FX-8370 - non E version few tests

[SPBHM]

Nice to post a single game at PClab.

Btw, next time try to not use deffamation/ad hominem as a mean to denigrate my saying by branding me an "AMD fan".

Hardware.fr updated their game suite, the FX8350 scores increased by 12.5% putting it at barely 10% of the 2500K in games, in applications we wont even discuss this point since this is now a massive win for the FX.

They still have not updated their old charts, that s why i posted their recent reviews using updated games, here the old scores, increase the FX score by 12.5% and do the conclusions you want knowing that this is the average of all their games and not a single game at x resolutions like your "exemple".

http://www.hardware.fr/articles/905-5/performances-jeux-3d.html

Now you can always pretend that games are going the lowly threaded route on the mid term or that the FX is already fully used in said games, in appllications the 8350 was already 25% better, and that s with unupdated softs, you think that it wont show in future games iterations.?..

if you are not an AMD fan I'm sorry, but that's something that transpires from posts, but I was mainly talking about this hopeful idea of games becoming good for the FX in the future I hear AMD fans talking about since 2011, and by now people forgot about the 8150, while the 2500K is relevant enough to beat the successor at 5GHz in many games;

games are becoming more threaded, but dependency on high per thread performance is not disappearing, games are all different from each other, both new games I posted (different sources) provide data against the generalization based on the hardware.fr test (which doesn't even include the 2500K), and as I pointed before, 60fps on watchdogs CPU test? your are doing it wrong hardware.fr

also they are not including AMD VGAs for some reason, when most people going for value are going to consider AMD VGAs, and tests have shown worse relative performance for AMD CPUs with AMD VGAs.

They are talking about Core i5 2500K, not Core i5 Haswell.


Also, thats something you dont see every day

http://gamegpu.ru/retro-test-gpu/crysis-2007-retro-test-gpu.html

they seem to have used the CPU benchmark scene included with the game, it's very extreme on the amount of things going on (far worse than normal game), but it's interesting I guess, because Crysis 1 is famous for not scaling with more than 2 cores, and I remember the Core 2 Duos doing a lot better than the AMD CPUs at the time? to bad they don't have anything older than SB (apart from the game)

 

[Abwx]

if you are not an AMD fan I'm sorry, but that's something that transpires from posts

I have gear from all brands but it is a fact that i prefer firms which have the most ethical records and in this respect only AMD suit this description, Intel or Nvidia, wich i both supported years ago, no more fits this requirement in my opinion.

but I was mainly talking about this hopeful idea of games becoming good for the FX in the future I hear AMD fans talking about since 2011, and by now people forgot about the 8150, while the 2500K is relevant enough to beat the successor at 5GHz in many games;

games are becoming more threaded, but dependency on high per thread performance is not disappearing, games are all different from each other, both new games I posted (different sources) provide data against the generalization based on the hardware.fr test (which doesn't even include the 2500K), and as I pointed before, 60fps on watchdogs CPU test? your are doing it wrong hardware.fr

they seem to have used the CPU benchmark scene included with the game, it's very extreme on the amount of things going on (far worse than normal game), but it's interesting I guess, because Crysis 1 is famous for not scaling with more than 2 cores, and I remember the Core 2 Duos doing a lot better than the AMD CPUs at the time? to bad they don't have anything older than SB (apart from the game)

Hardware.fr has its own wrongs and bias but all their infos are not to be rejected because of some lacks, as for the 8150 check its perfs before jumping to speedy conclusions, if you want we can compare the FX Vishera with IB, we can even have all this people in a single recent review :

http://www.computerbase.de/2014-09/amd-fx-8370e-im-test/2/#diagramm-performancerating-anwendungen

also they are not including AMD VGAs for some reason, when most people going for value are going to consider AMD VGAs, and tests have shown worse relative performance for AMD CPUs with AMD VGAs.

That was before Mantel was released but still , check computerbase.de review i linked, you ll have a good idea of where the 2500K stand currently, what transpires, as you said, from the posts i read here and there is that some people are either living with past but no more relevant impressions or are kept arguing using urban legends, i noticed that people insist on games as if it was the only usage of a PC, i guess that it s convenient to not talk of all possibles applications where the 2500K is largely outmatched by..a 8150.

 

[crashtech]

Visheras and Sandy Bridges are both blasts from the past, imo, but the defining characteristic of the 2500K, especially OCed, is that its per thread performance is pretty tough to beat. A full-tilt 9590 doesn't even start to eclipse the performance of a 2500K until about 6 fully loaded threads or higher.

 

[Abwx]

A full-tilt 9590 doesn't even start to eclipse the performance of a 2500K until about 6 fully loaded threads or higher.

Yet another urban legend....

Check Hfr MinGW-w64-GCC4.7.1 compilation test, according to the reviewer the soft use 6 threads but all the work is done on 4 cores with the two other cores being marginaly used.

So in essence you are saying that a 2500K has better perfs than a 4670K.

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

Edit : on an integer task like winrar a 2500k is 61% of a 9590.

http://www.computerbase.de/2014-09/amd-fx-8370e-im-test/2/#diagramm-winrar-integriert

You would need to overclock this 2500K to 5.4 to get the 9590 score assuming 100% scaling.

 

[crashtech]

Yet another urban legend....

Check Hfr MinGW-w64-GCC4.7.1 compilation test, according to the reviewer the soft use 6 threads but all the work is done on 4 cores with the two other cores being marginaly used.

So in essence you are saying that a 2500K has better perfs than a 4670K.

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

I am speaking in the general case, but you want to point out specific outliers to try and prove your point. That's fine, clearly you have a self-professed personal stake in this matter via what you perceive as "ethics." But if someone with no emotional attachment looks at all the available data, the picture becomes pretty clear that 8T Visheras only confer benefits in highly threaded loads. The specifics of this generality can be up for grabs depending on the software used, but it's not a controversial view to those who see clearly.

The problem with GCC is that while it might be perceived as more "fair," the fact is that same code will run faster on virtually all CPUs when compiled with ICC. Also, what a terribly limited way to be if we have to base our software purchasing decisions on what compiler was used.

 

[Abwx]

Your answer is basically that i use specifical cases, you answered before i posted computerbase link, check there, we are no more talking of GCC if this soft doesnt suit you, there are other exemples but an integer test is really good to estimate the perfs in fully Mthreaded games.

And why using a personal argument and talking about my mention of ethics wich was a response to SPBHM specific question that was somewhat OT but to wich i still answered to.?.

Does it have something to do with the numbers i posted.?.

What seems to me obvious is that people are somewhat confusing SB, IB and Haswell perfs as if the progress made in the last generation hold for the previous ones, of course i dont think that people are that stupid but simply that they do not bother doing researches and checking the numbers before doing bold claims that are easily demonstrated as being purely psychologicaly based.

 

[crashtech]

Also, when one peruses the Cinebench 11.5 results, we see a representative 2500K @4.6GHz scoring 7.2, and an 8350 @5.09GHz scoring 8.7. So 8350 wins, even in ICC compiled software. I'll leave it to readers to extrapolate from these results and draw their own conclusions on the relative per-core performance of these CPUs.

 

[crashtech]

And why using a personal argument and talking about my mention of ethics wich was a response to SPBHM specific question that was somewhat OT but to wich i still answered to.?.

I believe admitted bias is relevant. But I won't mention it anymore if you will refrain from passing ethical judgements on companies and then letting that judgement, which has nothing to do with absolute CPU performance, color your objectivity.

 

[SlowSpyder]

Also, when one peruses the Cinebench 11.5 results, we see a representative 2500K @4.6GHz scoring 7.2, and an 8350 @5.09GHz scoring 8.7. So 8350 wins, even in ICC compiled software. I'll leave it to readers to extrapolate from these results and draw their own conclusions on the relative per-core performance of these CPUs.

According to Cinebench the FX 8xxx and 9xxx's are four core CPU's.

 

[crashtech]

According to Cinebench the FX is a four core CPU.

Yeah, that gets a bit fuzzy, doesn't it. I kind of think of it as a 4C/8T unit myself, but most people that have them think they are octacores. What do you think?

 

[SlowSpyder]

Yeah, that gets a bit fuzzy, doesn't it. I kind of think of it as a 4C/8T unit myself, but most people that have them think they are octacores. What do you think?

I'm not sure how to classify it. I tend to go by module count. I think you can look at an AMD module as a 'core+'. But then again the module can't use all it's resources for single threaded applications, so maybe that isn't accurate either. I guess I'd say an AMD module is closer to two cores than a HyperThreaded Intel core, but still falls short of being two distinct cores, at least two distinct full cores.

 

[Abwx]

Also, when one peruses the Cinebench 11.5 results, we see a representative 2500K @4.6GHz scoring 7.2, and an 8350 @5.09GHz scoring 8.7. So 8350 wins, even in ICC compiled software. I'll leave it to readers to extrapolate from these results and draw their own conclusions on the relative per-core performance of these CPUs.

I guess that you did notice that the FXs do much better in integer related tasks than in FP intensive applications since it was designed as a server CPU mainly, but there s FP tasks like Povray where the FX8350 at stock manage to outmatch the 2500K by 60%, and to match the 4770K, i guess that looking at povray is quite fair since we can average this score with Cinebench which favour Intel CPUs, and dont tell me that it s an urban legend of mine, i wouldnt do such a statement if i had not enough elements but to make sure people dont conclude that ii m inventing check here, Maxon use much more from Intel than simply ICC :

http://www.youtube.com/watch?v=PzLxCo5qofo

 

[crashtech]

I guess that you did notice that the FXs do much better in integer related tasks than in FP intensive applications since it was designed as a server CPU mainly, but there s FP tasks like Povray where the FX8350 at stock manage to outmatch the 2500K by 60%, and to match the 4770K, i guess that looking at povray is quite fair since we can average this score with Cinebench which favour Intel CPUs, and dont tell me that it s an urban legend of mine, i wouldnt do such a statement if i had not enough elements but to make sure people dont conclude that ii m inventing check here, Maxon use much more from Intel than simply ICC :

http://www.youtube.com/watch?v=PzLxCo5qofo

Yes, the 8350 looks really good in POV-Ray also. Not a bad choice if all you do is run CPU intensive programs like that and can disregard efficiency.

 

[MiddleOfTheRoad]

Visheras and Sandy Bridges are both blasts from the past, imo, but the defining characteristic of the 2500K, especially OCed, is that its per thread performance is pretty tough to beat. A full-tilt 9590 doesn't even start to eclipse the performance of a 2500K until about 6 fully loaded threads or higher.

Dude, you're way off. My lowly FX-8320 keeps pace with my 3770K, which is considerably faster than a 2500k..... Enough of your Anti-FX bias, you've clearly never used one.

 

[Abwx]

Dude, you're way off. My lowly FX-8320 keeps pace with my 3770K, which is considerably faster than a 2500k..... Enough of your Anti-FX bias, you've clearly never used one.

Crashtech is obviously not of bad faith as aknowledged by his post above, there s a heavy tendency that is assuming that the first bulldozer iterations scores are a rule, it s just that 3 years later softs are noticeably more Mthreaded and that what was accurate at the time is no more relevant, since i mentionned Hardware.fr FX8350 getting 12.5% better games scores with slightly updated games i will add that even the 4770K gained 6% over the 4670K despite games being not fully Mthreaded, if we look at the available computing capabilities the 4770K is on a 20-25% better perfs course in respect of the 4670k while the 8350 has even more reserve left.

 

[Abwx]

Yes, the 8350 looks really good in POV-Ray also. Not a bad choice if all you do is run CPU intensive programs like that and can disregard efficiency.

You d be surprised by efficency numbers when estimating the real numbers...

The 8370E in isolation manage to reach or even outmatch in some cases the efficency of Intel s i5 HW offerings but the plateform is litteraly ruining a sizeable part of thoses progress, for the record i ll mention that the 8370E power comsumption at 3.3 is 80W for Prime 95 and 65W in Fritz where the 4670K has about 10% better perf/watt while the 8370E has better perf/watt in the two other chess benches used by Hfr, same with 7zip and Winrar and generaly integer tasks that scale well while the i5 has the edge in FP related softs.

At the risk of repeating myself i point that theses estimations are for the CPU only, they can be extracted from Hardware.fr tests since they also measure the CPU power drain at the 12V rail level wich allow very accurate computations.

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

 

[MiddleOfTheRoad]

Crashtech is obviously not of bad faith as aknowledged by his post above, there s a heavy tendency that is assuming that the first bulldozer iterations scores are a rule, it s just that 3 years later softs are noticeably more Mthreaded and that what was accurate at the time is no more relevant, since i mentionned Hardware.fr FX8350 getting 12.5% better games scores with slightly updated games i will add that even the 4770K gained 6% over the 4670K despite games being not fully Mthreaded, if we look at the available computing capabilities the 4770K is on a 20-25% better perfs course in respect of the 4670k while the 8350 has even more reserve left.

Still -- Crashtech constantly rambles about these things like they are set in stone..... When in real world usage the Vishera can keep pace with an equivalent Ivy Bridge for nearly everything outside of some older games.

 

[crashtech]

Dude, you're way off. My lowly FX-8320 keeps pace with my 3770K, which is considerably faster than a 2500k..... Enough of your Anti-FX bias, you've clearly never used one.

Depends on the usage, as a blanket statement, what you assert is not supported by evidence.

 

[crashtech]

You d be surprised by efficency numbers when estimating the real numbers...

The 8370E in isolation manage to reach or even outmatch in some cases the efficency of Intel s i5 HW offerings but the plateform is litteraly ruining a sizeable part of thoses progress, for the record i ll mention that the 8370E power comsumption at 3.3 is 80W for Prime 95 and 65W in Fritz where the 4670K has about 10% better perf/watt while the 8370E has better perf/watt in the two other chess benches used by Hfr, same with 7zip and Winrar and generaly integer tasks that scale well while the i5 has the edge in FP related softs.

At the risk of repeating myself i point that theses estimations are for the CPU only, they can be extracted from Hardware.fr tests since they also measure the CPU power drain at the 12V rail level wich allow very accurate computations.

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

Most people overclock Visheras in an effort to keep them competitive with newer offerings. Even AMD does this with the 95xx series. That's when the power consumption becomes truly problematic, as it also does on Intel CPUs when they are pushed too hard. It's nice that the 8370E is able to run in a lower TDP envelope, but as far as I can see it necessarily sacrifices some performance to do so.

I think part of what makes theses discussions so tough is that while Intel clearly shows a single thread lead, and AMD 4M/8T CPUs clearly do well at heavily threaded tasks, the reality is that most users fall somewhere in between these extremes 99% of the time. So usage case is a gray area and we can come up with counter examples all day about which case applies and which approach is better. I happen to be sold on having the right number high IPC cores for the job, but the fact that they are AMD or Intel has zero significance to me.

 

[Abwx]

Still -- Crashtech constantly rambles about these things like they are set in stone..... When in real world usage the Vishera can keep pace with an equivalent Ivy Bridge for nearly everything outside of some older games.

Years have passed, a lot of softs have been updated and some people are still stuck mentaly with old and no more relevant scores, not talking of Crashtech if course but more generaly of people who do not bother doing their homework, personaly i read all CPUs reviews whatever the brand or models.

 

[frozentundra123456]

Depends on the usage, as a blanket statement, what you assert is not supported by evidence.

Dont confuse them with facts and benchmarks. Some posters seem to value personal experience or feelings more than objective benchmarks, or at best want to focus on the benchmarks that fit their agenda and disregard the rest. And the fact that we are arguing about a two generation old intel processor vs top of the line current AMD one is quite telling. But as usual we have gotten far off topic.

 

[Abwx]

Most people overclock Visheras in an effort to keep them competitive with newer offerings. Even AMD does this with the 95xx series. That's when the power consumption becomes truly problematic, as it also does on Intel CPUs when they are pushed too hard. It's nice that the 8370E is able to run in a lower TDP envelope, but as far as I can see it necessarily sacrifices some performance to do so.

15 years or so ago CPUs had barely enough computing capabilities to satisfy fully even a modest user, upgrading from say 600MHz to 800Mhz did clearly improve the experience but currently?
Do you think that if a CPU does 8 or 5 or 3 in CBench there will be a difference in the experience ?.
The answer is none, moreover the CPU that does 3 will provide much more comfort if it has a SSD than the one that does 8 but is HDD dependent.

People who overclock to death are rather chasing records and personal satisfaction than anything else, if it was only for the perfs they would take dual sockets with massive core counts.

I think part of what makes theses discussions so tough is that while Intel clearly shows a single thread lead, and AMD 4M/8T CPUs clearly do well at heavily threaded tasks, the reality is that most users fall somewhere in between these extremes 99% of the time. So usage case is a gray area and we can come up with counter examples all day about which case applies and which approach is better. I happen to be sold on having the right number high IPC cores for the job, but the fact that they are AMD or Intel has zero significance to me.

The applications that are still ST are so because they dont need MT, if there was a lack they would have been promptly updated, it s not like there s no 1.5 or 2Ghz CPUs, tech firms know about it and they know that their softs wouldnt be attractive if providing poor experience, applications that benefit from MT are often Mthreaded without the users knowing it, just check Firefox, open a tab and look at the task manager, with my AM1 PC (4 cores at Athlon X4 1.7 level) all four cores are fired when i open even a page as simple as AT forums, currently i m using my old faithfull Pentium 2C 2.2, not as good as the AM1 but quite enough for the usage, i even do Spice simulation with it, not that fast but it takes me much more time to change components and values on the schematics than waiting for runs..

That, lengthily, said, is there applications where you feel that you are CPU limited.?..

 

[Abwx]

Dont confuse them with facts and benchmarks. Some posters seem to value personal experience or feelings more than objective benchmarks, or at best want to focus on the benchmarks that fit their agenda and disregard the rest. And the fact that we are arguing about a two generation old intel processor vs top of the line current AMD one is quite telling. But as usual we have gotten far off topic.

You wont negate that i provided numbers to sustain my points, i fully accept thoses numbers being contradicted but then they should be contradicted by other real numbers and not by statements that only express an opinion without providing the slightest technical data, can we agree at least on this point.?.

So please tell me what was disregarded, because one more time that is a general statement with no precisions, actualy if you had really read your post you would see that it s nothing else that an ad hominem attack, that is, you are questionning my credibility as a mean to question my numbers , i could do this as well but i prefer to have some credibility, numbers are facts but saying "facts" is not a fact...

 

[crashtech]

15 years or so ago CPUs had barely enough computing capabilities to satisfy fully even a modest user, upgrading from say 600MHz to 800Mhz did clearly improve the experience but currently?
Do you think that if a CPU does 8 or 5 or 3 in CBench there will be a difference in the experience ?.
The answer is none, moreover the CPU that does 3 will provide much more comfort if it has a SSD than the one that does 8 but is HDD dependent.

People who overclock to death are rather chasing records and personal satisfaction than anything else, if it was only for the perfs they would take dual sockets with massive core counts.



The applications that are still ST are so because they dont need MT, if there was a lack they would have been promptly updated, it s not like there s no 1.5 or 2Ghz CPUs, tech firms know about it and they know that their softs wouldnt be attractive if providing poor experience, applications that benefit from MT are often Mthreaded without the users knowing it, just check Firefox, open a tab and look at the task manager, with my AM1 PC (4 cores at Athlon X4 1.7 level) all four cores are fired when i open even a page as simple as AT forums, currently i m using my old faithfull Pentium 2C 2.2, not as good as the AM1 but quite enough for the usage, i even do Spice simulation with it, not that fast but it takes me much more time to change components and values on the schematics than waiting for runs..

That, lengthily, said, is there applications where you feel that you are CPU limited.?..

Not sure what case you are making here, but it doesn't seem to be one for the 8370E, from your observation we should be recommending the i3 or FX-6300?

 

[crashtech]

...i fully accept thoses numbers being contradicted but then they should be contradicted by other real numbers and not by statements that only express an opinion without providing the slightest technical data, can we agree at least on this point.?...

I know you weren't directly addressing me, but I want to respond. Surely you must know that the web is replete with numbers that indicate AMD's inferior position in general. Even the articles you cite, when read in their entirely, support this. If only the 8370's competition was the 2500K, then the decision would be a lot harder.

http://www.anandtech.com/bench/CPU/1062

http://www.anandtech.com/bench/CPU/1052

I ought to put at least one single-thread bench in here to make my point:

http://www.anandtech.com/bench/CPU/1028

Here's an interesting read that helps illustrate the difference between approaches, and a chart that maps out ranges of loading and approximate throughput for dozens of different CPUs. Should be taken with a grain of salt because the methodology is not revealed:

http://www.overclock.net/t/1493307/relative-access-to-execution-throughput-comparison-chart