AMD Sempron (Kabini) Quad Core AM1 @ $37

[Vesku]

Shame none of the AM1 boards so far declares ECC support, a ~$60-90 ECC capable CPU+Motherboard would be a nice niche imo.

 

[Abwx]

Sorry but even the cheap Asrock H81M-DGS ($49) is already a lot better than AM1 can offer. That and a G1820 we are talking about $103.

The cheaper AM1 5350 combo is $93, not a lot of price difference there.

Now mITX gona be cheaper on AM1, yeah.

On H81 you have full PCI-E, more satas and upgrade up to an I7...

No AVX no VT no AES on thoses celerons....besides if you want a good MB
for Kabini get an Asrock AM1H Itx....

Official TDP is 53W btw....

 

[Enigmoid]

What is the price of that 15W Haswell ??? Those Haswell SKUs are not in the same segment as Kabini.

They may have lower power than Kaveri or Core i SKUs but they have 50% more power consumption than Kabini. Not in the same segment.

Which is exactly why I said that Kabini has price (weighted by power) going for it and not much else.

Its not terribly efficient.

Its not high-performance.

Its an okay setup at a low price.

And power consumption is very low on all these platforms. You would have to be in a situation where 35W vs. 50W is a gamebreaker.

 

[USER8000]

In the UK,you can get an Asus AM1 mini-ITX motherboard for only £26. That means with a Sempron 3850,the total comes to only £55!

Power consumption looks quite low:

http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle

Hardware.info used a 150W pico-PSU,and power consumption was just 10W at the wall when the system was idling.

 

[Abwx]

In the UK,you can get an Asus AM1 mini-ITX motherboard for only £26. That means with a Sempron 3850,the total comes to only £55!

Power consumption looks quite low:

http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle

Hardware.info used a 150W pico-PSU,and power consumption was just 10W at the wall when the system was idling.

Nice tests that show what 25W TDP means by AMD specs and what 10W TDP means by intel specs....:biggrin:

 

[piesquared]

In the UK,you can get an Asus AM1 mini-ITX motherboard for only £26. That means with a Sempron 3850,the total comes to only £55!

Power consumption looks quite low:

http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle

Hardware.info used a 150W pico-PSU,and power consumption was just 10W at the wall when the system was idling.

Really liking these little Athlons and Semprons. Apparantley they are being very well recieved by Taiwan and customers alike.

 

[Sweepr]

-J1900 is about the same on CPU, uses less power and its pasive, about the same price too, but inferior in gaming.
-G1820 is better all around, but uses more power, it also cost about $10 more.

Intel's 2014 line-up leaves Kabini in a really akward position. I hope they have a new chip by early 2015 because 14nm Braswell will surely shake things up a bit.

 

[Abwx]

Which is exactly why I said that Kabini has price (weighted by power) going for it and not much else.

Its not terribly efficient.

Its not high-performance.

Its an okay setup at a low price.

And power consumption is very low on all these platforms. You would have to be in a situation where 35W vs. 50W is a gamebreaker.

Thoses results are quite different from the dutch site, why.???.

Because PClab use two different set ups , one with a 550W PSU
and the other one with a 90W PSU, looking at the dutch site BT
is not more efficient in CPU tasks and largely inferior in perf/watt
once the GPU kicks in.

 

[Enigmoid]

Thoses results are quite different from the dutch site, why.???.

Because PClab use two different set ups , one with a 550W PSU
and the other one with a 90W PSU, looking at the dutch site BT
is not more efficient in CPU tasks and largely inferior in perf/watt
once the GPU kicks in.

PSU is incredibly overpowered but Baytrail is definitely sucking less power than kabini.

Yeah, Gen 7 graphics are pretty bad when you look at them alone.

 

[AtenRa]

In the UK,you can get an Asus AM1 mini-ITX motherboard for only £26. That means with a Sempron 3850,the total comes to only £55!

Power consumption looks quite low:

http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle

Hardware.info used a 150W pico-PSU,and power consumption was just 10W at the wall when the system was idling.

Hell, that is exceptional performance and power consumption.

 

[Abwx]

PSU is incredibly overpowered but Baytrail is definitely sucking less power than kabini.

Yeah, Gen 7 graphics are pretty bad when you look at them alone.

Significantly only in GPU+CPU but the perf/watt is still better in this case,
and somewhat in CPU load but we should ponderate thoses numbers
by the perfs through thoses tests where Kabini has some lead rendering
BT marginal lower CPU comsumption pointless.

Besides THG use two MBs for Kabini , we dont know wich one was
used for power numbers....

 

[VirtualLarry]

The 5350+AM1 is $93, and thats what im talking here, the Sempron Quad+AM1 mb is about $74, Sempron dual will be at 50 or something?

http://www.newegg.com/Product/Produc...Size=10&page=1

5350 + MSI ITX AM1 mobo = $85 USD

 

[Vesku]

Intel's 2014 line-up leaves Kabini in a really akward position. I hope they have a new chip by early 2015 because 14nm Braswell will surely shake things up a bit.

If the following sites power and performance results are representative:

Power consumption looks quite low:

http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle

Hardware.info used a 150W pico-PSU,and power consumption was just 10W at the wall when the system was idling.

Then 28nm Kabini is quite a fine competitor to desktop Baytrail. The main issue with Kabini was getting enough chips out on the market, almost no integrated TSMC Kabini mini-itx boards for the months prior to the socketed Globlalfoundries versions showing up, and AMD's continued inability to convince OEMs to pay much attention to them.

 

[coercitiv]

Intel's 2014 line-up leaves Kabini in a really akward position. I hope they have a new chip by early 2015 because 14nm Braswell will surely shake things up a bit.

AMD claims Beema will drop TDP to 15W and still offer increase in performance (20-25% both CPU and GPU). The most important part will be pricing though, there we'll have to see if they remain competitive.

 

[Enigmoid]

Significantly only in GPU+CPU but the perf/watt is still better in this case,
and somewhat in CPU load but we should ponderate thoses numbers
by the perfs through thoses tests where Kabini has some lead rendering
BT marginal lower CPU comsumption pointless.

Besides THG use two MBs for Kabini , we dont know wich one was
used for power numbers....

What?

The J1900 doesn't exceed a delta of 10W. I see a delta of 4W for BT vs 12W for Kabini on the CPU only test. Kabini is 30% faster but is using significantly more power.

IMO, kabini is a half-finished product. It needs power gating/turbo and better power management.

Beema/Mullins should help.

 

[Abwx]

The J1900 doesn't exceed a delta of 10W.

Yes , because it has higher temperature at iddle hence the available
temperature delta is limited , if a little fan was used it would consume
significantly less at iddle while reaching the same max power comsumption
wich would result in a higher delta although with better perfs as well.

IMO, kabini is a half-finished product. It needs power gating/turbo and better power management.

Beema/Mullins should help.

Given that it s 28nm it performs very well , as for Beema/Mullins
not sure if improvements are due to extensive tweakings more
than eventualy GF s supposed better process.

 

[Enigmoid]

Yes , because it has higher temperature at iddle hence the available
temperature delta is limited , if a little fan was used it would consume
significantly less at iddle while reaching the same max power comsumption
wich would result in a higher delta although with better perfs as well.

You are saying that there is a tremendous amount of leakage going on?

Give me a break.

There is no way I'm going to believe than GF's 28nm allows AMD to get anywhere close to 22nm finfet in terms of efficiency.

The BT CPU is operating at a higher temperature and the temp delta between idle and load is almost twice as much on the BT platform. If a fan was used power would be almost the same (because a dinky little fan for these processors isn't going to draw any more than 1-2W); the load temps would drop as well and given the non-linear leakage vs. temp relationship the delta would shrink.

 

[Abwx]

You are saying that there is a tremendous amount of leakage going on?

How do you explain BT s 16W+ when iddle on the dutch site.?.

Remember this is with an efficient PSU and Kabini has lower
iddle power, explain also this if possible...

http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle

Power with cinebench :

http://nl.hardware.info/reviews/533...topplatforms-stroomverbruik-cinebench-115-max


With 3Dmark cloudgate :

http://nl.hardware.info/reviews/533...ms-stroomverbruik-3dmark-cloud-gate-gemiddeld

 

[blackened23]

http://www.newegg.com/Product/Produc...Size=10&page=1

5350 + MSI ITX AM1 mobo = $85 USD

http://www.newegg.com/Product/ComboD...=Combo.1596228

112$. I'd say 27$ would be worth it here.

I cannot fathom why either this AMD platform or the Bay Trail BGA platform cost nearly 90$. What a ripoff for what you get, considering the alternative that doesn't cost much more.

 

[AtenRa]

http://www.newegg.com/Product/ComboD...=Combo.1596228

112$. I'd say 27$ would be worth it here.

I cannot fathom why either this AMD platform or the Bay Trail BGA platform cost nearly 90$. What a ripoff for what you get, considering the alternative that doesn't cost much more.

You pay more for Mini-itx and for the lowest power consumption in desktop.

 

[Enigmoid]

How do you explain BT s 16W+ when iddle on the dutch site.?.

Remember this is with an efficient PSU and Kabini has lower
iddle power, explain also this if possible...

http://nl.hardware.info/reviews/533...goedkope-desktopplatforms-stroomverbruik-idle

Power with cinebench :

http://nl.hardware.info/reviews/533...topplatforms-stroomverbruik-cinebench-115-max


With 3Dmark cloudgate :

http://nl.hardware.info/reviews/533...ms-stroomverbruik-3dmark-cloud-gate-gemiddeld

Could be a mobo thing.

Not sure. But don't think its a passive vs. active cooling thing.

 

[coercitiv]

as for Beema/Mullins
not sure if improvements are due to extensive tweakings more
than eventualy GF s supposed better process.

Anandtech guessed tweakings.

AMD hasn’t disclosed how much the underlying architecture has changed, and I would guess the Puma cores are actually quite similar to Jaguar cores, but the net result is a 2X improvement in performance per Watt according to AMD.

Update: It sounds like most of the performance gains come from frequency increases, while power improvements happen at the SoC level.

 

[Abwx]

Could be a mobo thing.

Not sure. But don't think its a passive vs. active cooling thing.

Or a chip not as good as on other sites , remember that all chips do not use the same voltages so some chips can consume somewhat more in quite
a large variability, that said i prefer numbers from a relevant set up than
from THG who seems to produce hasardous not to say rigged results,
just look at the recent power numbers of BT in the AM1 review and
in a previous review using a BT that was supposed to consume less :

They did their job very well , i hope that they were rewarded
accordingly for such efforts.

 

[Vesku]

AMD claims Beema will drop TDP to 15W and still offer increase in performance (20-25% both CPU and GPU). The most important part will be pricing though, there we'll have to see if they remain competitive.

Availability is up there with price in importance. For example the A8-7600 looked good in reviews and expected MSRP at the beginning of the year but has yet to actually be purchasable.

Good to see that AM1 APUs seem to have a decent launch supply, hopefully the same thing is true with Beema.

 

[Abwx]

Anandtech guessed tweakings.

Undoubtly but it s still 28nm and the claimed numbers for Beema/Mullins would be quite possible with this same node if it was only for the benchs scores that were published , thing is that not only perfs are better but TDP is dramaticaly reduced , that s too much for a tweak that would use the same process.