The Official Kaveri Review Thread (A10-7850K, etc)

[coercitiv]

Look at the red and orange lines where they go up to over 70, run for a period then drops down to 60 where it stays until it finishes up. There rest of the chips are fairly stable (excluding the i7) in terms of draw until the work is finished.

There's one more thing to note about those power graphs: the red line corresponding to A10 6700T is only slightly higher than the orange one corresponding to A8 6500T, and the A10 finishes work faster. (more than 10 seconds faster)

This should translate in both better energy consumption over period and better task energy, yet the A8 figures in the last graph as having the better task energy consumption, which is... impossible.

 

[SViscusi]

There's one more thing to note about those power graphs: the red line corresponding to A10 6700T is only slightly higher than the orange one corresponding to A8 6500T, and the A10 finishes work faster. (more than 10 seconds faster)

This should translate in both better energy consumption over period and better task energy, yet the A8 figures in the last graph as having the better task energy consumption, which is... impossible.

Huh, good catch.

 

[ashetos]

I know all the hot talks are about race to idle and minimum power consumption where Haswell excels etc. Honestly, this is not good enough for me.

Are we close to having a portable machine that can game on battery for hours? I mean, what about battery life at maximum power consumption? I'm wondering if Kaveri will enable something like that with its mobile version, or if we're years behind in this scenario.

 

[Abwx]

I know all the hot talks are about race to idle and minimum power consumption where Haswell excels etc. Honestly, this is not good enough for me.

Are we close to having a portable machine that can game on battery for hours? I mean, what about battery life at maximum power consumption? I'm wondering if Kaveri will enable something like that with its mobile version, or if we're years behind in this scenario.

;;;;

http://wccftech.com/amd-kaveri-apu-...erformance-unveiled-mantle-45-faster-directx/

 

[ashetos]

;;;;

http://wccftech.com/amd-kaveri-apu-...erformance-unveiled-mantle-45-faster-directx/

All these scenarios have tons of idle time/resources. Also, video playback is nothing like gaming.

 

[Sweepr]

Regarding efficiency:

Competing against 14nm Broadwell mobile parts won't be any easier later this year.

 

[ashetos]

Regarding efficiency:

Competing against 14nm Broadwell mobile parts won't be any easier later this year.

To add some context here from the Tomshardware review:
"We can’t script the game tests, so 3D performance isn’t a component in these charts. You’re seeing the result of content creation, media encoding, productivity, and compression workloads."

This is CPU portion only, and race to idle is obviously a major factor here.

Also, intel cores being superior in both architecture and manufacturing process is well known.

Kaveri tries to compete as an APU, it has no hope to compete as a CPU.

 

[Erenhardt]

Regarding efficiency:

Competing against 14nm Broadwell mobile parts won't be any easier later this year.

It doesn;t give even a hint where mobile kaveri will be. Somewhere between intel and mobile richland, but no one knows where.

Binning makes wonders. My 125W TDP phenom can be downvolted to 1,25V (from 1,45 stock?) and uses less than 95W TDP FX6300. Downvolting cut 30 Watts from the wall. On top of downvolting I run it overclocked to 3,5 GHz@ 1,25V. Why it was not a mobile sample I have no idea

 

[Erenhardt]

Actually, if we assume that binning richland for efficiency (A10-6800 to A8-6500T) brings 25% improvement (233-172/233) and apply the same to kaveri then:
A8-7500T should have efficiency around 140Wh (184*0,75).
What is importans aswell is 6500T model is not a mobile chip - it is low power desktop part. I believe those are the chips that were close to be mobile, but not quite there, but are better than average desktop chip.

 

[Slomo4shO]

May be irrelevant but Micro Center has the A10-7850K bundled with BF4 for $130 (in-store only)

 

[Atreidin]

Regarding efficiency:

Competing against 14nm Broadwell mobile parts won't be any easier later this year.

The pedantic part of me that graded physics undergrads is annoyed they say "total power consumed" on a graph with units of Wh.

 

[Natfly]

Any word on when the a8-7600 will be available? Seen the reviews and it looks like it was released with the other kaveri based cpus but no one seems to have any for sale. It's not an oem only part is it?

 

[AtenRa]

Any word on when the a8-7600 will be available? Seen the reviews and it looks like it was released with the other kaveri based cpus but no one seems to have any for sale. It's not an oem only part is it?

It will be released in Q1, not specific date given from AMD. No it is not OEM only.

 

[Natfly]

It will be released in Q1, not specific date given from AMD. No it is not OEM only.

Cool, thanks

 

[USER8000]

The 45W A10 6700T is the same chip as the top Trinity mobile SKU rated at 35W. I suspect it might have failed binning and is anything between a 35W to 45W TDP.

The A8 7600 is not a binned chip for a 45W TDP - it is a 65W TDP chip with a configurable TDP which can be set at a 45W TDP. This means the motherboard is probably just using an offset voltage at a safe level to make sure all of the chips run at that voltage at the set clockspeeds.

It also means binned chips will probably easily hit the 35W TDP spec in mobile.

 

[Phynaz]

The A8 7600 is not a binned chip

All chips are binned. Period.

 

[USER8000]

The 45W A10 6700T is the same chip as the top Trinity mobile SKU rated at 35W. I suspect it might have failed binning and is anything between a 35W to 45W TDP.

The A8 7600 is not a binned chip - it is a 65W TDP chip with a configurable TDP which can be set at a 45W TDP. This means the motherboard is probably just using an offset voltage at a safe level to make sure all of the chips run at that voltage at the set clockspeeds.

It also means binned chips will probably easily hit the 35W TDP spec in mobile.

All chips are binned. Period.

Not binned for a 45W TDP. You know what I was meaning. Plus now edited to include the reference to the 45W TDP.

 

[SPBHM]

http://www.hardcoreware.net/kaveri-7850k-overclocked-benchmarks/2/
some 4.7GHz tests, even with OC still clearly slower than the i3 for some games as a CPU, like 45 to 60fps on AC4, but it's clearly faster for 7zip and POV-Ray

and with the IGP
http://www.hardcoreware.net/amd-kaveri-a10-7850k-apu-overclocked/2/

decent gain here and there, but still to slow for some games and the price they are asking.

but it have some good OC potential...

but Kaveri is really shining on the lower price point and TDP (7600), the 7850K needs a huge price drop.

 

[Erenhardt]

It may or may not be off topic, but I have to say here:
I hate binning. It takes away from getting new hardware. Like what EVGA did with their graphics. 158238 different SKUs, each faster from previous one by 1MHz on core, or memory.
The chip lottery could be a blessing, or big let down, but it was an experience.
Now you get what you pay for.
/rant

http://www.hardcoreware.net/kaveri-7850k-overclocked-benchmarks/2/
some 4.7GHz tests, even with OC still clearly slower than the i3 for some games as a CPU, like 45 to 60fps on AC4, but it's clearly faster for 7zip and POV-Ray

and with the IGP
http://www.hardcoreware.net/amd-kaveri-a10-7850k-apu-overclocked/2/

decent gain here and there, but still to slow for some games and the price they are asking.

but it have some good OC potential...

but Kaveri is really shining on the lower price point and TDP (7600), the 7850K needs a huge price drop.

Why overclocking is is causing fps regression?

 

[SPBHM]

Why overclocking is is causing fps regression?

the cases with a lower average look like within a reasonable margin of error (so the same result, just not gaining anything from the CPU OC), or maybe it throttles, it's unstable, no idea.

 

[ElFenix]

(facepalm) so....answer then: why even put TDP rating on CPUs if power consumption may exceed it ?

did you even read what you post?:biggrin:

what do you think this means? :

TDP is not the maximum power that the processor can dissipate

TDP . Thermal Design Power. The thermal design power is the maximum power a processor can draw for a thermally significant period while running commercially useful software

means that the processor can draw more than TDP, especially when running a "power virus"

 

[Haserath]

I know all the hot talks are about race to idle and minimum power consumption where Haswell excels etc. Honestly, this is not good enough for me.

Are we close to having a portable machine that can game on battery for hours? I mean, what about battery life at maximum power consumption? I'm wondering if Kaveri will enable something like that with its mobile version, or if we're years behind in this scenario.

Years if not decades.

There are several issues with making a portable gaming notebook last a long time on battery.

Battery technology is very limited and expensive to put into laptops. I'm unsure of the actual prices on batteries, but I think it's about $1 per Wh, maybe a little less. Then there's the volume, weight, heat, and durability issues they cause.

35W chips barely play modern demanding games with settings turned way down. Laptops generally only have 60Wh battery capacity, and 100Wh is almost never exceeded. So a system that uses 40W will only last at best 2 1/2 hours.

The gaming industry keeps putting more and more into graphics. Even with process improvements, the chips have to be made faster to keep up, so we aren't seeing power improvements. Valve is actually the only major gaming company that I know of that has gone on record as saying they think graphics are good enough(as in most people don't care for more).

One of the big innovations in battery life would be software. Mantle sounds very interesting because it does more with the same hardware. Hardware improvements are harder and harder to come by.

 

[know of fence]

So I hear Kaveri APUs aren't soldered either, but use some kind of white grease. Not sure, how reliable that source is, maybe someone with Japanese skills can help. It didn't go quite as planned either.

bit.ly/1evE2sS

 

[ashetos]

Years if not decades.

There are several issues with making a portable gaming notebook last a long time on battery.

Battery technology is very limited and expensive to put into laptops. I'm unsure of the actual prices on batteries, but I think it's about $1 per Wh, maybe a little less. Then there's the volume, weight, heat, and durability issues they cause.

35W chips barely play modern demanding games with settings turned way down. Laptops generally only have 60Wh battery capacity, and 100Wh is almost never exceeded. So a system that uses 40W will only last at best 2 1/2 hours.

The gaming industry keeps putting more and more into graphics. Even with process improvements, the chips have to be made faster to keep up, so we aren't seeing power improvements. Valve is actually the only major gaming company that I know of that has gone on record as saying they think graphics are good enough(as in most people don't care for more).

One of the big innovations in battery life would be software. Mantle sounds very interesting because it does more with the same hardware. Hardware improvements are harder and harder to come by.

Sounds pretty disapointing...

So with linear calculations like you did, we would need something like a capable system that draws only 20 Watts (monitor + wifi + sound + APU + storage) and a super light high tech battery that provides 80 Wh for only 4 hours of gaming...

Maybe we need a breakthrough with battery technology.

 

[Abwx]

Maybe we need a breakthrough with battery technology.

Two masses of electrons of one gramme at a distance of one metre have a repulsive force that is equal to the weight of a whole earth weighted at one metre of the ground, this say the difficulty , huge , to keep charges separated and thus have high capacities.