No.
http://www.notebookcheck.net/Review-Acer-Iconia-W4-820-2466-Tablet.114871.0.html
BT 3740 runs Prime 95 at max turbo (1.86 mhz) stable.
There is no downclocking to base clocks.
Luxmark will run at max turbo.
Under Prime and Furmark CPU drops to 1333 mhz
Note that some of the other BT tablets show massive throttling which appears to be temperature based, hitting over 70 degrees on the SOC; the iconia is around 50.
Needless to say these are extremely intensive tests.
for instance a J1900 consume at least 15W to do 1.8 in CB
No it doesn't.
And likewise the Z3770 consumes 2.5W to do 1.47.
We know because the SoC power has been measured.
http://www.anandtech.com/show/7314/intel-baytrail-preview-intel-atom-z3770-tested/
Edit: The software IA power monitor says 2.64 W on Enigmoid's post, too. I suspect that if someone here with a socketed BT would fire up HWiNFO, we would get numbers for Celeron J1x00.
HWmonitor doesnt report the real value.
The lower number i ve seen on CB with an efficient pico psu and a J1900 is 21.1W wich is low but the iddle power is quite high at 16.8W wich say that theses chips, on the DT versions, are leaky , much less surely for the cherry picked tablet variants but still the power under load will be about the same.
http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle
http://nl.hardware.info/reviews/533...topplatforms-stroomverbruik-cinebench-115-max
Are you following the thread?You clearly didn't read all of blackened and atenra's posts or you are intentionally misunderstanding. Read it again and you will see why pointing that out was necessary.
HWiNFO reads the CPU sensor outputs. They have been available since Sandy Bridge and are the same sensors that trigger TDP boosting and thermal throttling. If these power values are not accurate and under reporting, that implies a vast conspiracy between OEMs and Intel.
Here's my Prime95 blend results. My power meter (at the wall) reads ~190W loaded and ~80 watts at desktop. Taking into account power supply efficiency at 190W and 80W AC (87% minimum and ~80%), HWiNFO comes within a few watts of the estimated power consumption.
If you're claiming that a ~100mm^2 22nm FinFET die with power gating all over the place uses 5W when idle, you better have more substantial proof than wall measurements, which includes the whole platform.
Are you following the thread?
http://forums.anandtech.com/showpost.php?p=36319140&postcount=209
"Yes, Precisely."
That's the first words.
HWiNFO reads the CPU sensor outputs. They have been available since Sandy Bridge and are the same sensors that trigger TDP boosting and thermal throttling. If these power values are not accurate and under reporting, that implies a vast conspiracy between OEMs and Intel.
Here's my Prime95 blend results. My power meter (at the wall) reads ~190W loaded and ~80 watts at desktop. Taking into account power supply efficiency at 190W and 80W AC (87% minimum and ~80%), HWiNFO comes within a few watts of the estimated power consumption.
In the past, these numbers were known to be in error because they do not capture the static power losses, they only capture the dynamic power losses.
The way we could prove this was to put the CPU to task doing something (doesn't matter what) and note the reported power consumption value.
Then turn off the CPU HSF fan, or do anything to stifle airflow such that you can observe the CPU's temperature increasing. A situation which assuredly results in increasing power consumption by the CPU as the static power losses rise with operating temperature.
In the past we could easily observe the reported values failed to rise with rising temperatures. Proof beyond doubt that the reported power values were flawed and in error.
Maybe this has changed recently (in the last 6 months), but I'm doubtful.
At any rate, I recommend doing the test yourself as it is then easy to prove (or disprove) to yourself whether or not the reported power values can be expected to be accurate.
Well, even if Intel XTU is reporting an increase in TDP that seems correlated with temperature, I failed to see that with HWiINFO data from sensor reporting.Reported TDP grows by a few watts and shows a increase in ~5W at the wall and the CPU TDP meter by ranging from ~75C to 90C. The increase in leakage is comparatively subtle but it's definitely there and collaborated by my wall meter. Core clock was locked at 3.7 GHz @ 1.19 V and Prime95 SmallFFT was used to generate load.
Well, even if Intel XTU is reporting an increase in TDP that seems correlated with temperature, I failed to see that with HWiINFO data from sensor reporting.
Power is expressed in Watts on X Axys, Temps are in Celsius on Y axis.
What's even more interesting is that CPU Package Power for Haswell seems to drop in value as temperatures increase, indicating some sort of error or compensation.
Moving back to Kabini, reported "CPU Total Power" does seem to increse with temperature, although a clear correlation has yet to be established.
HWiNFO reads the CPU sensor outputs. They have been available since Sandy Bridge and are the same sensors that trigger TDP boosting and thermal throttling. If these power values are not accurate and under reporting, that implies a vast conspiracy between OEMs and Intel.
Here's my Prime95 blend results. My power meter (at the wall) reads ~190W loaded and ~80 watts at desktop. Taking into account power supply efficiency at 190W and 80W AC (87% minimum and ~80%), HWiNFO comes within a few watts of the estimated power consumption.
http://forums.anandtech.com/showthread.php?t=2341969&page=102On the other end, a combined Prime95 and FurMark will result in 28Watts on the J1900 and in 22Watts on the J1800. Core power consumption measured via CPUIDs HWMonitor showed 2.29Watts for the J1900 cores and 2.4Watts for the J1800 cores, while the package consumption was put at 6.85Watts for the J1900 vs. 6.54Watts for the J1900.
This oddity was consistent and I can only explain it by HWMonitor only measuring one of the two CPU blocks on the J1900, but the full GPU block (and remainder of the SoC), which is clocked a little higher on the J1900 under load.
If you're claiming that a ~100mm^2 22nm FinFET die with power gating all over the place uses 5W when idle, you better have more substantial proof than wall measurements, which includes the whole platform.
It does because PC lab used an efficient set up for the J1900 with a 90W PSU whose losses are totaling at worst 20% so tell me what is the part of the Soc in thoses amounts given the PSU and the basic MB with likely 4GB and a SSD as single other items, where does all thoses watts get dissipated.?.
You can see the scalings on MT below, so that s right but a 3740 or 3770 will stay within its TDP only at base frequency if all cores are fully loaded, without power comsumption numbers theses scores are not totaly meaningfull, for instance a J1900 consume at least 15W to do 1.8 in CB, no way a tablet BT can do 1.5 with less than 10W.
Maybe possible with the E1-Micro 6200T. But it really depends how much that costs and it would need a heatsink unlike the Raspberry Pi.
I'd love to see a fanless mini PC using the A4-Micro 6400T or A10-Micro 6700T a la AMD's nano PC concept:
But I just know that it's never going to happen.
edit: Interesting point about STAPM from the TechReport article:
"That gain need not come at the cost of battery life, either. On the contrary, a faster-running processor can complete tasks more quickly and thus spend more time at idle, which "actually saves energy in many common use-cases," the company tells us."
Obviously there's a limit (eg, lower efficiency as clock speeds and voltage increase), but it makes sense.
Speaking of cheaper desktops, I wish AMD would do something with unlocked multipliers for the lowest end cat core chips.
One example would be the current Kabini E1-2100 which comes passively cooled for MINI-ITX. However, if there is a desire for more performance the multiplier should be able to be changed and provisions should be made for a standard case fan to be attached to the passive heatsink.
Or in some cases the lower end Kabini could be given the larger passive heatsink of a higher end quad core:
(See the two example below: One is a E1-2100 and the other is a A4-5000. They appear to have the same PCB, but different size passive heatsinks.)
http://www.biostar.com.tw/app/en/mb/introduction.php?S_ID=712
Dual DDR3-1600
It supports 2 or 4 DIMM Slots DDR3 memory that features data transfer rates of DDR3 1600/1333. The processor support 2 memory channels and setting the memory multiplier, allows selection of the channel speed. These clearly uplift the band width and boost the overall system performance.
Copy + Paste typo, or are some of these chips dual-channel?
Speaking of cheaper desktops, I wish AMD would do something with unlocked multipliers for the lowest end cat core chips.
Copy + Paste typo, or are some of these chips dual-channel?
Hardware makers aren't tripping over themselves to provide AMD products so I think the potential warranty headaches of having unlocked SoCs soldered directly to the motherboard is one of the reasons that hasn't happened.
1. ) If necessary, the motherboard manufacturer could always limit the multiplier (or other settings like voltage).
2.) Since the cpu can't be changed out of the socket (with a BGA processor like the E1-6010), being able to change the multiplier makes the board more attractive IMO.