12/07/2011

Samsung Slate 7 issue (Continueing)



Samsung Slate 7 stands on totally different podium to general tablets like as iPad, Galaxy Tab using light O/S. Even with gorgeous outfit and attractive usability from regular PC to tablet, users measures and debates fundamental performance of PC and escalates their questions. I believe Samsung, even any of slate makers, should accept this user’s measurement scale to overcome current tablet centric market barrier. I bring this post from a site and translate in English.


Samsung Slate Series 7 looks outstanding Windows tablet so far in its specification and harmonious with performance, portability and battery life. But Slate 7 is in the middle of arguing stage between its specification and real performance. Privately I want to verify this arguing point by myself with measuring of encoding performance in 1080i resolution video file.

Some more of the devices currently on the power adapter is connected to the encoding time was measured from the heavy-load are as follows:
(Encoding, first of all, is largely dependent on CPU performance, better performance, take a short time.)

Desktop i7-2600K 3.4GHz / 1m37s (8 GB of RAM, Win 64-bit, No difference between high-performance mode & 97%).
Desktop i7-2600K 3.4GHz / 1m 40s (8 GB of RAM, Win 64-bit, balanced mode).
HA GTXX53 Lv69 i7-2670 2.2GHz / 2m04s (16 GB of RAM, Win 64-bit)
Desktop i5-2500 3.3GHz / 2s12s (4 gigabytes of RAM, Win 64-bit)
Dell XPS15 i7-2630QM 2GHz / 2m29s (6 GB of RAM)
Desktop AMD Phenom II X4 945 3.0GHz / 3m6s (RAM 12 GB)
ASUS G73JH i7-740QM 2.93GHz / 3m14s (RAM 16 GB)
Sony VAIO SE17 i7-2640M 2.8Ghz / 3m24s (RAM 8 GB)
Lenovo X220T i7-2620M 2.70GHz / 3m42s (8 GB of RAM, Win 64-bit, English)
Sony CW26FK, i5-520M 2.40GHz / 5m03s
Samsung Series 7 Slate i5-2467m 1.6GHz / 6m10s (the latest fw05 Firmware, UEFI Windows installation, measured in outdoor patio).
HP 2740P i5-520M, 2.40GHz / 6m22s
Sony VAIO SE17 i7-2640M 2.8Ghz / 6m23s (RAM 8 GB, bass mode)
Motion F5V i7-640um 1.2GHz / 6m28s
HP TM2 i5-470um 1.33GHz / 7m44s (RAM 8 GB)
ASUS EP121 i5-470um 1.33GHz / 7m53s (4 gigabytes of RAM, Win 64-bit Home Premium, Maximum processor state = 97%)
ASUS EP121 i5-470um 1.33GHz / 8m03s (4 gigabytes of RAM, Win 64-bit Home Premium)
Samsung Series 7 Slate i5-2467m 1.6GHz / 8m33s (fw05 latest firmware, and after about 10 minutes after boot up).
Apple MacBook Air fourth-generation Core 2 Duo 1.8GHz / 9m15s (2010 model year, 4 GB RAM, Win 32-bit)
Samsung Series 7 Slate i5-2467m 1.6GHz / 9m28s (the latest firmware, fw05, high performance mode, power connected)
Samsung Series 7 Slate i5-2467m 1.6GHz / 9m55s (the latest firmware, fw05, power saving mode, power connected).

Dell Allienware M11X R1 SU7300 1.73GHz OC / 10m18s
Dell XT2 U9400 1.4GHz / 11m31s (3GB RAM, Win 7 32-bit, 64-gigabyte SSD)
Fujitsu P1630 SU9300 1.2GHz / 15m24s(US model, Penryn 45nm)
Fujitsu Q550 Z670 1.5GHz / 42m53s (high)
Fujitsu Q550 Z670 1.5GHz / 42m22s (Maximum processor state = 97%)

In general, the devices with same CPU show quite similar performance. ASUS EP121 which already was pointed out the performance degradation due to throttling has mostly similar performance to HP TM2 which employs the same i5-470um CPU.

However, particularly Samsung Slate 7 shows quite wide range of encoding elapsing time from 6 minutes 10 seconds to 9 minutes 28 seconds.

Here are some of score points of CPU from cpubenchmark.net,

i7-2600K 3.4GHz - 10057 points
i7-2620M 2.7GHz - 3908 points (X220T)
i5-520M 2.4GHz - 2362 points (2740p)
i5-2467M 1.6GHz - 2172 points (Samsung Series 7 Slate)
i5-470UM 1.33GHz - 1515 points (EP121, TM2T)
Core 2 Duo 1.8GHz - 1000-point level (fourth-generation MacBook Air)

Even the above figures reflect the exact inverse of the encoding time, Samsung Slate 7 should supposed to be stand at 6m10s position according to its specification.

But reality is it depends deeply depends on room temperature around the device.
(The experiment in patio took 6 minutes 10 seconds but this elapsing time increased 30% when I moved to indoor room.)

I suppose this severe performance degradation due to throttling, this 9 minutes level should be considered to the 4th generation of MacBook Air nearly.
 

Eventually how to make it improvement?

If you've already noticed the above table which showed closely, the power management feature in Windows can make it through.
Even slight difference in Maximum processor state = 97% than 100%, there was key solution.

You can reach its setting through Control Panel - Power Management - Additional power management - Change settings - Change advanced power settings - Processor power management - Minimum processor state, the maximum processor state can be changed by:
(Windows default setting of the adapter connected, high-performance mode, the minimum, maximum processor state = 100%.)

As a result Samsung Slate 7, shows quite improvement according to only the maximum processor state changing.


At the same equipment, the latest firmware, fw05, Win 64-bit.

Samsung Series 7 Slate i5 2467m 1.6GHz, 7m13s / 7m56s (Maximum processor state = 97%) - 1.5GHz work reliably
Samsung Series 7 Slate i5 2467m 1.6GHz, 7m25s (Maximum processor state = 98%)
Samsung Series 7 Slate i5 2467m 1.6GHz, 8m06s (Maximum processor state = 95%)
Samsung Series 7 Slate i5 2467m 1.6GHz, 8m09s (Maximum processor state = 85%) - 1.3GHz work reliably
Samsung Series 7 Slate i5 2467m 1.6GHz, 8m32s (Maximum processor state = 80%) - 1.2GHz work reliably
Samsung Series 7 Slate i5 2467m 1.6GHz, 9m22s (Maximum processor state = 100%)

In high-performance (100%) state, the clock is changed dynamically between 800MHz ~ 2.3GHz and impacts to its performance as much as clock speed, but only 97% of setting allows enough to operate with reliable high performance at 1.5Ghz.

As a conclusion,

1. Samsung Series 7 Slate in common use and default setting do not perform its best condition as its specification.
2. To minimized degradation of performance, 97% of power management prevents severe loss of its performance.

No comments:

Post a Comment