Athlon 5350 mini video Review

[Vesku]

Do these reviewers not see a need to plunk down for an 80 plus low wattage power supply? Seems somewhat important given one of the big draws of these products is low power consumption.

 

[ethebubbeth]

Do these reviewers not see a need to plunk down for an 80 plus low wattage power supply? Seems somewhat important given one of the big draws of these products is low power consumption.

Yeah, I'd love to see a review comparing power consumption at the wall between various power supplies and a PicoPSU on a low wattage system like this.

 

[Shivansps]

Anadtech did not incluide a G1820 wth...

 

[OCGuy]

I agree, for any fair power consumption measurement, an 80 Plus certified PSU @ at least 20% load (preferably 50%) should be used and recommended.

 

[Vesku]

Interesting that the PClab's review shows J1800 boosts higher than J1900, Anandtech review used J1800 for single thread comparison and J1900 for multi thread comparison. Perhaps the Anandtech reviewer should go back and insert J1900 numbers in the single thread tests. Although, they are using a passive J1900 motherboard so perhaps they could borrow the AM1 fan for the single thread J1900 runs and add a caveat, or just caveat that the lower burst speed is a trade off for being a passive design.

 

[Abwx]

Interesting that the PClab's review shows J1800 boosts higher than J1900, Anandtech review used J1800 for single thread comparison and J1900 for multi thread comparison. Perhaps the Anandtech reviewer should go back and insert J1900 numbers in the single thread tests. Although, they are using a passive J1900 motherboard so perhaps they could borrow the AM1 fan for the single thread J1900 runs and add a caveat, or just caveat that the lower burst speed is a trade off for being a passive design.

Depending at wich temp Intel did set the throttling the celerons
are indeed badly used since the passive colling increase power
comsumption at iddle and limit its perfs once all threads are used,
a cheap active cooling is de facto mandatory.

 

[Sweepr]

I was actually surprised to see 2.0-2.42 quad-core Bay Trail-D (Celeron J1900) matching/beating this 4C/4T 2.05GHz Kabini in a few CPU benchies. 10W Pentium J2900 (2.42-2.67GHz) should have no trouble matching the fastest Kabini CPU-wise most of the time.

http://nl.hardware.info/reviews/5330/amd-am1-vs-intel-bay-trail-d-review-goedkope-desktopplatforms

 

[Abwx]

10W Pentium J2900 (2.42-2.67GHz) should have no trouble matching the fastest Kabini CPU-wise most of the time.

http://nl.hardware.info/reviews/5330/amd-am1-vs-intel-bay-trail-d-review-goedkope-desktopplatforms

Amazing that this very same link show that BT is close to Kabini
TDP wise and is surely not 10W but still you re quoting this
imaginary value as real, you think that it will get closer to 10W
once frequency is even more increased.??.

 

[Enigmoid]

Amazing that this very same link show that BT is close to Kabini
TDP wise and is surely not 10W but still you re quoting this
imaginary value as real, you think that it will get closer to 10W
once frequency is even more increased.??.

Idle power is way higher for the BT platform. Not sure if its the mobo or what but the delta is less than 10W.

 

[Abwx]

Idle power is way higher for the BT platform. Not sure if its the mobo or what but the delta is less than 10W.

I dont count BT iddle power as a valuable measurement in this case
as it is due to passive cooling mainly , with a little fan it would
substancialy decrease but it s quite possible that frequency is related
to temperature, in that case power comsumption is allowed higher
values on a short term like 5 minutes resulting in the measured 20W+.

 

[coercitiv]

Idle power is way higher for the BT platform. Not sure if its the mobo or what but the delta is less than 10W.

I don't think we should read much into the power numbers from most reviews, there are too many variables which could affect results.

I think we can all agree that Bay Trail will consume less under load while idle power consumption will probably be low enough for both platforms.

The delta between idle/load for the J1900 in the hardware.info review is indeed strange, they should have checked that. Then again Tom's Hardware had no problem posting a 13C idle temp for Kabini SOC. Some reviewers should know better.

 

[coercitiv]

I dont count BT iddle power as a valuable measurement in this case as it is due to passive cooling mainly , with a little fan it would substancialy decrease

I doubt chip temperature can affect idle power usage, when the most of the time the SOC is at sleep.

 

[Abwx]

I don't think we should read much into the power numbers from most reviews, there are too many variables which could affect results.

I think we can all agree that Bay Trail will consume less under load while idle power consumption will probably be low enough for both platforms.

The delta between idle/load for the J1900 in the hardware.info review is indeed strange, they should have checked that. Then again Tom's Hardware had no problem posting a 13C idle temp for Kabini SOC. Some reviewers should know better.

I will add to the above post that passive cooling will exacerbate
the chips variability, there s no two chips that are the same hence
without a fan to equalize the caracteristics the sites will indeed
post different results according to their samples quality and this even
if they use the same set ups.

 

[Enigmoid]

I will add to the above post that passive cooling will exacerbate
the chips variability, there s no two chips that are the same hence
without a fan to equalize the caracteristics the sites will indeed
post different results according to their samples quality and this even
if they use the same set ups.

You are undoubtedly true to the order of maybe 10-20% variation. However, these variations are much larger.

 

[Abwx]

You are undoubtedly true to the order of maybe 10-20% variation. However, these variations are much larger.

Possible that they used a process optimised for low power comsumption
for mobile devices that work at lower frequency than DT gear and we all know that for such features transistors with lower threshold voltage are necessary for the circuit to accomodate the lowest possible supply voltage, the downside is that if you increase voltage to reach higher frequencies leakage will go through the roof as the lower threshold inherently induce higher transistor residual conduction when it is switched off and this is the case even with power gatings as the gating is done by, yes, transistors.

Anyway it s just speculation, to be sure it would be necessary to have
access to the most secretive datas , that is , the spice parameters
of the transistors.

 

[SPBHM]

Anadtech did not incluide a G1820 wth...

funny thing is that they've included the rare G1101 (1156 2.2GHz) and not a single regular 1155/1150 Celeron, only the also less popular G465 (single core)


some of the others have the G1820, and if you can have both for comparable price and the worst case 15-20W+ are not a problem for you, the G1820 is a much better choice as expected,

but compared to the J1900 the 5350 is also a clearly better CPU, if the worst case 10W are not a problem, and pricing is close.

 

[Shivansps]

Yeah pclab review is what i was expecting, the G1820 is better, for a few more W, trading off AVX and AES for full pci-e and more satas.

J1900 a little worse on CPU, and a lot worse on IGP, but using less power and pasive.

Personally i have a hard time figuring out why i should buy a 5350, the Semprons dual/quad on AM1 in other hand are just cheaper, they gona sell a lot because of that.

 

[Abwx]

Yeah pclab review is what i was expecting

I have no doubt about this claim being accurate...

 

[atakall]

OP: (1) What are the underclocking/undervoltage options & are there restrictions (eg., the Asus E35M1 e-350 limited underclocking via the fsb to 90mhz at its lowest; settings also allowed for undervoltage).
(2) Could you attempt to underclock/undervolt your 5350 and provide power readings for the underclock version? For example, maximum undervolt at stock power readings. Power readings at maximum undervolt @ 1.6ghz.

Since the 5350 (quad @ 2.05ghz; 600mhz GPU) is merely being a better binned part than that of the Athlon 5150 (quad @ 1.6ghz;600mhz GPU) & the Sempron 3850 (quad @ 1.3ghz; 450mhz GPU) , it should provide for the best case scenario as to underclocking (although this isn't always the case). These measurements would be interesting for those looking for maximum efficiency that don't necessarily need the highest clock part but prefer to maintain it as an option given the cost difference among the chips (eg., purchase 5350 but operate it undervolted at 1.6ghz instead of purchasing the 5150). For my particular usage, I'm looking to replace my e-350 with a 5350 in a WHS 2011 setup (& it seems this was not too uncommon of a use for the e-350).

 

[Shivansps]

I have no doubt about this claim being accurate...

yeah it was very unrealistic to expect the J1900 to be better than a 2ghz Kabini and a 2Ghz kabini to be better than a G1820, i was kinda expecting the Haswell GT1 to be better too, especially when AMD stop targeting the Celeron G1620 after the G1820 was released and focused on BT-D.

There was no much suprise on power figures either, BT-D/Kabini/Haswell, except that J1900 seems to be using a lot of idle power for some reason, may be related to bad VRMs.

You where expecting something else?

 

[Abwx]

The results are somewhat in line with what expected for Kabini since the 5350 was already existing under the A6 5200 moniker and we knew its perfs, much less for BT celeron iterations wich are obviously above the official 10W claim, by how much exactly is unknown since the available power measurements do not isolate the different parts contribution but anyway THG numbers seems to me among the less reliable for whatever plateform including of course BT.

As a comparison another site did an AM1 review with more realistic
number but there s no comparison with identical products, weird thing
is that they use a 1050W PSU , 200W more than THG s.

http://www.eteknix.com/amd-kabini-am1-athlon-5350-apu-fs1b-review-winning-sub-60/11/

Edit : actualy the PSU is 400W for Kabini and 1050W for the two other systems.
This 400W PSU has an efficency close to 90% at 20% load, that is 80W output
power , roughly double the AM1 consumption but still efficient enough at theses
power levels.

http://www.ginjfo.com/dossiers/test...ht-power-e9-400w-silence-et-economie-20120112

 

[Abwx]

Another review made with this very same highly efficient PSU but this time there is numbers for the 5350, a J1800 and a A8 7600.

Idle :

Loaded by Luxmark :

The necessary luxmark score for perf/watt estimation :

http://www.golem.de/news/kabini-fuer-am1-im-test-gesockelte-jaguare-fuer-jedermann-1404-105670.html

 

[Abwx]

From anand s review this graph :

Digging about this "bench" :

Intel is a sponsor and member of the BenchmarkXPRT Development Community, and was the primary developer of WebXPRT* 2013 and MobileXPRT* 2013.

http://demo.motivace.com/intelretai...t-CompareTablets/index.html?culturecode=en-US

 

[atakall]

Underclocking/Undervolting

I'm address my previous post regarding the underclockability of this APU. It seems as though it can be done but is dependent on the features of the motherboard.

The ASRock AM1B-ITX bios currently does not allow one to change the clock or voltage for the core. For a full video of the ASRock AM1B-ITX firmware, seehttp://www.youtube.com/watch?v=YHoY5o1w0N8 @ 11 seconds. It appears the bios only provides for setting the speed/voltage of the RAM. These features may be accessible via ASRock's A Tuning software or ASRock Hybrid Booster but that's merely a conclusion derived from the description of the software (although A Tuning appears to just provide software access to the firmware level features).

The Asus boards, OTOH, do provide for frequency and voltage changes. The Asus AM1I-A manual explicitly discusses these firmware options, and contains pics of the bios screen with these options (see p.43 of the PDF manual), including changes to the multiplier, APU frequency, APU voltage & NB frequency.

Hopefully ASRock implements these options b/c at this point their AM1 boards are the only ones with 4 SATA ports.

Op, could you install the Hybrid Booster and A Tuning to see if it provides access to changing the multiplier, vcore, etc.?

 

[monstercameron]

now this video/demo should be in the advertising for the athlon.
https://www.youtube.com/watch?v=0Y9WBv5-R-Y
^ titanfall running decently on amd am1 5350 @720p