FX chips better than i5 for video editing?

[Abwx]

The best would be something like a 5960X but that is obviously not in the same price or performance bracket.

Not sure that it s a good investment considering the total price, 71% more perfs in applications than a 8350 at 5-6x the price for the CPU, dont know how much times for the MB and RAM...

http://www.hardware.fr/articles/924-19/indices-performance.html

Note also that you will have to look at the individual program. Some are still singlethread heavy.

What are thoses "some" exactly.?

Because about all the single thread benches i see here and there are extracted out of Multithreaded softwares, generaly CB 11.5 or R15, not denying that there are softs wich are badly trhreaded but i d like to know wich ones are annoyingly slow on a FX other than the "some" that are undefined...

On Hardware.fr suite there are a few according to their own words but still the FXs performs quite well in all their tests :

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

 

[AtenRa]

With a 'cheap' [$30/$40] cooler you can put one of this $70 6 cores [X5650] @ 3.9/4Ghz,


If you don't like to read that a $70 cpu with 4+ years beat all FX cpu's, well.. deal with it.

The same 4+ years CPU will beat every currently Intel CPU up to $300 in MT loads.

So ???

 

[Cerb]

Please, educate yourself in this thread.

With a 'cheap' [$30/$40] cooler you can put one of this $70 6 cores [X5650] @ 3.9/4Ghz, and at that point you'll have a system that can at least equal a i7 4770/4790K @ 4.5/4.6Ghz in multithread.

The alleged $70 superior CPU has been magically unmasked as a $100+ CPU, that still requires substantial OCing to make faster. :| That's my shocked face, at this revelation.

As you can see on anandtech cpu comparison a stock i7 990X will beat a FX9590 stock in single thread.
And that's a 3.5/3.7GHz x58 i7 vs a 4.7GHz FX cpu.

And the used price for that i7 is not far off from the new price of that FX.

X5650 have a similar performance as a i7 990X @ same MHz.
So from that comparison chart you can make a educated guess how a 4Ghz X5650 will quick Fx8xxx or even the 220w FX beast.

Not so great, given the speed is 2.4GHz. If you're going to bring it to 3.5-4GHz, the FX deserves to be brought to 5.2-5.5GHz for the same comparison. OC v. OC or stock v. stock. And, if it needs to be OCed to meet or beat a stock CPU, the cooler costs must be counted, as well.

If you don't like to read that a $70 cpu with 4+ years beat all FX cpu's, well.. deal with it.

The problem is falsehoods, like it being a $70 CPU that's faster. You're doing the same w/ Intel as we keep seeing from AMD fanboys. It's a $70 CPU that, with more money and time, can be made to perform faster. Big difference.

 

[Mk pt]

The same 4+ years CPU will beat every currently Intel CPU up to $300 in MT loads.

So ???

So....

I bought a couple weeks ago a old (5 years!? ) intel for $70 and it kicks AMD FX cpu's @ss in single and multithread.

So if you want single thread performance buy a good i3 or cheap i5..

If you want multithread performance pay for the best - X99 platform or if you're on a small budget buy a X58 motherboard and a Xeon X5650.

My first post was accurate..

 

[Cerb]

So....

My first post was accurate..

Except for missing the necessary OC for performance v. optional in the new platforms, and the first statement has the cost wrong and the ass kicking part wrong. The old Xeon is a good value, but if it requires such unfair comparisons to look good, it's not kicking the competitor's ass, just saving you $100-250 by not buying new.

 

[crashtech]

At this point I want to vote "none of the above" for the purpose suggested in the OP, and recommend instead a 4C/8T Xeon E3-1231v3. You get about 95% of the MT perf of the 9590, with about 125% of the ST perf, using much less power, for about $250. No overclocking needed.

 

[Abwx]

At these level ther s no substancial difference between a 8350 and a 9590, so why pointing the 220W variant rather than a regular 125W.?.

After all a 8350 will provide about 95% of the MT perfs of this Xeon, you see, it can work the other way around..

 

[crashtech]

At these level ther s no substancial difference between a 8350 and a 9590, so why pointing the 220W variant rather than a regular 125W.?.

After all a 8350 will provide about 95% of the MT perfs of this Xeon, you see, it can work the other way around..

Actually there is no nefarious intent, merely using AMD's flagship to show them in the best possible light, and to simulate, if you will, the maximum potential of the 8350 et al. But if we were to show the scenario you mention, the ST calculation becomes much worse for AMD. And even though to constantly dismiss this metric, fact is that most multipurpose PCs spend most of their time under partial and not full load, and that is where the ST measurement becomes useful in predicting performance.

 

[Abwx]

fact is that most multipurpose PCs spend most of their time under partial and not full load, and that is where the ST measurement becomes useful in predicting performance.

This will predict absolutely nothing since you dont know if the PC will be ST or MT loaded when computing ressources are needed, a PC may be idling while i m moving the pointer to click on whatever settings, and in that case ST perf doesnt matter at all, once you click "run" of whatever compute heavy task you can be sure that it will be Mthreaded and in this case what matter is the available throughput, so you could have 2 high IPC cores and 4 cores with each 60% of the ST perf of the formers, the second case will be faster if in MT and the higher ST perfs of the first quoted cores was an insufficent parameter to predict the part of the perfs that actualy matters.

 

[crashtech]

This will predict absolutely nothing since you dont know if the PC will be ST or MT loaded when computing ressources are needed, a PC may be idling while i m moving the pointer to click on whatever settings, and in that case ST perf doesnt matter at all, once you click "run" of whatever compute heavy task you can be sure that it will be Mthreaded and in this case what matter is the available throughput, so you could have 2 high IPC cores and 4 cores with each 60% of the ST perf of the formers, the second case will be faster if in MT and the higher ST perfs of the first quoted cores was an insufficent parameter to predict the part of the perfs that actualy matters.

I disagree, multiple instances of ST benches in particular can better simulate a partial load than extrapolating back from a 100% load. Unfortunately partial load is where the performance penalty of CMT vs SMT becomes more apparent, so from your point of view it would be better to emphasize performance at 100% load, where AMD's modular design is performing at its best.

 

[Cerb]

The CPU won't be idle after you move the mouse and click something, though. That's when it does a burst of work. Where is the multithreaded client-side Javascript engine (sad or not, this is one of the most relied on technologies, today)? Where are the multithreaded GUIs? How many client-side Java or Citrix applications are multithreaded? Even many of us that can use many threads tend to do so by way of many processes, which will still behave like ST. Twiddling one's thumbs due to a heavy multithreaded compute task is a rarity. Twiddling one's thumbs because a web page takes 10s when it could take 6s is very common, OTOH. The common multithreaded tasks people have to wait on are almost all interactive, as well, where both the performance of each thread, and the performance of the aggregate, matter (gaming and content creation).

And, the performance penalty of CMT is not what is apparent. It's the performance penalty of a speed demon running slow. There's no good reason AMD couldn't have targeted lesser clock speeds, and had as good as or superior IPC than Stars, per int unit.

 

[Mk pt]

Not so great, given the speed is 2.4GHz. If you're going to bring it to 3.5-4GHz, the FX deserves to be brought to 5.2-5.5GHz for the same comparison. OC v. OC or stock v. stock. And, if it needs to be OCed to meet or beat a stock CPU, the cooler costs must be counted, as well.

X5650 is a 2.66GHz cpu with turbo to 3GHz.
And 4Ghz is really easy to achieve.

And X5650 price as dropped to $65.

Not so great, given the speed is 2.4GHz. If you're going to bring it to 3.5-4GHz, the FX deserves to be brought to 5.2-5.5GHz for the same comparison. OC v. OC or stock v. stock. And, if it needs to be OCed to meet or beat a stock CPU, the cooler costs must be counted, as well.

Right, oc vs oc, can you put a FX-8xxx or FX-9590 @ 5.5Ghz for daily use?
Using which cooling and with what vcore [that doenst 'kill' the cpu over time]?

A X5650@ 4GHz is as easy as use a average cooler and stock [or less] vcore, for daily use.


Still, if a stock 990x beats a stock FX9590.. that's 3.5/3.7 vs 4.7/5ghz, you really think a 4Ghz 990x will not be able to equal or beat a 5.5Ghz FX?

That's a 14%/8% [base/turbo] oc for 990X and 17%/10% oc for FX9590.

If a 3.5/3.7GHz 990X is able to beat a 4.7/5GHz FX9590, it means that for any additional 100Mhz in both CPU'S, the performance increase will be great in the i7 990x.
So with that 17%/10% oc increase vs 14/8% at most will make the FX9590 @5.5Ghz loose to i7 990X for a similar distance that they have in stock.

The problem is falsehoods, like it being a $70 CPU that's faster. You're doing the same w/ Intel as we keep seeing from AMD fanboys. It's a $70 CPU that, with more money and time, can be made to perform faster. Big difference.

Falsehoods is saying 'with more money and time can be made to perform faster'.

I spend $70 and I have more performance that any FX cpu. That's a fact.
I loose a couple hours in oc because I wanted to see where my X5650 could get for daily use, it's running @4.1ghz and I tested/did benchmarks @ 4.2GHz.
But if you want to achieve a lower point [4ghz] and don't try to understand what max oc you can achieve you don't need to loose so much time.

$70+time spent in oc for better performance than a $170/180 FX8xxx or >$200 FX9590? Or even $300 i7 47xxK?

Yes, that's what I did.

Even if I didn't had a X58 board and decent cooler, why not a $70* CPU + $50 Cooler + $180 board vs a $180 cpu or >$200 cpu +$100/120 board, when I'll get more performance with first option?

*current X5650 price is $65


The real problem is AMD FX is not a good choice even for multithread.
And for some persons that's difficult to accept.



And I'm not against AMD neither I'm a Intel fan.
I wish AMD could sell nice cpu's, better that any Intel cpu.
I like the underdogs and dislike those corporations that have a monopoly, or are close to get that.

But my money doesn't grow on trees so I most compare and choose what's best for my budget.
That's why I have a R9 290 - AMD you see - it was the bang for the buck at that time.
Because I'm not one of those who say AMD GPU has drivers issues because I had a pair of hd4890 too, without issues. As a matter of fact I had more issues with drivers with my GTX570 than with 4890's or with my r9 290 now.

 

[Enigmoid]

Not sure that it s a good investment considering the total price, 71% more perfs in applications than a 8350 at 5-6x the price for the CPU, dont know how much times for the MB and RAM...

http://www.hardware.fr/articles/924-19/indices-performance.html

What are thoses "some" exactly.?

Because about all the single thread benches i see here and there are extracted out of Multithreaded softwares, generaly CB 11.5 or R15, not denying that there are softs wich are badly trhreaded but i d like to know wich ones are annoyingly slow on a FX other than the "some" that are undefined...

On Hardware.fr suite there are a few according to their own words but still the FXs performs quite well in all their tests :

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

Maybe there is a world outside the all inclusive perf/$ world where |performance| matters too.

I never said that it was the best value. I merely said it was the best.

Here is an example though for photoshop.

We use two distinct Photoshop benchmarks, one of which fully taxes each processor’s x86 cores using well-threaded filters, and another that is OpenCL-optimized to leverage graphics resources. Don’t compare the black and red bars above—they’re only together to save space (and your scrolling finger).

Calling this the OpenCL version of our benchmark is a little disingenuous, since we already ran our benchmarks based on each processor's on-die graphics engine using this same workload. Here, all platforms are accompanied by Nvidia’s GeForce GTX Titan, so the differences are wholly attributable to CPU interaction.

Looks like the OpenCL test is singlethreaded.

There may be other examples but they are often not reviewed.

 

[Cerb]

X5650 is a 2.66GHz cpu with turbo to 3GHz.

Yup. Bad memory. Not enough to change performance, though.

And 4Ghz is really easy to achieve.

By just plugging it in, with no added cooler?

Right, oc vs oc, can you put a FX-8xxx or FX-9590 @ 5.5Ghz for daily use?
Using which cooling and with what vcore [that doenst 'kill' the cpu over time]?

I don't, and I wouldn't. But, I also wouldn't run an old Xeon OCed by 50%. The new chips seem to be doing well at low Vcores, from I read, though. I'd be using the Xeon at stock speed, too.

Still, if a stock 990x beats a stock FX9590.. that's 3.5/3.7 vs 4.7/5ghz, you really think a 4Ghz 990x will not be able to equal or beat a 5.5Ghz FX?

That's a 14%/8% [base/turbo] oc for 990X and 17%/10% oc for FX9590.

In some things, no (fewer). In some, yes (more). In particular, where the FX is stronger, it should remain so. The OC will still increase its performance. Don't you do TBs-worth of AES per day?

The i7-990x also tends to cost 2-3x the FX-9590 today, used. I can't see why, given that you can beat it for that cost with brand new stuff, but some people are just crazy, and older high-end CPUs keep their cost, even used. The i7-960 is still >$200, as well, and the i7-930 >$100. Boggles my mind, it does, even for an upgrade market.

If a 3.5/3.7GHz 990X is able to beat a 4.7/5GHz FX9590, it means that for any additional 100Mhz in both CPU'S, the performance increase will be great in the i7 990x.
So with that 17%/10% oc increase vs 14/8% at most will make the FX9590 @5.5Ghz loose to i7 990X for a similar distance that they have in stock.

Where the FX strong, it will still be, with the OC (but still usually lose, being basically 4C v. 6C). In the 9000 case, the stock cooler is sufficient to do it, too. Those two CPUs would generally be neck and neck in most things, the Intel having a slight edge--but it would be more than slight if the AMD were to be compared against at stock.

Falsehoods is saying 'with more money and time can be made to perform faster'.

I spend $70 and I have more performance that any FX cpu. That's a fact.

And the rest cost nothing?

or if you're on a small budget buy a X58 motherboard and a Xeon X5650.

Implying a fair price comparison at that $70.

Even if I didn't had a X58 board and decent cooler, why not a $70* CPU + $50 Cooler + $180 board vs a $180 cpu or >$200 cpu +$100/120 board, when I'll get more performance with first option?

You might. There are variables; and to compare costs, they should be controlled for as well as possible, though, to compare different parts. A $70 CPU needing a $50 cooler makes for $120, whereas the compared-to option needs no extra cooler. Though, TBH, I don't see why a $120 mob should be needed, unless you're keeping all the cores busy a lot of the time (if you're making money from doing so, you won't be buying used, and should go straight for X99). Just don't get 6C, and you can have a mobo and 4C8T CPU for $312, brand new, with 6Gbps SATA and USB 3. 6C is goo if you keep it fully loaded a lot, but the returns are going to be quite diminsihing.

The real problem is AMD FX is not a good choice even for multithread.
And for some persons that's difficult to accept.

That much is true as generally applicable statement. The FX has a few decent niches due to price points, but is not a good genera purpose value, and thus the low market share.

Because I'm not one of those who say AMD GPU as drivers issues because I had a pair of hd4890 too, without issues. As a matter of fact I had more issues with drivers with my GTX570 than with 4890's or with my r9 290 now.

I'm likely in for more driver issues with the GTX 970, too. AMD just got stuck with ATI's old rep...but I can't turn down fans not running at all.

 

[BSim500]

Yeap and by doing that you have more time your PC powered on using more energy

Meanwhile back in the real world:-

As "Essence of War" already said, your "race to idle" claim is wildly false if your CPU is using 60% more energy for hours on end then finishes only 2% faster vs even an old i5-3470 Ivy Bridge clocked 800MHz slower...

My original post wasn't even about that anyway, it was about the hilarious lack of common sense of those who claim running multiple games & video encodes at the same time is "typical usage" - but strangely they bring such a usage scenario up only when it suits AMD's, primarily to try and find something for idle cores to do during less well threaded games to skew the scores in favor of CPU's with more numerous but weaker cores... Intel fanboys on the other side of the fence could be equally dishonest and start quoting Starcraft 2 benchmarks as the "average typical" gaming example... :thumbsdown:

 

[AtenRa]

As "Essence of War" already said, your "race to idle" claim is wildly false if your CPU is using 60% more energy for hours on end then finishes only 2% faster vs even an old i5-3470 Ivy Bridge clocked 800MHz slower...

Toms review didnt do multi-tasking, they just run a single bench at a time. Big difference

 

[BSim500]

Toms review didnt do multi-tasking, they just run a single bench at a time. Big difference

I think you'll find most people run only one benchmark at once for obvious reasons...

 

[AtenRa]

I think you'll find most people run only one benchmark at once for obvious reasons...

People in real life use more than one application at a time for obvious reasons

 

[BSim500]

People in real life use more than one application at a time for obvious reasons

Yes they do, but they still don't run several artificial benchmarks together simultaneously...

 

[Abwx]

Meanwhile back in the real world:-

Haswell for the perf and Ibridge for the perf/watt.??.

Nice one when one knows that IB has 15-20% better perf/watt than Haswell wich would be at about 260-270 on this chart, the 8370E should be not so far...

Yes they do, but they still don't run several artificial benchmarks together simultaneously...

So in short thoses benchmarks are not representative of real world usage..so much for tests that are supposed to be used as reference...for real world usages..

 

[Sweepr]

Haswell for the perf and Ibridge for the perf/watt.??.

Nice one when one knows that IB has 15-20% better perf/watt than Haswell wich would be at about 260-270 on this chart, the 8370E should be not so far...

Except it doesn't. Oops.

 

[BSim500]

Haswell for the perf and Ibridge for the perf/watt.??.

No idea what you're talking about. There wasn't a single Haswell anywhere in the review (3770K / 3470 is Ivy Bridge not Haswell):-
http://www.tomshardware.com/reviews/fx-8350-vishera-review,3328-16.html

So in short thoses benchmarks are not representative of real world usage..so much for tests that are supposed to be used as reference...for real world usages.

The ONLY time I ever see "Oh yeah, well what if you wanted to play Watch Dogs and encode 500x 4K video's whilst scanning your system for Malware and printing a 200 page PDF document in the background all at the same time" style proclamations is by those wildly distorting what they believe "the average person does" for the purposes of skewing benchmarks in favor of high-core-count-but-low-IPC CPU's. It's not even a "benchmark" (since no-one has provided any figures or even attempted to do so) it's just a personal anecdote.

I searched for "video encoding whilst gaming benchmark" and the one single site that attempted to do so was Tech Report whose results were probably not quite what you meant... There were no other sites, benchmarks or any attempt at objectively benchmarking what you claim are the "one true benchmarks". Why? Because no-one's interested in doing it from day to day outside of benchmark cherry-picking! It's a "solution looking for a problem".

 

[Abwx]

Except it doesn't. Oops.

I assure you that if you use the same set up and measure the actual CPU comsumption it does..

http://www.hardware.fr/articles/913-9/cpu-consommation-overclocking.html

 

[Phynaz]

4.8GHz at 1.375V on 22nm process, you really find this serious, would you advise such settings.?.

Btw, at this frequency and using their numbers, that is 1.04V at 3.2Ghz and 30.8W measured with P95, they got this frequency at 80W TDP with P95 , quite an achievement in the waiting for the CPU to be damaged by this overvoltage, have you a few other exemples of the same barrel.?.

You moved the goal posts. I replied to perf/$. The Pentium smokes anything FX in perf/$. Period.

Have any proof that is an unadvisable overclock?

 

[Idontcare]

I assure you that if you use the same set up and measure the actual CPU comsumption it does..

In what way does that matter?

(versus at-the-wall consumption for the entire platform, which is what the customer is paying the electrical company for when they use their computer to get the job done?)

It seems like an academic argument at best (who has the best architecture if you ignore XYZ, etc. etc.) but one that isn't relevant once you start down the road of "real world application and use models of actual customers" because those "actual customers" aren't just paying for the electricity the CPU uses.