[AT] AMD A10-7800 Review: Testing the A10 65W Kaveri

[Essence_of_War]

http://www.anandtech.com/print/8291/amd-a10-7800-review-65w-kaveri

Gives the 95W A10-7850k a pretty good run for its money at a lower TDP.

There didn't appear to be any actual platform power consumption benches though?

 

[ShintaiDK]

There is some powernumbers here:
http://www.bit-tech.net/hardware/2014/07/31/amd-a-10-7800-review/8

And some here:
http://www.techspot.com/review/856-amd-a10-7800-kaveri/page8.html

 

[Lyfer]

Good CPU for htpc at 65w.

 

[Fjodor2001]

More power consumption tests here:

http://hexus.net/tech/reviews/cpu/72709-amd-a10-7800-28nm-kaveri/?page=8

 

[rainy]

Good CPU for htpc at 65w.

I would say that A8-7600 at 105 dollars (A10-7800 - 155) could be even better.
http://www.extremetech.com/computin...us-better-pricing-and-lower-power-consumption

 

[frozentundra123456]

Pretty superficial test actually. Only one demanding game was tested, no i5 was used in the cpu comparison, and no comparison to a cheap cpu and a discrete card. The reviewer did not even test any of the games included A10 bundle. Would have been nice to at least see how the bundled games would have fared with the A10.

As lyfer said, the best choice for a SFF light gaming box perhaps, but overall seems like a niche product. Still too expensive as well. Should be 120.00.

 

[Fjodor2001]

38 vs 31 W at idle for the A10-7800. Just shows how much impact the motherboard and the rest of the test setup has on the idle power consumption.

 

[ShintaiDK]

38 vs 31 W at idle for the A10-7800. Just shows how much impact the motherboard and the rest of the test setup has on the idle power consumption.

Even hexus doesnt compare apples to Apples on the FM2+ platform. Look in their test setup. Different board, different drivers

The lazyness of review sites is astonishing.

 

[NTMBK]

Seems like a good part for a gaming HTPC.

 

[Erenhardt]

A8-7600 is stealing the show!
Also, not a fan of the premium for K chips ;/

 

[Essence_of_War]

Seems like a good part for a gaming HTPC.

Yeah, potentially?

1280x1024 is certainly more pixels than 1280x720 (which is as low as I'm interested in going on a HTPC situation), and the actual FPS are reasonably good, but most living room TVs are natively 1920x1080, which is ~60% more pixels. I think the APUs would have a hard time keeping up at native res.

 

[DrMrLordX]

Those Hexus power consumption numbers seem odd. The 7850K is turning in lower power consumption at 2D load and 3D load than the 65W 7800.

Could this be the 7850K doing the infamous "throttle to 3 ghz" routine while the 7800 is not?

 

[ShintaiDK]

Those Hexus power consumption numbers seem odd. The 7850K is turning in lower power consumption at 2D load and 3D load than the 65W 7800.

Could this be the 7850K doing the infamous "throttle to 3 ghz" routine while the 7800 is not?

Looked at this?

http://hexus.net/tech/reviews/cpu/72709-amd-a10-7800-28nm-kaveri/?page=2

 

[Qwertilot]

Could be more than that though. I mean you can see how incredibly close the two come performance wise on the Anand tech benchmarks. Identical for some tests and never really more than a very small lead overall.

Frankly if its a 'real' 65w vs 95w you'd want to know what they've done with extra 30w of power budget Maybe it is just that the overall design is much happier at a 45/65w power draw. Maybe bandwidth issues throttling things, maybe its turbos everywhere, goodness knows.

Quoted power draw never the most meaningful thing in the world of course

 

[sm625]

A 34 watt increase at the wall vs a 54W TDP i3. Shocking. All that extra power for less performance. I really want to see the data showing a G3258 + discrete outperforming kaveri in Perf/watt. I am confident it will but have yet to see the data.

 

[Roland00Address]

Numbers From Anandtech

What is the point of the 65w AMD A10 - 7800 for $155 dollars vs 65w AMD A8 - 7600 for $105 again? What are you gaining for those $50 dollars, not much on the gpu even though it has 33% more calculation units.

Gaming numbers (can you say bottlenecked)

Difference from A10 vs A8 in % --- A10 vs A8 Average FPS all games at 1280x1024

2.22% --- 78.3 vs 76.6 F1 2013 (Processor Graphics)

2.75% --- 67.2 vs 65.4 Bioshock Infinite (Processor Graphics)

9.78% --- 66.2 vs 60.3 Tomb Raider (Processor Graphics)

5.14% --- 59.3 vs 56.4 Sleeping Dogs (Processor Graphics)

-3.08% --- 22 vs 22.7 Company of Heroes 2 (Processor Graphics)


-----

Can AMD get some newer fabs as soon as possible, that and some high speed memory bandwidth technology such as stacked DRAM. You can have the best engineers and the best design ever, but if the tools you have to make it suck this bad, then what is the point?

What an utter waste of die space.

 

[Enigmoid]

A 34 watt increase at the wall vs a 54W TDP i3. Shocking. All that extra power for less performance. I really want to see the data showing a G3258 + discrete outperforming kaveri in Perf/watt. I am confident it will but have yet to see the data.

http://www.anandtech.com/show/7003/the-haswell-review-intel-core-i74770k-i54560k-tested/2

http://www.anandtech.com/bench/product/1223?vs=836

Sad to say but for pure CPU performance HW can be as efficient or better.
Despite a massive 34W idle, the 4770k uses 113W under load and gets 46.7 fps x264 second pass while the 5350 gets 11.1 fps. HW is 4.21x faster; for the 5350 to be more efficient it would have to be using less than 26.8W.

I am pretty sure that the AT HW system is also including the idle power of a dgpu, TR x264 pass 2 is around 87W on igp, 102 without (more efficient mobo, different dgpu). Using the igp numbers, the 5350 would have to consume about 21W or less.

http://techreport.com/review/24879/intel-core-i7-4770k-and-4950hq-haswell-processors-reviewed/7

 

[SPBHM]

Numbers From Anandtech

What is the point of the 65w AMD A10 - 7800 for $155 dollars vs 65w AMD A8 - 7600 for $105 again? What are you gaining for those $50 dollars, not much on the gpu even though it has 33% more calculation units.

Gaming numbers (can you say bottlenecked)

Difference from A10 vs A8 in % --- A10 vs A8 Average FPS all games at 1280x1024

2.22% --- 78.3 vs 76.6 F1 2013 (Processor Graphics)

2.75% --- 67.2 vs 65.4 Bioshock Infinite (Processor Graphics)

9.78% --- 66.2 vs 60.3 Tomb Raider (Processor Graphics)

5.14% --- 59.3 vs 56.4 Sleeping Dogs (Processor Graphics)

-3.08% --- 22 vs 22.7 Company of Heroes 2 (Processor Graphics)

and it can be worse, Anandtech used DDR3 2133, a lot of people will use 1600 or 1866, making the performance difference probably less noticeable

 

[ShintaiDK]

Another review:
http://www.hardwarecanucks.com/foru...ews/67043-amd-a10-7800-kaveri-apu-review.html

Almost same performance, but also almost same power consumption as the 7850K.

 

[Ajay]

Well, as long as AMD is stuck @ 28nm, we just aren't going to see even good power numbers. 20nm, if it doesn't totally suck (as some rumors have suggested) will help, but I don't expect anything impressive till 14FF (where Intel will still dominate since it can leverage more design $$s to improve actual xtors/mm2 on silicon, vs best case numbers derived from the PDK).

AMD is fighting the good fight, but what are they spending on x86 development compared to Intel? As the node sides get smaller, design expenditures become even more important.

 

[know of fence]

Numbers From Anandtech

What is the point of the 65w AMD A10 - 7800 for $155 dollars vs 65w AMD A8 - 7600 for $105 again? What are you gaining for those $50 dollars, not much on the gpu even though it has 33% more calculation units.
Gaming numbers (can you say bottlenecked)
What an utter waste of die space.

This is a good observation, but TDP is just the maximum threshold. Having more steam processors even though they are bottlenecked by both the CPU and DDR3 bandwidth, ideally means that you can tweak them to run at a lower and thus more power efficient frequency. Basically you get the same performance as the A8 at lower power consumption.

Also every system will be bottlenecked by something, current high end cards are also bandwidth limited, so it barely makes sense to OC the top GPU, though a GDDR5 OC should give them a measurable boost at high resolutions/with AA cranked up.

The tragedy of this HTPC dream however is that there aren't any 80+ certified 200W power supplies, and all these low idle savings just heat up an over-sized or crappy PSU. If only there was a company that sold an appropriately sized quality case+power brick, we could have an all purpose PC that idles well below any CPU+dGPU combination aside from Optimus laptops.
Also consoles have about twice as many stream processors along with faster memory, but going down a notch and playing in 720p is perfectly serviceable even letterboxed on a 16x10 monitor.

 

[ElFenix]

wonder how it compares to the 7700k

 

[cbn]

I would say that A8-7600 at 105 dollars (A10-7800 - 155) could be even better.
http://www.extremetech.com/computin...us-better-pricing-and-lower-power-consumption

Can the iGPU on the A8-7600 be overclocked?

 

[monstercameron]

Can the iGPU on the A8-7600 be overclocked?

k indicates unlocked parts. so most likely not, maybe in amd overdrive.

 

[cbn]

k indicates unlocked parts. so most likely not, maybe in amd overdrive.

Thanks. I knew the cpu could not be overclocked.....and maybe the GPU doesn't need to be since it has 384 GCN stream processors at 720 Mhz.

With that mentioned, I would be very interested to see how A8-7600 compares to A8-6600K which is a $99 processor.

Unlocked quad core (A8-6600K) with 256 VLIW4 stream processors (overclockable)

vs

Locked quad core (A8-7600) and 384 GCN stream processors