AMD FX-8350 Review: Does Piledriver Fix Bulldozer's Flaws?

[proffet]

I see that DDR3-1600 is still good enough..

 

[blazorthon]

were not talking about the FX-4170 (I can destroy that argument later..)
and the 8-core FX-8350 doesn't exactly win by large margins in half of the bench suites that utilize all cores.
(I expected more..)

and don't give me that disable 1/2 the module nonsense either..
in-gaming for better performance.

You should have mentioned the FX-4170 in that context and complaining about an eight-core CPU's performance in lightly threaded games was not helping you point.

In the workloads that fully taxed all eight cores as much as reasonably possible, it did very well. Whether or not it won by huge margins isn't the point, how well it did is the point. It doesn't need to win by huge margins. Heck, it didn't even need to win. All that it needed to do was beat the FX-8150 in power consumption and performance and in that, it has succeeded ,although not as much as I expected it to. Like I said before, the L3 cache seems to have too little performance to make a difference because even the A10-5800K can game almost as well as this CPU and that's with no L3 and a lower CPU frequency and the same CPU micro-architecture. If anything, either disabling the L3 (if possible) or overclocking it (this is easy) might be important for Vishera to really distinguish itself from Zambezi.

EDIT: Disabling one core per module lowers power consumption while increasing lightly threaded performance, albeit at the sacrifice of highly threaded performance (no big deal for a lightly threaded workload anyway). That's not nonsense.

EDIT again: You can't destroy my FX-4170 argument, but you can go ahead and try.

 

[proffet]

A10-5800K w/ HD 6670 will be a great unit for my son..
(can it X-Fire with a higher card.?)

passing up on Piledriver.

 

[proffet]

EDIT: Disabling one core per module lowers power consumption while increasing lightly threaded performance, albeit at the sacrifice of highly threaded performance (no big deal for a lightly threaded workload anyway). That's not nonsense.

EDIT again: You can't destroy my FX-4170 argument, but you can go ahead and try.

I got to pick on you...
the FX-4170 is crap too.

 

[Teslarifle]

The only benchmark I ever check on these is Skyrim, since I am big player of Bethesda RPGs. What I take most from this review is that the i5-2500k is still overwhelmingly a much better value for exceptional gaming than anything else available, especially if you can find it for sale under $170. The Phenom 955BE is also still a great value for adequate performance if you can find it for $70 or less.

 

[davemaster84]

AMD Has done a really good work; despite the Intel's favoritism the reviwer seem to have, I think the 8350 is an exc ellent chip for an enthusiastic builder. It's finally an alternative between the 3570 and the 3770k we amd followers have been waiting for

 

[blazorthon]

A10-5800K w/ HD 6670 will be a great unit for my son..
(can it X-Fire with a higher card.?)

passing up on Piledriver.

The A10-5800K uses Piledriver and no, it can't CF with anything higher than a 6670, at least not according to AMD. I've never tried it and I don't know of anyone successfully trying it.

 

[blazorthon]

I got to pick on you...
the FX-4170 is crap too.

Your opinion of it is completely irrelevant as is mine. It performs right about where it should in gaming for its price, although its power consumption is obviously very high. It also performs similarly to this quad module 8350 in most games and that's what makes it relevant to the conversation, but not our opinions of it.

 

[proffet]

fair enough...
(good talk.. )

 

[gondor]

It’ll be interesting to see how AMD makes use of it in future architectures, particularly as the same paper shows a gradual reduction of clock efficiency after a peak at 3.3 GHz.

Any electronic resonant system is usually only resonant at a certain (single) frequency. The further you deviate from it, the worse it performs. This is not something specific to AMD. As a consequence of that, the chips they make exhibit largest savings from resonant clock mesh at a certain frequency. All chips are produced the same, with parts binned after the fact.

I'm sure AMD can retarget the rclk mesh for higher frequency with next generation product (Kaveri with Steamroller), should they expect an increase in average operation frequency from that of Trinity/Piledriver.

 

[Guest]

It's basically an 8150 + 400mhz. There's been no / to little improvement like I predicted. This is a very disappointing offering from AMD. Would not buy.

 

[davemaster84]

It's basically an 8150 + 400mhz. There's been no / to little improvement like I predicted. This is a very disappointing offering from AMD. Would not buy.

I thikn you're wrong, it's not only 400mhz more, it's lower power consumption, more overclocking radio and most of all, around 15% core to core improvement. If you compare an 8150 at 4.2 ghz to an 8350 at the same speed the 8350 will get better result in overall. Only in gaming the difference seems to be not so big, but it's a very good, inexpensive alternative.

 

[manicmike]

What I don't understand is why AMD didn't add PCIe 3.0 support to this one... It would have helped in benchmarks like Skyrim & WoW:MoP.

I'm glad to see a definite improvement over the BD line, but I think for the time being I'll stick with my FX-6100, it does everything I need it to and then some. It's no high-end gamer-super-enthusiast-wtfHAXXX rig, but it plays my games acceptably for the time being.

It will be interesting to see if time has an effect on these processors as it did in the article Better With Time? The A8-3870 and the Pentium G630, One Year Later: Quo Vidas, Llano? A Look Back and Ahead (Which aactually inspired me to switch from the G630 to the A8-3870 for my kid's Christmas presents this year)

@ Chris Angelini - Don't suppose you'd be willing to retry a few (not necessarily all) benchmarks on Windows 8 when it drops would you? From what I've been reading, it's more lightweight and sees performance gains for both intel and AMD, but the latter's gains are supposed to be greater. I don't think by a long shot that it will somehow catapult AMD into the lead, but I'm curious as to how it would effect each score (or perhaps do a side by side)

 

[dscudella]

I know Newegg is your official pricing partner, but I was able to pick up a 3570K at a local micro center for $190 this summer. The FX8350 should be priced around $180 IMP. Yup just went to their website and they're still showing the 3570K for $189. So I think the Newegg price is suspicious.

You can't count on Microcenters pricing because 80% of the population does not live by one. Everyone can shop on newegg.

 

[murambi]

I just dont understand some of the reasoning with TH writers. why spend money on software that costs above 2k $ and then later spend 200$ on a processor to accomplish the work. I work alot with animation and beleive you me single threaded performance is very important. Most motion works artist that I know of use i7 processors and also the extreme editions. So to pitch Vishera as a multithreaded viable option just feels abit wrong

 

[cangelini]

[citation][nom]manicmike[/nom]What I don't understand is why AMD didn't add PCIe 3.0 support to this one... It would have helped in benchmarks like Skyrim & WoW:MoP. I'm glad to see a definite improvement over the BD line, but I think for the time being I'll stick with my FX-6100, it does everything I need it to and then some. It's no high-end gamer-super-enthusiast-wtfHAXXX rig, but it plays my games acceptably for the time being. It will be interesting to see if time has an effect on these processors as it did in the article Better With Time? The A8-3870 and the Pentium G630, One Year Later: Quo Vidas, Llano? A Look Back and Ahead (Which aactually inspired me to switch from the G630 to the A8-3870 for my kid's Christmas presents this year)@ Chris Angelini - Don't suppose you'd be willing to retry a few (not necessarily all) benchmarks on Windows 8 when it drops would you? From what I've been reading, it's more lightweight and sees performance gains for both intel and AMD, but the latter's gains are supposed to be greater. I don't think by a long shot that it will somehow catapult AMD into the lead, but I'm curious as to how it would effect each score (or perhaps do a side by side)[/citation]
We have a bunch of coverage planned for the Win8 launch, including FX-8150 in 7 and 8. I was planning on doing some tests with FX-8350 as well, though perhaps a little later.

 

[TeraMedia]

I still can't get past the fact that the Intel IB L3 cache latency is 19 clock cycles, while the 8350 is 76... four times as long at same clock frequency. And for L2, it's 12 cycles vs. 20. L3 might not be *that* important, but L2 sure is. And the difference in single-threaded performance (1.6:1.1 = 1.45) roughly mirrors the difference in clock cycle latency at L2 (20:12 = 1.67). Fix the L2 cache latency, and does that fix the single-threaded performance? If I were AMD, I would be putting a lot of R&D into improving cache latency right now. The cache serves as the framework for your CPU's performance. If it's slow, nothing is going to help the CPU go faster.

 

[blazorthon]

What I don't understand is why AMD didn't add PCIe 3.0 support to this one... It would have helped in benchmarks like Skyrim & WoW:MoP.

I'm glad to see a definite improvement over the BD line, but I think for the time being I'll stick with my FX-6100, it does everything I need it to and then some. It's no high-end gamer-super-enthusiast-wtfHAXXX rig, but it plays my games acceptably for the time being.

It will be interesting to see if time has an effect on these processors as it did in the article Better With Time? The A8-3870 and the Pentium G630, One Year Later: Quo Vidas, Llano? A Look Back and Ahead (Which aactually inspired me to switch from the G630 to the A8-3870 for my kid's Christmas presents this year)

@ Chris Angelini - Don't suppose you'd be willing to retry a few (not necessarily all) benchmarks on Windows 8 when it drops would you? From what I've been reading, it's more lightweight and sees performance gains for both intel and AMD, but the latter's gains are supposed to be greater. I don't think by a long shot that it will somehow catapult AMD into the lead, but I'm curious as to how it would effect each score (or perhaps do a side by side)

PCIe 3.0 most certainly would not have made much of a difference and it would require new chipsets and motherboards, so it wouldn't have mattered if AMD added it because then you'd need new motherboards with support for it to test it. Besides, the CPU doesn't decide what type of PCIe is supported for the AM3+ socket, that's all on the chipset, so maybe AMD will introduce PCIe 3.0 with a new chipset, although probably not.

 

[blazorthon]

I just dont understand some of the reasoning with TH writers. why spend money on software that costs above 2k $ and then later spend 200$ on a processor to accomplish the work. I work alot with animation and beleive you me single threaded performance is very important. Most motion works artist that I know of use i7 processors and also the extreme editions. So to pitch Vishera as a multithreaded viable option just feels abit wrong

Why mention i7s, let alone Extreme Editions, if you want to talk about single threaded performance? Heck, i3s and i5s can match them in that. Furthermore, most, modern professional software most certainly can be highly-threaded, so of course that plays a huge role here.

 

[maxinexus]

AMD learned from 8150 mistake. I'm glad it is priced at $199 and not $259. If they would have priced Zambezi 8150 at 199 as well it would not get that much scrutiny. Good comeback but we need old FX

 

[burmese_dude]

Gotta root for AMD to be competitive. Otherwise, it'll all be Intel.

Just look at wrestling, used to be entertaining and fun. It was the best when WCW competed against WWF. Now that it's WWE and there's no competition, it sucks like hell.

 

[sugetsu]

When it to productivity and content creation is the adobe CS6 master collection better suited for Intel or AMD if the user is on an $800 budget? With $800 I can't afford to go with an i7; my choices are i5 or the new FX. Remember that I have to go with an NVIDIA card for cuda acceleration and the best thing I can find for this limited budget is a GeForce 9800 GT HDMI 1GB 256-bit DDR3 for $60.

Any suggestions?

 

[andrewcarr]

Theirs still one big reason I wouldn't get the AMD processor. The gaming performance is still vastly sub par and this alson will keep me from even considering this CPU for my next build.

 

[razor512]

The main issue with AMD now is their focus on multithreading rather than single core performance. Compare the speed of the FX chips to the Phenom II, you will see that you have more cors but nearly the same or sometimes slower performance, this means you are taking more cores to get the same amount of work done. the problem with this is that applications that cant use many cores will lose performance.

I currently have a AM3+ board, and overclock, and I am sure many people on Phenom II systems are in the same boat.

While the FX chips have higher clock speeds and can also be overclocked, the IPS is lower to there is less of a benefit to to having a 600-800MHz overclock on the FX chip as compared to the Phenom II

that is why I want to see the benchmarks done with both the FX chip and the Phenom II overclocked (most of the Phenom II x6 chips can hit 4GHz with the stock heatsink, and 4.2GHz with a more powerful cooler such as the H80), and many articles are showing around 4.8GHz for the FX chip (and upwards of 5-5.1GHz with liquid cooling)

Since you are talking to a audience who are likely to have their CPU's overclocked, a good piece of data to have is, will an overclocked FX beat my overclocked Phenom II, and if to, will it beat it by enough to be worth investing in a upgrade.


Another thing to test is the shared core performance loss. Since the FX chips are essentially 4 hyper threaded cores, the load on 1 core will have a significant impact on the performance of the other core next ot it, for example have cinebench do a single threaded run and set the affinity to the first core, then using another application, (eg prime 95), have it also only do 1 core then repeat the test, each time changing which core prime 95 uses, you will see that depending on which core it uses, it prime 95 uses a core on the same core module, it will lower the cinebench score a lot more than if it uses a core that is not on the same core module. (A phenom II CPU will not have this issue)

PS a core I7 has this issue also, but a core I5 and other non hyper threaded intel chips do not have this issue. The main difference is that intel tells you that it is a 4 core CPU running 8 threads instead of lying and saying 8 cores running 8 threads.

Because of the bahavior of the current FX chips, I don't see how they can call it 8 cores.


Can you call a building 2 houses if both share the same living room?

 

[murambi]

Why mention i7s, let alone Extreme Editions, if you want to talk about single threaded performance? Heck, i3s and i5s can match them in that. Furthermore, most, modern professional software most certainly can be highly-threaded, so of course that plays a huge role here.

its like buying a lorry without an engine and placing a 4 * 4 engine to run it. sure it will move but........ you get my point?