Energy-efficient AMD A10-6700T APU Goes on Sale This Week(TPU)

[AtenRa]

http://www.techpowerup.com/189355/energy-efficient-amd-a10-6700t-apu-goes-on-sale-this-week.html

"AMD's energy-efficient socket FM2 desktop APU, the A10-6700T, is expected to go on sale later this week. Its selling point is the 45W TDP, with no reduction in core-count, or other components. Based on the 32 nm "Richland" silicon, the A10-6700T features four "Piledriver" x86-64 cores clocked at 2.50 GHz, with TurboCore frequency of 3.50 GHz, and 4 MB of total L2 cache. It features a Radeon HD 8650D graphics core with 384 stream processors, and untouched clock speeds of 760 MHz core, and 844 MHz boost. The chip also features a dual-channel DDR3-2133 MHz integrated memory controller, and a PCI-Express gen 2.0 root complex. The A10-6700T is expected to be priced around $150."

 

[SPBHM]

45w is nice but, as low as 2.5GHz and $150 not so much, I wonder if it's actually going to use less power than a regular Haswell i3.

 

[Abwx]

45w is nice but, as low as 2.5GHz and $150 not so much, I wonder if it's actually going to use less power than a regular Haswell i3.

It will use 45W.

 

[AtenRa]

With 2133MHz memory and 760MHz for the iGPU it will perform almost like a A10-5800K in games (using the iGPU only) and that means it will still be faster than Haswell Core i3. And only with a 45W TDP and all of the Ritchland features, thats one hell of a HTPC/light game APU.

 

[SPBHM]

It will use 45W.

I'm talking about real world and power usage.

With 2133MHz memory and 760MHz for the iGPU it will perform almost like a A10-5800K in games (using the iGPU only) and that means it will still be faster than Haswell Core i3. And only with a 45W TDP and all of the Ritchland features, thats one hell of a HTPC/light game APU.

i3 for $135 will have a 54W TDP, 3.4GHz haswell cores and GT2 (20EUs), I'm pretty sure it's going to be faster for games where the 2.5GHz CPU is a problem...

and again, I'm not convinced this 45W richland is really going to use less power

 

[rainy]

45w is nice but, as low as 2.5GHz and $150 not so much, I wonder if it's actually going to use less power than a regular Haswell i3.

I'm just curious: it's A10-6800K with significant lower clock on CPU side (slightly lower GPU) or modified A10-5750M with support of DDR3 2133 MHz?

Btw, I agree about price - 110-120 would be OK in my opinion.

 

[Abwx]

I'm talking about real world and power usage.

i3 for $135 will have a 54W TDP, 3.4GHz haswell cores and GT2 (20EUs), I'm pretty sure it's going to be faster for games where the 2.5GHz CPU is a problem...

and again, I'm not convinced this 45W richland is really going to use less power

A 65W richland effectively use 65W since the whole
plateform use 91W when loaded and whatever
what intel does , unless your concern is that
it consume less than the competition CPUs
you re quoting i dont know why since this give
no clue about AMD s products caracteristics...

http://www.cowcotland.com/articles/1411-8/test-processeur-amd-a10-6700.html

Why would they release a 45W that use as much as their
65W model.?.

 

[ElFenix]

following intel's terrible practice of reusing model numbers for a lower performing low power model.

 

[SPBHM]

A 65W richland effectively use 65W since the whole
plateform use 91W when loaded and whatever
what intel does , unless your concern is that
it consume less than the competition CPUs
you re quoting i dont know why since this give
no clue about AMD s products caracteristics...

http://www.cowcotland.com/articles/1411-8/test-processeur-amd-a10-6700.html

Why would they release a 45W that use as much as their
65W model.?.

65w richland vs 55w ivy

(probably unfair because of how slow the hd2500 is)

but this should be fair

so I'm not conviced 45w richland can use less power than 54W haswell, particularly with the IGP+CPU loaded.

 

[Abwx]

65w richland vs 55w ivy

(probably unfair because of how slow the hd2500 is)

but this should be fair

so I'm not conviced 45w richland can use less power than 54W haswell, particularly with the IGP+CPU loaded.

This THG review is bogus and dishonnest...

Would an honnest site write this :

A10 6700 exhibits an impressive 25 W drop compared to AMD's A10-5800K in our Metro: Last Light benchmark. But that's hardly an achievement next to the 61 W Intel s core i3

Despite this :

Since they used Metro to test game PWR comsumption ,
so basicaly 90% more comsumption for 150% more perfs....


But let s look at the web browsing comsumption in detail :

What are theses spikes.?..

Why the other AMD CPUs need to use their full CPU capability
despite being more powerfull at stocks than the 6700 to simply
browse internet pages.?..

They used a 1200W PSU and one cant even be sure
that they didnt recycle numbers for the i3 using
another PSU in previous reviews where they didnt
even specify its comsumption.

Edit : What makes me dubious of THG honnesty
is that they didnt show the iddle power but
Hardware.fr test show the intel i3 3240
iddling at 42W wich is more than the THG
test with a loaded CPU.

 

[Topweasel]

I don't know Intel didn't lower the TDP really with Haswell. Whereas AMD cut off almost 20w.

I know what you are trying to establish and I know why you are stating it. But you are using a higher TDP chip then it's competitor and how it lost in benchmarks for power consumption and then using those results to establish that you doubt AMD could use less power with a TDP that is lower then the competitor. Sure the power usage is substantially lower on the I3 then its max compared to the Richland part. But now you are talking 20w drop. Even if the draw during basic stuff is the 45w (which it won't be) we are still only talking about a handful of w difference.

Also it's not fair to compare GPU loaded circumstances. When a GPU craps out like a 2500 would in those tests the CPU's are going to rest for many a cycle. It wasn't going 100% across all cores and GPU's because the CPU is going to wait for the GPU to finish what it has before sending it more.

Now I am not so sure on this but I have seen enough mobo reviews in my life to know there can be some drastic differences in CPU power usage dependent on the board. Is the board in the review for the I3 a good or bad one for power, and how about the Richland board? I pretty much stopped reading Tom's back in the day when they ran a stacked test against AMD with their week long performance and stability, which the AMD one was doing fantastic, the Intel kept crashing, died, was operated on for 2 days, several new CPU's/mobo's/memory later, they got it back up and running. They decided to "reset" the test then, ran 1 more application so the Intel system looked better slightly due to HT. Then at the end glossed over the Intel platform issues (blaming parts providers), called it the winner because while it was dreadfully behind in 2 of the applications, it accomplished more on the third application they started purely because they knew it would either take a significant chunk out of the other two processes or it wouldn't get much done. After that I couldn't take them seriously. So much had changed in such a little time on that site.

Edit: This was a response to SPBHM

 

[el etro]

This could be real(the power consumption of Richlands when web-browsing) given the low IPC Piledriver has(I prefer too Anand's reviews especially when covering GPUs). AMD 32nm(the GloFo Process is very good) vs tri-gate Intel 22nm(Intel manufacturing processes is better ones I know) makes an unfair architecture competition for the eficiency run(ok, we already know intel's make more perf/transistor, especially when with HT).
Power management of Richland deserves congratulations. Will be good to see this litle improvement on Kaveri next year.
According to benches i see, Kaveri's GPU will be up to 300% faster than a10 5800K GPU.

 

[monstercameron]

This could be real(the power consumption of Richlands when web-browsing) given the low IPC Piledriver has(I prefer too Anand's reviews especially when covering GPUs). AMD 32nm(the GloFo Process is very good) vs tri-gate Intel 22nm(Intel manufacturing processes is better ones I know) makes an unfair architecture competition for the eficiency run(ok, we already know intel's make more perf/transistor, especially when with HT).
Power management of Richland deserves congratulations. Will be good to see this litle improvement on Kaveri next year.
According to benches i see, Kaveri's GPU will be up to 300% faster than a10 5800K GPU.

there is no way that is accurate.

 

[Abwx]

The PWR graph show that AMD APUs load climb to at least 50%
for the 4 cores or 100% for two cores while the i3 seems loaded
at a rate of no more than 50% of a single core , that would be
logical if an i3 core had 4 times the IPC of an AMD core....

 

[frozentundra123456]

there is no way that is accurate.

Yea, I was going to say "what benches?".

 

[el etro]

Yea, I was going to say "what benches?".

7730 review(20% less memory dependent plus 10% AVG more perf/CU) and Kabini IGP(1/4 CUs of Kaveri IGP from top SKU) vs A10 5800k. I made my calculations based on this.

 

[frozentundra123456]

7730 review(20% less memory dependent plus 10% AVG more perf/CU) and Kabini IGP(1/4 CUs of Kaveri IGP from top SKU) vs A10 5800k. I made my calculations based on this.

Kaveri is supposed to have the same number of shaders as a 7750, and a 7750 is about twice as fast as the A10. So max performance could be twice A10. But that is a gddr5 7750. A DDR 3 7750 loses about 50% in performance, and clocks will probably be lower in the apu.

So I would say 50% absolute Max performance over the A10, and probably something more like 30 to 40 percent.

 

[monstercameron]

Kaveri is supposed to have the same number of shaders as a 7750, and a 7750 is about twice as fast as the A10. So max performance could be twice A10. But that is a gddr5 7750. A DDR 3 7750 loses about 50% in performance, and clocks will probably be lower in the apu.

So I would say 50% absolute Max performance over the A10, and probably something more like 30 to 40 percent.

your speculation seems more plausible. makes me wonder if just maybe, AMD has some kinda bandwidth cure up their sleeves for kaveri...

 

[Abwx]

your speculation seems more plausible. makes me wonder if just maybe, AMD has some kinda bandwidth cure up their sleeves for kaveri...

CPU + dGFX :

RAM --> CPU cache --> RAM --> GPU RAM --> GPU --> GPU RAM --> RAM


Kaveri , with hUMA :

RAM---> CPU cache --> RAM --> GPU --> RAM

 

[el etro]

Kaveri is supposed to have the same number of shaders as a 7750, and a 7750 is about twice as fast as the A10. So max performance could be twice A10. But that is a gddr5 7750. A DDR 3 7750 loses about 50% in performance, and clocks will probably be lower in the apu.

So I would say 50% absolute Max performance over the A10, and probably something more like 30 to 40 percent.

I admit i think make my calculations a little wrong, but the performance of Kaveri IGP may be higher than this. Hoping they reduce the CPU bottleneck at low resolutions to turn all games playable.

 

[beginner99]

makes me wonder if just maybe, AMD has some kinda bandwidth cure up their sleeves for kaveri...

GDDR5 memory controller.

 

[Gikaseixas]

Nice little chip for people who don't want to undervolt. On the negative side, once loaded the cores and gpu will most likely consume the same amount of power as the regular A10-6700.
IMO it's almost perfect for HTPC use.

 

[ShintaiDK]

GDDR5 memory controller.

Nope.

 

[inf64]

4x64bit for quad channel DDR3? Also GDDR5 could still be supported in HW even tho AMD decided against official support- earlier SOGs say Kaveri supports it .

 

[ShintaiDK]

4x64bit for quad channel DDR3? Also GDDR5 could still be supported in HW even tho AMD decided against official support- earlier SOGs say Kaveri supports it .

No quadchannel either.