AMD Beema/Mullins Launch Thread

[cbn]

On that E1 Beema, I think having it be dual bin would be a good idea also.

Example: my E1-2100 is a low voltage processor (1.0 volts in the ECS KBN-I/2100 BIOS IIRC)

With the Beema E1 (which I am assuming it is low voltage as well), it would be great if there were also a standard voltage setting with an increased clock rate. (A toggle switch in the BIOS could enable the second setting).

 

[Idontcare]

Ah, I an excuse to tinker around with my case! So I unplugged the second fan on my NH-D14 and my exhaust fan so only one fan that I can easily block is running. I had to actually screw around with my overclock to find a range of temperatures.

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.

Non-grammar edit: Maybe this is something new with Haswell.

Edit 2: Cripes, had to use a pair of chopsticks to put the fan power back on.

Edit 3: I thought that a 5W increase in leakage seemed low for a 15C swing, but then I found this slide.

It would fit with your leakage measurements on Ivy Bridge, I think.

Excellent, and well done!

Good to know things have changed in time and the reported power numbers are more realistic in that they at least capture some aspect of the static power.

I wonder if this is true for Haswell in part because Intel moved the power deliver on-package with FVIR, and as such all the necessary hardware is now present and accounted for to robustly make such power measurements?

 

[coercitiv]

Here's some interesting power usage sensor data from a Cinebench 11.5 run. The Athlon 5350 CPU was clocked at 1.5Ghz and undervolted to 1V. Data was collected with HWiNFO 64 4.37

Average reported power draw for CPU Total Power was 2.67W, while power usage measured at the wall had a delta of 6W (14W idle, 20W load). Considering PSU efficiency @ 75-80%, VRM efficiency @ 90%, some 1-1.5W used by RAM and MB under load, and the sensor data makes sense.

 

[atakall]

The Athlon 5350 CPU was clocked at 1.5Ghz and undervolted to 1V.

Excellent information. I had asked about underclocking the 5350 previously but it went unanswered. What motherboard did you use to underclock the 5350? Or did you use software? Was that the lowest stable undervolt you could get at 1.5Ghz?

Thanks.

 

[coercitiv]

Excellent information. I had asked about underclocking the 5350 previously but it went unanswered. What motherboard did you use to underclock the 5350? Or did you use software? Was that the lowest stable undervolt you could get at 1.5Ghz?

The CPU was undervolted with the maximum allowable offset on my motherboard: -300mV. It was done via BIOS and it is also the lowest stable voltage for the CPU at that clock speed and on this motherboard. (stable in Prime 95)

You should also note that static leakage seems to be taken into the equation by the AMD sensors. I am using the platform without a fan on the stock cooler, and reported CPU power usage increases as temps go higher. This is visible on the Cinebench chart, but I should confirm it with data taken while the CPU is being actively cooled.

To my knowledge only the Asus boards allow adjusting CPU voltage right now: AM1I-A and AM1M-A. I have the mITX model. The mATX model seems to be better equipped, allowing NB voltage tuning as well as a wider range for CPU voltage offset. LLC might also help a bit. Software underclocking is available via AMD drivers, athough I did not make use of it.

 

[coercitiv]

You should also note that static leakage seems to be taken into the equation by the AMD sensors. I am using the platform without a fan on the stock cooler, and reported CPU power usage increases as temps go higher. This is visible on the Cinebench chart, but I should confirm it with data taken while the CPU is being actively cooled.

Here's the data. Average reported CPU Total Power was 2.9W for the Hot Run and 2.6W for the Cold Run.

For the hot run I preheated the CPU using stress software until it reached 50C. For the cold run I used the case fan at high RPM to keep the CPU cool and the rest of the board with it.

 

[Maxima1]

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.

Wait. So just the temperature itself was wasting energy? How much is that improved TIM for Devil's Canyon? Why would anyone not want it then?

 

[EightySix Four]

Wait. So just the temperature itself was wasting energy? How much is that improved TIM for Devil's Canyon? Why would anyone not want it then?

Transistor leakage increases at higher temperatures, resulting in higher power draw.

 

[pw257008]

Coming soon, an A6-6310:

http://www.bestbuy.com/site/hp-15-6...ory-500gb-hard-drive-black-licorice/6534009.p

 

[frozentundra123456]

Just doesnt seem like a good fit. I just purchased a HP laptop with a full power i3 for the same price. Seems like the ideal form factor for these processors, as well as atoms, is 11-12 inch or less models. Each to his own I guess, but personally I would not be willing to give up that much performance in a laptop of this size. In something smaller, if battery life is good, sure.

 

[pw257008]

Just doesnt seem like a good fit. I just purchased a HP laptop with a full power i3 for the same price. Seems like the ideal form factor for these processors, as well as atoms, is 11-12 inch or less models. Each to his own I guess, but personally I would not be willing to give up that much performance in a laptop of this size. In something smaller, if battery life is good, sure.

I suspect they'll come down to $330 or so on average, as that's where the Kabini A6's tend to sell. Though even at that price, I'd easily pay $30 more for an i3, so I'm with you on the whole.

 

[coercitiv]

personally I would not be willing to give up that much performance in a laptop of this size

So in other words you also don't see the point in any 15.6" notebook equipped with i3 ULTs, right?

 

[frozentundra123456]

So in other words you also don't see the point in any 15.6" notebook equipped with i3 ULTs, right?

Not really in the cheap, 300 to 400 dollar range. Just a plain i3 (or i5) is very good performance for the money. Laptops in this price range are not going to be fancy ultrabooks with ultralow power processors for super long battery life. So you might as well get the best possible performance at a low price.

 

[ninaholic37]

The 15W A6-6310 should fit in 11.6" but would probably be rare and overpriced just like the 15W A4-5000 11.6" laptops were. They'd probably try and sell A4-6210 11.6" laptops for $600 and A6-6310 11.6" laptops for $700-800 (if they even make any) so they can keep their lower end 11.6" laptops at $400.

 

[rainy]

Coming soon, an A6-6310:

http://www.bestbuy.com/site/hp-15-6-...rice/6534009.p

Price is similar or even identical to Lenovo laptop with A6-5200 from last year.

http://www.bestbuy.com/site/lenovo-1...specifications

Btw, honestly I don't know what is base clock of A6-6310 - cpu-world showing it as 1.8 GHz, notebookcheck as 2.0 GHz and AMD webpage just maximum clock (2.4 GHz).

http://www.cpu-world.com/CPUs/Puma/AMD-A6-Series A6-6310.html
http://www.notebookcheck.net/AMD-A-Series-A6-6310-Notebook-Processor.115403.0.html
http://products.amd.com/en-us/Noteb...eon™+R4+Graphics&f3=2400&f4=4&f5=FT3&f6=15+W&

 

[strata8]

Price is similar or even identical to Lenovo laptop with A6-5200 from last year.

http://www.bestbuy.com/site/lenovo-1...specifications

Btw, honestly I don't know what is base clock of A6-6310 - cpu-world showing it as 1.8 GHz, notebookcheck as 2.0 GHz and AMD webpage just maximum clock (2.4 GHz).

http://www.cpu-world.com/CPUs/Puma/AMD-A6-Series A6-6310.html
http://www.notebookcheck.net/AMD-A-Series-A6-6310-Notebook-Processor.115403.0.html
http://products.amd.com/en-us/Noteb...eon™+R4+Graphics&f3=2400&f4=4&f5=FT3&f6=15+W&

It's 1.8 GHz. Initial reports said 2 GHz but all the laptop listings I've seen have said 1.8 GHz.

Seems to me that the A6-6410 might just be an A4-6310 with turbo enabled for the CPU and GPU. Makes me wonder if we'll see lower end turbo parts a month or two later like what happened with Bay Trail-M.

 

[monstercameron]

It's 1.8 GHz. Initial reports said 2 GHz but all the laptop listings I've seen have said 1.8 GHz.

Seems to me that the A6-6410 might just be an A4-6310 with turbo enabled for the CPU and GPU. Makes me wonder if we'll see lower end turbo parts a month or two later like what happened with Bay Trail-M.

the e2-6210 and e1-6110 are also there.
here are a few upcoming skus!

they took the listing down for the hp elitebook with a 12.5" fhd display and ssd...

 

[Shivansps]

Last time they where selling A4-1200 and A4-1250 for about +$350

 

[Idontcare]

Wait. So just the temperature itself was wasting energy? How much is that improved TIM for Devil's Canyon? Why would anyone not want it then?

Might be too old to be of interest to you but you can check out this thread on the topic of power-consumption versus operating temperature.

 

[Homeles]

Wait. So just the temperature itself was wasting energy? How much is that improved TIM for Devil's Canyon? Why would anyone not want it then?

In addition to what others have chimed in on, it's a positive feedback loop. The IC consumes power, which causes it to heat up, which causes it to consume even more power, which makes it heat up even more, which results in even more power consumption...

The most efficient servers and super computers use exotic cooling methods, as the lower temperatures result in lower power consumption.

 

[Roland00Address]

Wait. So just the temperature itself was wasting energy? How much is that improved TIM for Devil's Canyon? Why would anyone not want it then?

Think of it like this.

CPUs work on a process where they are on or off based off comparing the voltages of one chip state vs another. When anything in chemistry is heated, any element, any state of matter the electrons become more excitable and more reactive, literally the electrons are moving more.

What is electricity fundamentally? Electricity is the movement of electrons, electricity is the accumulation of charged particles that can be directed as a current from one place to another.

Increasing the temperature for a cpu adds more "static noise" to the measuring of point a vs point b, thus it is harder to measure the result of what the cpu just did.

-----

If you need less energy delta for a cpu to tell what is on and what is off, then you need less power to run the cpu.

Running the cpu at a lower temperature yields a smaller energy delta between on and off, thus keeping a cpu cool means a cpu needs less power to operate.

 

[NostaSeronx]

I think it should be mentioned that Puma adds;
RDRAND

I have heard that AVX2 and BMI2 will also be added to Puma+ on the 20nm node.

 

[NTMBK]

I have heard that AVX2 and BMI2 will also be added to Puma+ on the 20nm node.

Nope, Puma+ is the core in Beema and Mullins on 28nm:

 

[NostaSeronx]

Nope, Puma+ is the core in Beema and Mullins on 28nm

Nope, 28nm = Puma and 20nm = Puma+.

When the marketing team doesn't clarify with the engineering team we get these mistakes.

AMD's marketing fixed it after the May 4th update, but sometimes the marketing team still calls it Puma+ when it should be Puma.

The 20nm SoC makes use of Puma+ CPU cores and the 28nm Soc makes use of Puma CPU cores.

28nm, Tick -> Puma
20nm, Tock -> Puma+
20nm, Tick -> Leopard? (or Cheetah)
14nm, Tock -> Cheetah? (or Cheetah+)

 

[pw257008]

Well, the HP with an A6-6310 that was listed at $360 or so seems to actually sit exactly in the A6-5200 price range at $329.99.