Kabini Rumors

[Arachnotronic]

Thanks for the link. Finally some hard data when it comes to power:

Not bad! I wonder how it will compete with Bay Trail-M, which is supposed to run up to 2.7GHz and have an SoC TDP of 4W - 6.5W

 

[Vesku]

Google Chrome Flash playback: 100% CPU in use, some frame drops.
Google Chrome HTML5 playback: Approx 20% CPU used by Chrome processes
IE10 HTML5: 5-7% CPU
IE 10 Flash: Approx 10%


Darn you FLAAAAAAAAASH! Wouldn't be so bad if it was more consistent, shouldn't have to run IE to get good optimizations.

Not bad! I wonder how it will compete with Bay Trail-M, which is supposed to run up to 2.7GHz and have an SoC TDP of 4W - 6.5W

We'll find out next year.

 

[Arachnotronic]

Google Chrome Flash playback: 100% CPU in use, some frame drops.
Google Chrome HTML5 playback: Approx 20% CPU used by Chrome processes
IE10 HTML5: 5-7% CPU
IE 10 Flash: Approx 10%


Darn you FLAAAAAAAAASH! Wouldn't be so bad if it was more consistent, shouldn't have to run IE to get good optimizations.



We'll find out next year.

Actually, we'll find out in holiday 2013

 

[Arkadrel]

Test done on high-performance profile under battery power.
Test 2 – ALU
Standing usage immediately before test: 7.9W (7% utilization, ave 1Ghz clock in quiescent state)
Test max power drain (ALU Test): 12.3W (94% CPU load by Crystalmark Process)
Test power: 4.4W
Normalised: 790mw/Core/Ghz
Test 2 – FPU
Standing usage immediately before test: 8.1W (7% utilization, ave 1Ghz clock in quiescent state)
Test max power drain (FPU Test): 13.1W (94% CPU load by Crystalmark Process)
Test power: 5W
Normalised: 900mw/Core/Ghz

Estimated max power usage of 4 Temash CPU-cores at 1Ghz, 100%: 3.6W (+/- 10%)

Estimated max power usage of 4 Temash CPU-cores at 1.4Ghz, 100%: 5W (+/- 10%)

That is better than I thought it would be, esp at 1,4ghz.
>> A 5w cpu is scoreing like 1.39 in cinebench 11,5. <<

Makes the AMD A6-1450 look pretty good.

Not bad! I wonder how it will compete with Bay Trail-M, which is supposed to run up to 2.7GHz and have an SoC TDP of 4W - 6.5W

The graph shows D&M (D being 10w) go upto and perphaps over 2,7ghz.
Maybe at 4w-6.5w its only does half that?

Even so a potential ~1,8-2ghz baytrail M will probably slaughter a 1.4ghz Jaguar quad, and use less power because of the fab advantages Intel has.

Still those are 6-8months away right?

 

[Exophase]

Similar laptops, even other model Acer Aspire v5s with 11.6" screen, HDDs and 1.5V DDR3 use substantially less power at idle (http://www.anandtech.com/show/6467/...view-the-death-and-rebirth-of-the-netbook/4); it's about 6W vs 8W. Which makes me seriously question the methodology here of determining SoC power consumption by subtracting peak from idle. Even AMD says that Temash uses 1.2W at idle, if not in its deepest sleep.

 

[monstercameron]

Power consumption aside it seems like the battery is quite small. From amazon listing it says it has a 3-cell 2650mah, so assuming 7.4v, it has a 20whr battery - unless I have messed up somewhere - and It states 3.5 hour battery life.

http://www.amazon.com/Acer-Aspire-V5-122P-0600-11-6-Inch-Touchscreen/dp/B00CM1AAP0/ref=amtcd_B00CM1AANW_B00CM1AAP0

 

[Arkadrel]

Power consumption aside it seems like the battery is quite small. From amazon listing it says it has a 3-cell 2650mah, so assuming 7.4v, it has a 20whr battery - unless I have messed up somewhere - and It states 3.5 hour battery life.

http://www.amazon.com/Acer-Aspire-V5-122P-0600-11-6-Inch-Touchscreen/dp/B00CM1AAP0/ref=amtcd_B00CM1AANW_B00CM1AAP0

Yep, its disappointing they waste production costs on weirdly named thunderbolt ports, instead of spending a bit extra on battery.

Hopefully others are smarter, and have versions with 6cell battery and no thunderbolt ports to cut costs if needed. Same with that back light keyboard crap lmao.

If your doing low cost devices, prioritizing where you spend production costs should be important.

 

[monstercameron]

Unless the battery was at 11.6v, then it would be 30whr...so maybe I did it wrong afterall

 

[mikk]

Thanks for the link. Finally some hard data when it comes to power:

Update:

Update 13:40 13 May 2013: Test 3 – Prime95 Torture Test (Blend – includes memory testing)
Standing usage immediately before test: 6.7W (7% utilization, ave 1Ghz clock in quiescent state)
System max power drain during test : 17.8W (91% CPU load by Prime95 Process)
Test power: 11.1W

Test 4 – Prime95 Torture Test (Small FFT – minimal memory usage)
Standing usage immediately before test: 7.2W (7% utilization, ave 1Ghz clock in quiescent state)
Test max power drain : 15.8W (91% CPU load by Prime95 Process)
Test power: 8.6W
Normalised: 1700mW / Core / Ghz

Test5 – Prime95 Toture (Blend) + Cinebench OPenGL
Standing usage immediately before test: 7.2W (7% utilization, ave 1Ghz clock in quiescent state)
System max power drain during test : 21.1W
Test power: 13.9 (CPU, GPU, Memory)

Another test:

Leistungsaufnahme


Leerlauf: (Messungen auf Windows-8-Desktop)

• Energiesparmodus, minimale Helligkeit, WLAN aus: 5,3 Watt
• Ausbalanciert, maximale Helligkeit, WLAN aus: 6,8 Watt
• Höchstleistung, maximale Helligkeit, WLAN an: 10,3 Watt

Last: (Messungen mit Höchstleistung, maximaler Helligkeit und WLAN an)

• Cinebench R11.5 Single (~1,4 GHz CPU): 13,7 Watt
• Cinebench R11.5 Multi (~1,1 GHz CPU): 14,9 Watt
• Prime95 large FFTs (~1,0 GHz CPU): 16,9 Watt
• FurMark (~1,2 GHz CPU): 21,2 Watt
• Prime95 + FurMark (~0,85 GHz CPU): 21,9 Watt
• 3DMark 06: 19,5 Watt

http://www.notebookcheck.com/Im-Test-AMD-A6-1450-APU-Temash.92206.0.html

 

[inf64]

The power virus tests are pointless for devices like that . It's good to see them but they have zero relevance for usage of the chip like that.

 

[strata8]

http://www.notebookcheck.com/Im-Test-AMD-A6-1450-APU-Temash.92206.0.html

"Despite the low price of around 450 euros Acer has a high-quality IPS display with touch screen installed, which dissolves with 1366 x 768 pixels."

Awesome! I've been waiting for cheap laptops with decent screens.

edit: Going by the review the A6-1450 has similar single-threaded CPU performance than the E2-1800 and much better multi-threaded and GPU performance. Not bad for a chip that (hopefully) has a much lower TDP.

 

[Cerb]

Power consumption aside it seems like the battery is quite small. From amazon listing it says it has a 3-cell 2650mah, so assuming 7.4v, it has a 20whr battery - unless I have messed up somewhere - and It states 3.5 hour battery life.

3-cell in series is a nominal 11.1V, so that's more like 30-35Wh. Not impressive, but that's how lightweight ones go. A 6-cell option, or better, a machine designed for a 6-cell, would be good for all-day use, without harming the battery's service life (much like Atom, Zacate, or ULV Cores).

Given that AMD's performance estimates, compared to Bobcat, appear dead on, on, and power looks alright...what's the pricing on these like?

 

[monstercameron]

Power draw under max load is a bit high, also the system seems to be a huge addition to the power draw.

Min brightness and WiFi off is 5.3w
Max brightness and WiFi on is 10.3w

How bright is that display?

 

[Olikan]

Power draw under max load is a bit high, also the system seems to be a huge addition to the power draw.

Min brightness and WiFi off is 5.3w
Max brightness and WiFi on is 10.3w

we are really reaching the point where CPU power usage is baraly relevant

 

[mikk]

The power virus tests are pointless for devices like that . It's good to see them but they have zero relevance for usage of the chip like that.

They are not pointless for gaming. With GPU+CPU usage power draw is much higher than with CPU load alone. Such tests are important to reveal throttling or heat issues. In this notebookcheck review the CPU throttled for whatever reason. Heat or TDP issue. Gaming tests don't look good.

 

[monstercameron]

They are not pointless for gaming. With GPU+CPU usage power draw is much higher than with CPU load alone. Such tests are important to reveal throttling or heat issues. In this notebookcheck review the CPU throttled for whatever reason. Heat or TDP issue. Gaming tests don't look good.

Chippy's was able to clock it up to 1.4ghz and have the GPU max out when plugged in during his love stream, maybe there is a trick to it.

 

[Arkadrel]

Given that AMD's performance estimates, compared to Bobcat, appear dead on, on, and power looks alright...what's the pricing on these like?

I believe a fudzilla article mentioned 40-70$ range or so.

The A6-1450 is probably closer to 40$ mark.

we are really reaching the point where CPU power usage is baraly relevant

Power usage of screen and HDD, odd things like added thunderbolt chipsets ect... it all adds up.

 

[Cerb]

I believe a fudzilla article mentioned 40-70$ range or so.

I was thinking pricing for the whole computers.

 

[monstercameron]

I was thinking pricing for the whole computers.

up to $550

 

[Khato]

http://www.notebookcheck.com/Im-Test-AMD-A6-1450-APU-Temash.92206.0.html

Hooray for an actual good review! The numbers look to somewhat line up with the earlier information, but it's actually well presented and properly conducted. Which means that numbers can safely be compared to their other reviews.

Main point of interest being the system power usage. From which we can derive the dynamic load power usage:

Cinebench R11.5 single (~ 1.4 GHz CPU ): 3.4 Watt
Cinebench R11.5 multi (~ 1.1 GHz CPU ): 4.6 Watt
Prime95 large FFTs (~ 1.0 GHz CPU ): 6.6 Watt
FurMark (~ 1.2 GHz CPU ): 10.9 Watt
Prime95 + Furmark (~ 0.85 GHz CPU ): 11.6 Watt
3DMark 06: 9.2 watts

Rather much as expected - for pure CPU loads the power draw is reasonable and kept in check by dropping down towards the base frequency of 1GHz. However soon as the GPU is added into the equation power draw goes way beyond the ~4.5W target of a tablet SoC. Sure it has the performance to merit that power usage, but that doesn't change the fact that it's way too high. Will be interesting to see how it performs in a tablet power budget.

 

[Exophase]

Main point of interest being the system power usage. From which we can derive the dynamic load power usage

The difference between balanced mode w/wifi off and maximum performance mode w/wifi on is 3.5W. That's wifi on but idle. It's not going to use 3.5W, it's not even going to use a tiny fraction of it. The only explanation is that moving away from balanced mode is preventing the CPU from entering as low power of a state, which means that it's probably not a good point to use as a zero for SoC power consumption. Even the other modes are probably not really close to zero.

I expect that keeping it at maximum performance could also be preventing unused cores from being power gated, which would skew the single threaded test closer to the peak.

It's weird that it only lets you go up to 1-1.1GHz with all the cores pegged but GPU not used, when using the GPU exposes a much higher power ceiling. I wonder what's regulating this.

 

[Khato]

The difference between balanced mode w/wifi off and maximum performance mode w/wifi on is 3.5W. That's wifi on but idle. It's not going to use 3.5W, it's not even going to use a tiny fraction of it. The only explanation is that moving away from balanced mode is preventing the CPU from entering as low power of a state, which means that it's probably not a good point to use as a zero for SoC power consumption. Even the other modes are probably not really close to zero.

I expect that keeping it at maximum performance could also be preventing unused cores from being power gated, which would skew the single threaded test closer to the peak.

Correct, going from balanced with WiFi off to high performance with WiFi on increases power usage by 3.5W. But why would using that highest idle power consumption as a baseline for determining load dynamic power be non-ideal? Other than the fact that both the performance measured earlier in the review and the corresponding dynamic power would likely decrease if using the balanced profile instead. I will agree that it seems a bit odd for the single-threaded cinebench at 1.4 GHz to show a dynamic power usage of 3.4W while the multi-threaded cinebench at 1.1 GHz only bumps up to 4.6W... but yeah, can only speculate as to why that might be. It can't just be an issue with power gating because that would push up the baseline as well. And it doesn't make sense that much higher voltage would be needed for operation at 1.4 GHz since I'd expect Temash to be the exact same die as Kabini which can operate at far higher frequencies.

It's weird that it only lets you go up to 1-1.1GHz with all the cores pegged but GPU not used, when using the GPU exposes a much higher power ceiling. I wonder what's regulating this.

Agreed. My guess is that the power profiles on this particular product aren't exactly implemented correctly. Since it certainly seems like maximum performance mode is allowing the GPU to run rampant while still reigning in the CPU.

 

[inf64]

Note from the notebookreview:

Not without reason why AMD has integrated a turbo (dock) mode, which can raise the clock speed to 100 MHz increments up to 1.4 GHz. To enable this, the notebook must (whether on battery or AC power) "Balanced" or "High performance" power profile in the run. Unfortunately, the turbo turned out in our measurements are only partially effective: the full 1.4 GHz processor reaches only single-threaded load, with each additional thread decreases the average frequency by about 100 MHz. We address this point deliberately from an average, since all four cores constantly fluctuate between base and turbo clock rate. Absence of other test equipment we currently can not tell if this is a special feature of the Aspire V5-122, or the general behavior of the A6-1450 APU - 1.4GHz at constant performance would be significantly better TDP -. Temperature or conditional Downclocking appears to us in pure CPU load but not plausible (see Power consumption / temperature).

The clock in MTed workloads is ~1Ghz and max Turbo is achieved only in ST tasks. When more threads become active Turbo drops down in 100Mhz intervals (per active thread). So take this in consideration when viewing the MTed benchmark results.

 

[Exophase]

Correct, going from balanced with WiFi off to high performance with WiFi on increases power usage by 3.5W. But why would using that highest idle power consumption as a baseline for determining load dynamic power be non-ideal? Other than the fact that both the performance measured earlier in the review and the corresponding dynamic power would likely decrease if using the balanced profile instead.

It depends what you actually intend to be measuring with load dynamic power. If the idle measurement doesn't actually have the CPUs at their lowest power states then should it be considered to be the start of the dynamic range instead of part way into it? We don't even know what's causing the difference but it's pretty safe to assume that it's at least partially due to the CPU not being driven to as low states as it could be, and in that case why should that not be considered partway into the dynamic range? Besides that, if the idle isn't even really very idle (I've seen 7% usage listed and that's pretty high) that's another factor..

People are using derivations like this to try to estimate how low the power consumption could be in a tablet. They're basically saying that the measurement with maximum performance profile while idle represents the SoC itself using essentially nothing and the big power consumption can be blamed on the design of the laptop. We've already seen reviews take the difference here and call it "CPU TDP." That's not accurate if the CPUs are still using a big chunk of power while idle. It's also not really fair to blame the CPU if it's happening because the software is artificially preventing it from entering power states that'd be truly representative of idle consumption.

I expect the balanced profile would show less power consumption with the single threaded loads but same power consumption with the fully MT ones. Assuming that they're allowed to hit the same clocks. I expect balanced to simply be more dynamic, allowing for more aggressive power savings for idle cores, mainly at the expense of switch time to get into the higher performance modes. Tests with these sorts of profiles tend to show they're only very slightly behind the full performance modes in measured performance.

I will agree that it seems a bit odd for the single-threaded cinebench at 1.4 GHz to show a dynamic power usage of 3.4W while the multi-threaded cinebench at 1.1 GHz only bumps up to 4.6W... but yeah, can only speculate as to why that might be. It can't just be an issue with power gating because that would push up the baseline as well. And it doesn't make sense that much higher voltage would be needed for operation at 1.4 GHz since I'd expect Temash to be the exact same die as Kabini which can operate at far higher frequencies.

Having a higher baseline due to not power gating those cores is exactly what I'm saying. That could (at least partially) explain the 3.5W difference from balanced and a much smaller power scaling when going from one to four cores.

Their policy for clocking strictly based on active cores seems pretty rudimentary, and well behind what Intel is using (more similar to what ARM SoCs can do right now). You would think they would at least take into consideration if the 3D part of the GPU is in use or not.

 

[SPBHM]

any comparison on the same clock (and number of cores) with bobcat?