Do AMD cpus at least give a smoother desktop experience w/more cores?

[TheELF]

The comparison to the Core i3 Ivy was only to illustrate that even with a HUGE node advantage of FineFets, and still the AMD Kaveri on a inferior process on 28nm PLANAR was still faster and with better perf/watt. Even in CPU workloads, not to mention iGPU performance. Ohh, dont forget at the same TDP of 55W.

Yes let's not forget that.
The TDP is supposed to be for the whole of the chip, CPU and IGPU TOGETHER,kaveri heavily throttles the one when the other is under load exactly for benchmarking like the ones you do,just so it appears to be faster.
But if someone was to look at your very own Luxmark:sala bench that uses both of them together you would get closer to the truth, suddenly the i3-3225 and the a8-7600 are very close both in performance and in power draw in a scenario that completely favors a multicore CPU and a good GPU no less,but I guess people should only use one of the components at a time no need to be greedy and actually use a CPU for everything at the same time.
And yes, all the games have in build benches that only push the GPU as well.

And then some people wonder what all that "biased" talk is all about...

 

[ShintaiDK]

Standard operating procedure: cherry picked benchmarks comparing AMD to a 3 generation old Intel product.

Exactly same since Bulldozer release. Pick the best case benchmark and keep chanting how great it is. And hope nobody questions it.

 

[ShintaiDK]

Speaking of the devil,looks like he benchmarked the Core i3 6300:
http://forums.anandtech.com/showthread.php?t=2456159&highlight=

So he actually already benchmarked the top end Core i3 and the first reply was from the same poster,who called him biased. Emm,WTF??

I3 6300 isn't "the top end". That's for the i3 6320.

 

[DAPUNISHER]

So basically all of the overclocking scene comes down to luck?
:\

Certainly not. While there are always winners and losers in the CPU silicon lottery. Because so many share their experiences, an average expectation can be determined for a particular part. Most that follow the advice provided, will have no difficulty hitting the mean overclock speed with optimal settings. Getting a part that will exceed the average easily, or fail to manage the average without unsafe settings, is where good and bad luck enter.

It is part of the hobby. One this forum provides a discussion platform for.

Not surprising then, that perf/watt is so much more attractive these days

It has always been attractive to silent PC enthusiasts, and those needing a SFF PC. And of course it is integral to mobile. As long as case and component cooling are adequate, it is less essential to some. Though it makes a great bullet point for advertising and marketing.

 

[frozentundra123456]

Exactly same since Bulldozer release. Pick the best case benchmark and keep chanting how great it is. And hope nobody questions it.

Yes, Aten's very own charts dont back up his claims. Looking at the graph he posted for x264, a best case benchmark for AMD, the i3 3225 takes 80 watt hours to finish the benchmark, while the A8 7600 takes the same, so the AMD is similar in efficiency in that particular benchmark, not more efficient. What he conveniently fails to mention is that in his very own chart, the Haswell i3 finishes the task in 60 watt hours.

 

[mrmt]

My FX was clocked at 4.4GHz, pretty high on the wrong end of the efficiency curve. I'm sure if I clocked it at something like 3.6Ghz I could get a huge jump in efficiency, and probably about match your score. I don't know I could be as efficient, but I know it'd be a lot closer than at 4.4GHz.

I doubt you can match the efficiency. Vishera was trounced on the server market by Sandy Bridge-EP and EP on the server market exactly because of poor efficiency, and that was on frequencies much lower than those you are aiming competing against IVB.

 

[Yuriman]

Yes, Aten's very own charts dont back up his claims. Looking at the graph he posted for x264, a best case benchmark for AMD, the i3 3225 takes 80 watt hours to finish the benchmark, while the A8 7600 takes the same, so the AMD is similar in efficiency in that particular benchmark, not more efficient. What he conveniently fails to mention is that in his very own chart, the Haswell i3 finishes the task in 60 watt hours.

It IS sortof a point in AMD's favor that on 28nm, their CPUs are of similar efficiency to Intel's first generation 22nm products - that's not terrible. Yes, it's meaningless to say "well, it's great for what it is", since that's not a compelling argument for picking one product over another, but nobody complains that Ivy Bridge is a power hog, even if it's not competitive with Skylake. So long as there's still some other positive characteristic, Kaveri isn't obsolete. The performance is not bad; there are still some cases where Kaveri could be a compelling buy, given that the iGPU is generally better than equally-priced Intel chips. The i3 6100 certainly closes up some niches where AMD was the better choice, and for AMD to keep selling chips, they're going to need to get their refresh out soon.

It's also true that AMD parts arguably have too low a TDP to make full use of both iGPU and CPU, which seems nonsensical in a desktop.

 

[AtenRa]

Did you measure the actual CPU usage? Because it somehow dont seem to align with how review sites ended up. Excuse for no 3225 there, but the world had moved on. However I doubt 3225 to 4330 made that big of an impact.

http://techreport.com/review/25908/amd-a8-7600-kaveri-processor-reviewed/12

Did you bother to see what you are posting ??? Those are TOTAL SYSTEM ENERGY consumption.

Also, techreport only uses 140 seconds single pass, my benchmark is using first and second pass and the entire bench last for more than 40-50 minutes.

Ohh, their results of Core i3 4330 match mine as well but you forgot to see that some how.

I assume you used 2400Mhz with the Kaveri and 1333 or 1600Mhz with the IB?

2133MHz for Kaveri and 1600MHz for IvyBridge, the official supported memory clocks.

 

[AtenRa]

Yes, Aten's very own charts dont back up his claims. Looking at the graph he posted for x264, a best case benchmark for AMD, the i3 3225 takes 80 watt hours to finish the benchmark, while the A8 7600 takes the same, so the AMD is similar in efficiency in that particular benchmark, not more efficient. What he conveniently fails to mention is that in his very own chart, the Haswell i3 finishes the task in 60 watt hours.

A8-7600 at 55W TDP is FASTER than Core i3 3225 while they both use the same energy to finish the benchmark. That makes the A8-7600 more efficient with a higher perf/watt than the Core i3 3225.

 

[ShintaiDK]

No it doesn't. Performance/watt is exactly the same in your homebrew test. A test with a 10W increment btw. And it doesn't matter how long you ran your tests. The 3225 in your case also use more power than it should due to the Maximus V Gene. That's why I asked you how you measured it. Because its obvious you didn't meansure CPU usage.

And if you wanted to show an actual view of the CPUs themselves as you claim. Then you wouldn't use 2133Mhz on Kaveri either. You already decided to compare products with years in between to claim your wrongfully outcome.

 

[ShintaiDK]

Yes, Aten's very own charts dont back up his claims. Looking at the graph he posted for x264, a best case benchmark for AMD, the i3 3225 takes 80 watt hours to finish the benchmark, while the A8 7600 takes the same, so the AMD is similar in efficiency in that particular benchmark, not more efficient. What he conveniently fails to mention is that in his very own chart, the Haswell i3 finishes the task in 60 watt hours.

Yes, just one of many things. But the sole cherry picking alone removes any relevance to his "testing" for good. Also his test is no different than what we know from review sites, yet we somehow need to see it again in the narrow AMD pro selection that is with whatever changes done.

 

[Yuriman]

In this case Shintai is right (not to imply anything!) - if they take the same amount of energy to do a task, they have the same efficiency. You nicely isolated that from performance, which is another matter entirely.

 

[AtenRa]

In this case Shintai is right (not to imply anything!) - if they take the same amount of energy to do a task, they have the same efficiency. You nicely isolated that from performance, which is another matter entirely.

Sorry but efficiency is measured from performance and energy usage.

Example from Green 500 SuperComputers, Efficiency is measured from performance (FLOPs) and power consumption or Perf/watt.

Just because Core i3 3225 finished one task using the same amount of energy as the A8-7600 doesnt mean it has the same efficiency. In order to measure Efficiency you also count the performance you have using the same amount of energy or at the same performance the amount of energy used.

In the case of x264,
At the same energy used the faster CPU has a higher Efficiency or
At the same performance the lower energy usage has the higher efficiency

You need both metrics of energy and performance to derive to Efficiency.

 

[frozentundra123456]

Well, if you want to talk about a variety of tasks, here is the full set of benchmarks for an i3 vs the A8-7600 at 65w. anand bench.

I also agree with Yuriman, you can pick some esoteric definition if you like, but in a general scientific sense, "efficiency" in regards to power would be the area under the curve of power consumption vs time to complete the task. How fast the task is finished is another issue. But in any case, it you want to look at performance, the benches I listed show that a modern Haswell i3 slaughters the A8-7600 in all but a few cpu benchmarks. And that is with the A8 at 65 watts, not the 45 watts you picked in your example to make the efficiency look better.

 

[Yuriman]

I'm thinking "efficiency" in terms of performance per watt, as fruzentundra said; If you take 25% longer to do a task, but use 20%* less energy per unit of time (resulting in the same total energy used to do the task), the performance per watt is the same. By that metric, "watt-hours to finish the benchmark" in a benchmark that takes variable time to finish should be synonymous with "performance per watt".

 

[TeknoBug]

And that is with the A8 at 65 watts, not the 45 watts you picked in your example to make the efficiency look better.

And with the BIOS setting to 45W, the performance efficiency is reduced a bit compared to that Anand benchmark.

 

[Sweepr]

A8-7600 draws an extra 30W compared to Core i3-4130 @ Luxmark CPU and gaming according to Hardware.fr. Looking at the performance numbers there's no way it's more efficient in general.

Standard operating procedure: cherry picked benchmarks comparing AMD to a 3 generation old Intel product.

Preferably with 1 year old drivers and/or crippling the Intel system with slower memory (don't forget to blame the chipset and claim 'maximum officially supported clocks').

 

[AtenRa]

Well, if you want to talk about a variety of tasks, here is the full set of benchmarks for an i3 vs the A8-7600 at 65w. anand bench.

I also agree with Yuriman, you can pick some esoteric definition if you like, but in a general scientific sense, "efficiency" in regards to power would be the area under the curve of power consumption vs time to complete the task. How fast the task is finished is another issue. But in any case, it you want to look at performance, the benches I listed show that a modern Haswell i3 slaughters the A8-7600 in all but a few cpu benchmarks. And that is with the A8 at 65 watts, not the 45 watts you picked in your example to make the efficiency look better.

I'm thinking "efficiency" in terms of performance per watt, as fruzentundra said; If you take 25% longer to do a task, but use 20%* less energy per unit of time (resulting in the same total energy used to do the task), the performance per watt is the same. By that metric, "watt-hours to finish the benchmark" in a benchmark that takes variable time to finish should be synonymous with "performance per watt".

Guys, just read what you are typing

You just said it yourself that Efficiency is performance (Time) per watt (Energy).

If you use the same energy but complete the task faster than your opponent you are more efficient. It is not that hard to understand.

or

If you finish the same task at the same time but you are using LESS energy than your opponent you are more efficient.

For the x264 benchmark,

A8-7600 using the same energy to finish the benchmark faster (less time) = A8-7600 is more efficient than Core i3 3225, simple as that.

 

[Yuriman]

Your graphs are deceptive, in that case. The power charts seem to indicate that the i3 uses the same amount of power to do the same amount of work.

If the i3 takes longer, but does it with equally less energy, that would indicate it has the same performance per watt. And, the chart seems to imply that you're measuring two things - total time, and total power.

What are you actually measuring, in this chart?

 

[AtenRa]

Your graphs are deceptive, in that case. The power charts seem to indicate that the i3 uses the same amount of power to do the same amount of work. If the i3 takes longer, but does it with equally less energy, that would indicate it has the same performance per watt. And, the chart seems to imply that you're measuring two things - total time, and total power, and efficiency is the ratio of these two factors.

What are you actually measuring, in this chart?

The above graph display two different things,

On the left is the performance in fps (Frames per second - Higher is better) and on the right we have the energy consumption in Wh (less is better) to finish the benchmark.

By using those two we can calculate the efficiency (perf/watt) or perf/Wh

 

[Yuriman]

Reworded my post to make a bit more sense.

Watts, or watt hours? That is to say, if you had a battery with 80 watt hours of energy in it, and you ran both the i3 and the A8 system on this battery, both would be able to finish the task right as the battery ran out. The i3 system would run longer, but would do the task using the same amount of energy because it uses less energy per time. So, did you mean watts, or watt hours?

Because, 80 watts and 80 watt-hours are very different things.

 

[TheELF]

For the x264 benchmark,

A8-7600 using the same energy to finish the benchmark faster (less time) = A8-7600 is more efficient than Core i3 3225, simple as that.

x264/7zip is also very inefficiently coded which is why you use them as examples instead of winrar or others that actually do proper multithreading.

Why not use itunes to compare efficiency, that's at least something that a user would do on a $100 CPU ,nobody is gonna buy them to do professional video encoding,users are more then ok with quicksync or software like format factory.

 

[TheELF]

Reworded my post to make a bit more sense.

Watts, or watt hours? That is to say, if you had a battery with 80 watt hours of energy in it, and you ran both the i3 and the A8 system on this battery, both would be able to finish the task right as the battery ran out. The i3 system would run longer, but would do the task using the same amount of energy because it uses less energy per time. So, did you mean watts, or watt hours?

Because, 80 watts and 80 watt-hours are very different things.

Are there even that many people that do two or more pass (trans)codings?

 

[AtenRa]

Reworded my post to make a bit more sense.

Watts, or watt hours? That is to say, if you had a battery with 80 watt hours of energy in it, and you ran both the i3 and the A8 system on this battery, both would be able to finish the task right as the battery ran out. The i3 system would run longer, but would do the task using the same amount of energy because it uses less energy per time. So, did you mean watts, or watt hours?

Because, 80 watts and 80 watt-hours are very different things.

The graph clearly says Wh.

Wh = Watts x time

 

[AtenRa]

Ehm lets do it like in school,

Car A Ferrari, needs 10 seconds to run from A to B and it uses 1 Gallon of petrol

Car B Toyota, needs 12 seconds to run from A to B and it uses 1 Gallon of Petrol

Which has higher efficiency ???



Edit: Even easier to understand

Efficiency = Performance / Watt

CPU A produces 100 Frames per second by using 60 Watt

CPU B produces 140 Frames per second by using 60 Watt

Calculate the efficiency of both CPUs.