All new build or drop in FX8350?

[Vic Vega]

...

 

[NTMBK]

Drop in the 8350 and don't look back. It's weaker in gaming for sure, but for running multiple VMs it will do a good job (much better than your current quad core). You have plenty of fast DDR3 and SSDs- the rest of your system is just fine, no need to upgrade it.

 

[jhu]

I have never felt the system was underpowered and I do everything on this system from gaming to day to day work as a SAN Systems Engineer, complete with multiple VMs running and lots of other "IT crap" open, to simple Web/YT/Email.

Wait until your computer can't adequately do what you want. Or at the very least, wait until Steamroller.

 

[adnank77]

This Comparison would help you decide:

http://www.anandtech.com/bench/Product/697?vs=102

As you can see, the CPU/Cores intensive tasks see big performance increase (up to double), while the less CPU intensive tasks see slight increase ..

I would vote that FX8350 is a good investment, however, if you are a primarily gamer, my vote would go for a new Graphics Card ..

 

[JWade]

to me if you like the case, and the purpose of buying a high-er end board is the ability to upgrade it in the future, but the fx8350. use the money you would spend on all new system and get bigger SSDs, maybe better video cards

 

[The Alias]

get the 8350 games better multitasks better and will most definitely make your system feel more snappy

 

[Vic Vega]

...

 

[AnonymouseUser]

The 8350 would be a nice boost, especially if overclocked.

I must say, though, I am mighty impressed to see an MSI board survive 3 years of overclocking.

 

[inf64]

Drop in the 8350 and don't look back. It's weaker in gaming for sure, but for running multiple VMs it will do a good job (much better than your current quad core). You have plenty of fast DDR3 and SSDs- the rest of your system is just fine, no need to upgrade it.

I disagree with the bold part. Here is very CPU bound gaming suite which hardware.fr used for their 8350 review:
http://www.hardware.fr/articles/880-17/moyennes.html

FX8350 scored 128.5 while 3.7Ghz Deneb 980 X4 scored 111.3pts. So 4.1Ghz Deneb would be scoring (with perfect scaling) ~123.3pts, still lower than stock FX8350 (which runs at the same average clock of 4.1GHz in games). OC the 8350 to around 4.5-4.6Ghz and 8350 increases the lead over 4.1Ghz Deneb by around ~17%. So no matter if it's stock vs stock or OC vs OC, 4 module Piledriver is a better gaming chip than Deneb

 

[PPB]

I think he says weaker compared to an Intel solution. In that case it makes sense.

If you can switch PHII to an 8320/8350 as an drop in upgrade, with no need to change motherboard, it's a no brainer.

 

[crashtech]

Even in single-thread performance, the (stock to stock) increase from the 965 to the 8350 is over 25%. Obviously, in heavily threaded loads the 8350 kills the 965. It seems like a reasonable upgrade to me.

 

[parvadomus]

Isnt the MSI 890FXA-GD70 AM3? You need an AM3+ mobo for 8350 to work.

 

[inf64]

The board does support FX CPUs, but oddly not the 8350 (at least officially). FX8320 is supported.
http://www.msi.com/product/mb/890FXA-GD70.html#/?div=CPUSupport

 

[Durvelle27]

After looking at the CPU support list. Your board doesn't support the FX 8350 but it does support the FX 8320

 

[bgt]

I have a 965 here on the same mobo, only rev 1, but the 8350 is absolutely snappier in everything over the 965. Only have no experience with the 890 chip.

 

[bgt]

The board does support FX CPUs, but oddly not the 8350 (at least officially). FX8320 is supported.
http://www.msi.com/product/mb/890FXA-GD70.html#/?div=CPUSupport

power consumption???? 125W vs 140W.

 

[krumme]

The most satisfying upgrades in my life have been the drop in cpu upgrades. Dont know but perhaps its because its cheap and fast and you keep your old gear that has become a bit like family
With your usage pattern its a no brainer imho. Even if its only a year or two.

 

[bgt]

old gear that has become a bit like family

Nicely said

 

[Vic Vega]

...

 

[Rvenger]

Newegg - FX-8320 plus use the mobile google wallet $15 code and get it for $145 shipped. Great deal. You should be able to get 4.3 - 4.5ghz with ease.

 

[Vic Vega]

...

 

[jvroig]

If you are already set on getting an FX-8350 as a drop-in replacement for your 965 @ 4.1, I would say good choice.

I have an old rig (not as old as yours, ~2 years old, although some parts like the vid card are a bit older) similar to yours, but it used to run a 1090T. Could only clock it reliably up to 3.9GHz.

Almost a month ago, was able to see an FX-8350 for very cheap, and since the board in that old rig can accept it, got the FX as a straight CPU-swap.

Even against an overclocked 1090T @ 3.9, a stock FX 8350 was no worse off in the benchmarks I ran for Windows and 2 Linux distros (CentOS 6.4, Fedora 17), a mix of single- and multi-threaded benches, slightly more ST though. I just quickly crunched some of the data I collected before and after the swap, and broken down per OS (all are 64-bit versions), the results are:

Advantage of stock FX-8350 vs  1090T@3.9
Win 7 Pro:   +6.5%
CentOS 6.4:  +0.55%
Fedora 17:   +1.58%

So even before OC'ing the FX-8350, it's already no worse off than the highest my 1090T could go. In ST scenarios, your 965 has 200Mhz more power and so will be slightly better off, but in MT scenarios it won't be a match for the 1090T @ 3.9, so most of these results will be a "worst case" for you and you can expect much better improvement even at stock.

Also, at stock, AMD's CnQ is on for the FX, as well as the Linux scaling governer, cpufreq, being set to on. It is not a problem for most of the benches, but in certain ST benchmarks like Oracle, it lowered some scores by 10-40% (this is why CentOS 6.4 only shows <1% improvement). The 1090T@3.9 has both CnQ and cpufreq turned off. Turning off cpufreq (but leavning CnQ on) would increase linux scores by a few % overall.

Here comes the complicated part. The benches, depending on what they are and on what OS, show non-uniform improvement. Some benches show big increases, some ST benches are as you'd expect (slight disadvantage for Piledriver) while some actually show a hefty increase (20%, meaning clock for clock Piledriver cores would score better than Stars).

For example, in general the PHP benches favor Piledriver cores vs Stars, but the Python benches show the opposite picture. Results, but without the WHIRLPOOL results mixed in:

Advantage of stock FX-8350 vs  1090T@3.9
PHP Hash Bench:    +9.35%   //max turbo speed at stock of the FX is 4.2, only
                            //7.7% faster than 3.9, so seems PHP loves Piledriver
                            //cores more than Stars

Python Hash Bench: -11.72%   //seems Python hates Piledriver cores
                             //or could just be the hashlib library which does.

I haven't run further tests yet on whether it is indeed only Python's hashlib that hates Piledriver cores, or the python interpreter itself.

For PHP benches that don't deal with crypto, results are even more advantageous for Piledriver:

Advantage of stock FX-8350 vs  1090T@3.9, PHP Bench, non-crypto
Windows 7:   +15.89%
CentOS 6.4:  +13.28%
Fedora 17:   +17.33%

Clearly, the PHP interpreter loves Piledriver cores more than Stars. Results above contain 50% ST benches and 50% MT benches (simply the MT-equivalent of the ST ones); the ST benches, if separated from MT, range from 10%-24% (from all OSes) better. Again, since the clock advantage of the FX at stock is only 7.7% (and we're already assuming it will run at the max Turbo freq for the duration of the ST benches), everything above that shows favoritism to the PD core vs Stars.

Another example of "it's complicated" is the WHIRLPOOL hash algorithm, which is why I removed it from the overall hash bench results I presented earlier. The hash bench (both the PHP and Python versions whose results I presented above) is composed of the following hash algos: MD5, SHA1, SHA224, SHA256, SHA384, SHA512, RIPEMD160, and WHIRLPOOL. All these algos, except WHIRLPOOL, show consistent increase in performance, as expected from their platform (Win/CentOS/Fedora). That is, the individual results don't deviate from what you would expect based on the other relevant benchmarks in the particular OS tested.

However, WHIRLPOOL, like python, shows a consistent regression on all platforms, more pronounced on Linux:

Performance regression of stock FX-8350 vs 1090T@3.9, WHIRLPOOL hash algo only:
Win 7:      -9.00%
CentOS 6.4: -24.38%
Fedora 17:  -14.18%

There's a perfectly good explanation for this. Unlike the rest of the hash algos benched, WHIRLPOOL is based on a block-cipher. And in all other crypto benches I made, block ciphers (like blowfish) suffered a significant regression from Stars->Piledriver. Clearly, if your machine is meant to be a block-cipher churning super-machine, Piledriver cores are out of the question.

There are some other interesting data points, but you get the idea. Overall, as the first set of results would say, it is definitely no worse than a 1090T@3.9. The power savings are also significant. Compared to the 1090T@3.9, the FX 8350 idle power consumption is ~40W less. Load power consumption is even better, at ~70W less.

Clearly, with overall performance no worse than an OC'd 1090T, but with significant idle and load power consumption, it is (mostly) a no-brainer as a cheap CPU-swap.

The only possible drawback is if your specific use-case suffers from Piledriver cores vs Stars, such as (from my initial data gathered) block ciphers and python.

Anyway, good luck and tell us how it goes for you after you do upgrade.

 

[Icecold]

The FX 8350 is a good upgrade for anything up to and including the highest Phenom X6. Even in single threaded benchmarks(and even more surprisingly pretty much all games I tested but YMMV) my FX 8350 beats my 1090t.(even with the 1090t overclocked to 4GHz and the FX 8350 stock). I would definitely suggest the FX 8350 or 8320 as an upgrade to a quad core Phenom II.

Phenom II was and still is a very solid chip, though, I'm typing this from one of my 1090t machines.

 

[Arkaign]

Go for the 8320! Always nice to have a drop in upgrade, and the price is wonderful. I'd just run it at stock, will match nicely with the 5850s for now, and you can look at 7950OC 3GB as a potential upgrade down the line. Power efficiency and heat go a little over the top with oc, but the stock performance is overall very respectable.

 

[oceanside]

Wait until your computer can't adequately do what you want. Or at the very least, wait until Steamroller.

Sage advise... currently I'm draining every last iota of life from this P4 socket 478 AGP system before upgrading. I use it daily, and unfortunately, it just won't DIE! D: