 7950s so GPU bottlenecking should not be a factor.

I am not trying to cast doubt on the i5 here - it's a great processor that's done very well for me. Just saying that for me personally the FX unexpectedly turned out to be a better fit for the things I do and those people who dog-pile threads like this saying the FXs are uniformly inferior to Intel parts should probably pause and consider that it isn't always the case.

 

[james_2k]

those people who dog-pile threads like this saying the FXs are uniformly inferior to Intel parts should probably pause and consider that it isn't always the case.

well it is usually. 9/10 times. to be fair.

 

[Marine-RX179]

I am not trying to cast doubt on the i5 here - it's a great processor that's done very well for me. Just saying that for me personally the FX unexpectedly turned out to be a better fit for the things I do and those people who dog-pile threads like this saying the FXs are uniformly inferior to Intel parts should probably pause and consider that it isn't always the case.

But it is greatly inferior in games that use 4 cores or less, when enough GPU grunt used with it (which has been proven when with GTX690 and other multi-GPU set-ups for games in general). It's just that majority of the time it is out of sight due to people using using highest settings in game with a mid-high range card with the GPU bound pushes the average frame rate down to the point (i.e. 25-45fps) of having faster CPU that can do constant 55-60fps wouldn't matter, comparing to a slower CPU which does...say 25-50fps on average; but the difference is when it is CPU bounded, lowering graphic details won't help much with increasing frame rate.

Also, if people uses 120Hz instead of 60Hz monitor, the FX-8's weakness in the use less than 4 cores games would be VERY apparent. Not trying to be a fanboy or anything, but if I was using a FX-8 instead of my i5 for Guild Wars 2 (which only uses 4 cores and each only uses 80% max on average), I would be gutted since I got a 120Hz monitor.