Observations with an FX-8350

[JimPhreak]

real-time 1080P transcoding for DLNA clients, IPTV, file server, Firewall ( sidewinder/MFE) , SSL VPN Termination (F5 APM), Cisco IOU/Dynamips/Olive Lab . the price difference between a 3770 ( K doesn't have IOMMU) and 8350 bought a Samsung 840 SSD that all my guests OS's run on.

Clearly for your needs you found the right chip. I'd be willing to bet the majority of those looking at desktop CPU's aren't doing half of what you are, at least not with any kind of regularity.

 

[itsmydamnation]

Clearly for your needs you found the right chip. I'd be willing to bet the majority of those looking at desktop CPU's aren't doing half of what you are, at least not with any kind of regularity.

yes and that why i had a go at his over generalization, when you generalize that far you give everyone the right of reply , so i did. These are technical forums vague generic statements are of little value. Also as i said my main rig is an I7 (old 920-C0), built my brother a 3770 3 days ago, so im hardly a buy AMD and then convince myself my workloads are optimized for it as stated.

 

[JimPhreak]

yes and that why i had a go at his over generalization, when you generalize that far you give everyone the right of reply , so i did. These are technical forums vague generic statements are of little value. Also as i said my main rig is an I7 (old 920-C0), built my brother a 3770 3 days ago, so im hardly a buy AMD and then convince myself my workloads are optimized for it as stated.

Touche :thumbsup:.

 

[2is]

yes and that why i had a go at his over generalization, when you generalize that far you give everyone the right of reply , so i did. These are technical forums vague generic statements are of little value. Also as i said my main rig is an I7 (old 920-C0), built my brother a 3770 3 days ago, so im hardly a buy AMD and then convince myself my workloads are optimized for it as stated.

Your main system being an i7 you fit into the "most people" generalization.

 

[bgt]

Ofcourse the 3770 is a fast CPU but my experience with the 2500K is just........soso.
This is just my experience. Benchmarks mean little if your overall feeling is different. With what I do the FX experience is a great 1. Nothing wrong with the 2500K. When Intels next chip is coming out......I will buy it to try and use. I am not biased, I am too old for that, and I always like to see beyond the benches. I found them to be a guide to what to expect from a chip. And power usage..............I change my main system every half year=no issue.
BTW best wishes to you all.

 

[SPBHM]

Ofcourse the 3770 is a fast CPU but my experience with the 2500K is just........soso.
This is just my experience. Benchmarks mean little if your overall feeling is different. With what I do the FX experience is a great 1. Nothing wrong with the 2500K. When Intels next chip is coming out......I will buy it to try and use. I am not biased, I am too old for that, and I always like to see beyond the benches. I found them to be a guide to what to expect from a chip. And power usage..............I change my main system every half year=no issue.
BTW best wishes to you all.

feeling?
I see that you use the 2500K with some Intel H67 board... considering how huge OC was for this CPU, you are really limited... you were the one complaining about how hot (a stock 2500k?) the 2500K was and you don't care about power usage?

if you are only doing some heavy rendering work then sure, the 8350 is probably clearly faster...

but if you consider that the 2500k can probably be overclocked further and that this is with something like 200MHz less
http://www.anandtech.com/bench/Product/698?vs=288

I still don't understand how you can call it "slow", sure if you use mainly involve some of the best cases for the FX, and you are not affected by the higher power usage and you are comparing 3.3 vs 4GHz, it's clearly better,

 

[inf64]

feeling?
I see that you use the 2500K with some Intel H67 board... considering how huge OC was for this CPU, you are really limited... you were the one complaining about how hot (a stock 2500k?) the 2500K was and you don't care about power usage?

if you are only doing some heavy rendering work then sure, the 8350 is probably clearly faster...

but if you consider that the 2500k can probably be overclocked further and that this is with something like 200MHz less
http://www.anandtech.com/bench/Product/698?vs=288

I still don't understand how you can call it "slow", sure if you use mainly involve some of the best cases for the FX, and you are not affected by the higher power usage and you are comparing 3.3 vs 4GHz, it's clearly better,

I don't get what you are trying to do here? The guy already bought his new system and he is happy with it. He certainly won't go back to 2500K now so your posts on this topic are hardly relevant. And your link to AT bench shows 2500K vs 8320,not 2500K vs 8350 . I'd say FX8350 is overall much better product. Sure it cannot provide the fps that 2500K does but it's way more than enough and it trounces it in other workloads(and the trend is that MTed apps are becoming more prevalent as time goes by).
Note that 8350 he runs uses stock clock so he is not OCing or racking up power/heat.

 

[SPBHM]

I don't get what you are trying to do here? The guy already bought his new system and he is happy with it. He certainly won't go back to 2500K now so your posts on this topic are hardly relevant. And your link to AT bench shows 2500K vs 8320,not 2500K vs 8350 . I'd say FX8350 is overall much better product. Sure it cannot provide the fps that 2500K does but it's way more than enough and it trounces it in other workloads(and the trend is that MTed apps are becoming more prevalent as time goes by).
Note that 8350 he runs uses stock clock so he is not OCing or racking up power/heat.

how relevant is your post, how relevant is his post calling the 2500k slow and complaining about the temperature? (without showing anything to support)
he posted a result of his own tests showing the stock 8350 using a lot more power (http://forums.anandtech.com/showpost.php?p=34431958&postcount=281)

if you read my post you will understand why I posted the link, I tried to find any test of both CPUs overclocked and found nothing, so I tried to post the closest clock possible comparison I could find, if you look carefully the stock 2500 does well in some of the most popular MT softwares, in others not really

I'm just answering to what I believe is not a fair comparison between both products, and trying to understand why,

 

[inf64]

He said he does not care about power. FX draws more power but performs better,stock vs stock. In some MT software way better than 2500K. Sure 2500K wins in some poorly threaded stuff(such as games) but people don't just game and even if he does game 8350 is just a fine gaming CPU-just not as good as SB/IB though.

Note that user had both CPUs. Do you own a FX to compare it with intel part you own? If not then his comparison is more valid since he had access to both,sorry.

 

[Homeles]

It's not valid at all. It's an anecdote.

 

[SPBHM]

He said he does not care about power. FX draws more power but performs better,stock vs stock. In some MT software way better than 2500K. Sure 2500K wins in some poorly threaded stuff(such as games) but people don't just game and even if he does game 8350 is just a fine gaming CPU-just not as good as SB/IB though.

Note that user had both CPUs. Do you own a FX to compare it with intel part you own? If not then his comparison is more valid since he had access to both,sorry.

he have both (but unfortunately that Intel h67 board is a big limitation for the K CPU), so that's what I wanted to see, something to support his opinion that the 8350 was a great upgrade, but the only thing he posted was the 8350 using 90w more!
he don't care about power but complains about temperature? both are related...

on the real world not everything (I would say the most typical usage scenario uses quite poorly) can use amazingly well 8 threads, that's why (combined with the higher IPC) the 2500 even at a lower clock and with 4 c/t does so well in so many applications (even in things like winrar, which actually uses well 4+ cores) as you can see on the tests, and that's why I don't think calling the 2500 a bit slow and seeing the 8350 doing better on basic stuff doesn't make sense.

having access to something doesn't mean you are necessarily doing a fair comparison, sorry.

 

[SlowSpyder]

I liked this thread more before it became AMD vs. Intel. One person says AMD is better for him and the calvary comes in... let's continue to see how this chip does for IDC, good or bad.

 

[Rvenger]

He said he does not care about power. FX draws more power but performs better,stock vs stock. In some MT software way better than 2500K. Sure 2500K wins in some poorly threaded stuff(such as games) but people don't just game and even if he does game 8350 is just a fine gaming CPU-just not as good as SB/IB though.

Note that user had both CPUs. Do you own a FX to compare it with intel part you own? If not then his comparison is more valid since he had access to both,sorry.

If haters would actually own an FX cpu they would find out they aren't crap and they aren't slow either.

 

[The Alias]

on the real world not everything (I would say the most typical usage scenario uses quite poorly) can use amazingly well 8 threads, that's why (combined with the higher IPC) the 2500 even at a lower clock and with 4 c/t does so well in so many applications (even in things like winrar, which actually uses well 4+ cores) as you can see on the tests, and that's why I don't think calling the 2500 a bit slow and seeing the 8350 doing better on basic stuff doesn't make sense.

having access to something doesn't mean you are necessarily doing a fair comparison, sorry.

you never know he may be a us like me that pulls up several gaming sites while playing games to check stats, or uses his pc in a way like that . You can't nullify someone's opinion based on actual usage especially when that person has experience in something you don't .

 

[Abwx]

If haters would actually own an FX cpu they would find out they aren't crap and they aren't slow either.

Viral marketing did take its toll with the "single thread perf is all" mantra.

As pointed by Inf64 softs are more and more multithreaded and people
dont realize at wich point a 8350 has so much more processing power
than a i5.

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/10

 

[Homeles]

If haters would actually own an FX cpu they would find out they aren't crap and they aren't slow either.

Us "haters" are those of us who rely on the opinions of third party reviewers that back their findings with reliable and accurate benchmarks.

The people who don't see it as crap are the fools that believe that subjective testing is a legitimate way to differentiate between two products.

Viral marketing did take its toll with the "single thread perf is all" mantra.

As pointed by Inf64 softs are more and more multithreaded and people
dont realize at wich point a 8350 has so much more processing power
than a i5.

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/10

http://en.wikipedia.org/wiki/Cherry_picking_(fallacy)

 

[grimpr]

Wheres the cherry picking? Piledrivers architecture was clearly designed for MT apps in mind and it succeded there, the FX-8350 is a stronger chip than Intels quadcores, it stays in Core i7 class in the most heavy multithreaded apps and usage, case closed. Lets move on.

 

[BrightCandle]

There is the real potential that the software tests done by review sites don't represent what people do with their computers. However so far it seems benchmarks are far more likely to utilise many threads than the average piece of software.

Multithreading remains difficult, not a great deal has changed in the way we design software. The last couple of years has seen a decent growth in functional programming which should deliver a lot of parallel execution gains, but at the cost of some efficiency. But the languages that make this easy and utilise multiple threads are still far from mainstream.

 

[TheRyuu]

It's the x87 instruction set IIRC and AMD took out native support for space reasons as x87 is pretty much irrelevant today.

I'm not so sure this is true (as others I think have pointed out already).

The FPU throughput should remain unchanged from previous generations. The only reason it would perform slower than a 6-core phenom is because the phenom can execute 50% more FP instructions (since the FPU is shared in a module on bulldozer).

And even using SSE2 for floating point math doesn't make bulldozer any faster.

My 2500K would not overclock to 4.5Ghz. It was too hot also. Had to buy a big cooler to cool it. Still got over 65C when fully loaded. Wanted to buy an i7 3770 but HT is a weak point of Intel. I have a 3225, 2100,2100T here for use in HTPC's but I find HT to be very chaotic in use. Sometimes it really acts like a dual cpu and in another instance it is a quad core. So for me a true core is better.

I would think that SMT (HyperThreading) is a strong point for intel cpu's. It allows for better resource usage for heavily threaded applications. There should be, near as makes no difference, zero performance impact on stuff that doesn't thread so well. SMT gives you gains of up to ~25% for threaded applications, and it should be anything but chaotic.

There is the real potential that the software tests done by review sites don't represent what people do with their computers. However so far it seems benchmarks are far more likely to utilise many threads than the average piece of software.

It all depends on what you use. Certain benchmarks like the x264 benchmark exactly reflect the numbers you would get if you ran x264 (they do run 2-pass encodes though, unless you absolutely need to hit a specific file size you should be using the crf encoding mode (which is 1-pass) but that's beside the point here I suppose). Things like perhaps a dedicated streaming box would probably do well with AMD in this case.

That being said certain benchmarks are probably useless, it's just provides one more way to compare CPU vs CPU with all else being equal (e.g. the same benchmark). Even if it has no real bearing on the real world it does allow you to take known values for one platform and compare it against another for that specific benchmark which can allow you to get a general idea of the performance characteristics of a given CPU.

 

[Xpage]

You mean those synthetic memory benchmarks aren't uber important for my day to day life?


I think a FX series or intel Sandy/Ivy bridge would be fine for most any person. My father uses a E-350 for the web and he's happy with it. Most people do not care about the differences we are all talking about as most users are ignorant of most things inside their computer.

We are the 1-2% of people who care about relatively small differences between processors. As long as the processor is faster than something 4-5 years ago people will be happy. But then again user experience is due to the OS (see vista nightmare), and most users will be happier with a SSD and a fresh windows install in their old computer than a new processor.

 

[bgt]

He guys, please don't make this FX8350 vs 2500K!! I have them both and each has it advantages. And I never said the 2500K becomes hot in normal mode....only when you overclock it. And I now changed the MB to an Intel board, had an Asus P8P67-M before.
Please read my experience as it is, it is not a statement against the 2500K. The FX really performs well. The 2500K also but on another level.

 

[SPBHM]

you never know he may be a us like me that pulls up several gaming sites while playing games to check stats, or uses his pc in a way like that . You can't nullify someone's opinion based on actual usage especially when that person has experience in something you don't .

that's why I tried to ask him more than once to explain how he concluded the FX 8350 was great and 2500k not really, I don't really agree with his argument that the 2500k is to hot and things like that, also the dislike for HT makes no sense,

there are many usage scenarios where the 2500k is clearly the best, and others in which the FX will be clearly faster (unfortunately using a lot more power), but I think, for the vast majority of PC users the qualities of the i5 have a larger appeal, the way he put his opinion it seems the FX is by far a superior product, when there are clear shortcomings,

 

[guskline]

I have 2 2500k machines, a Bulldozer 8150 machine and a PileDriver 8350 mchine. Clearly the 2500k rigs are the best gamers and held a decent edge over the 8150. The 8350? Not as big of an edge if at all.

 

[parvadomus]

A real life example where the FX8350 loses out due to abit of serial code in the multithreaded application is chess.

As you can see the application scales with cores, even 16 threads as shown here.

I got your point, but that example is terrible, Fritz chess doesnt have that limitation (if you watch carefully this benchmark scales almost linearly to core count).
The problem with fritz chess is that its a "branch predictor killer". No CPU can predict branches correctly in this almost random branch picker benchmark. However, Intel CPUs, thanks to a shorter pipeline can recover much much faster, and thus have a very high IPC compared to BD for this particular benchmark.

 

[guskline]

BTW, my 8350 @4.6 Ghz score 14,524 in Fritz chess.