Atom Z3370-D 3Dmark Cinebench11.5 etc

[mikk]

I disagree the dual 1GHz should be fine
http://www.3dmark.com/is/187221
my c-50 9w scores
3DMark Score 10053.0
Graphics Score 12461.0
Physics Score 5998.0

so the a4-1200 should score a little higher around 7200 for the physics score.

compared to a nexus 10
est. a4-1200 | exynos 5 octa
graphics :12500 | 7952
physics :7000 | 8127
overall :11500 | 7990

and the exynos 5 dual is clocked higher, faster memory and uses a similar amount of power.
exynos 5 source http://www.anandtech.com/show/6875/3dmark-for-android-performance-preview

Physics score is 2x-3x slower. It's the same with Cinebench 11.5 or Geekbench. The CPU is 2x-3x slower, it's not competitive.

I have a tablet with a Dual core 1.5GHz ARM and crapy iGPU, im sure A4-1200 dual core 1GHz is way faster, not to mention the iGPU performance.

The only problem i see is Windows price, but they could use Ubuntu.

Basically every decent Quad ARM is faster than AMDs tablet CPU.

According to here: http://www.cpu-world.com/news_2013/...nch_J1750_J1850_and_J2850_ULV_processors.html J1850 is a notebook CPU, not tablet and has TDP of 10W.
These figures show it has CPU and GPU performance considerably lower than A4-5000, and is a lot more expensive ($82).
The performance matches Intel's marketing information, more or less.

Lots of false statements here. J1850 is an entry Desktop model. According to Intels price list it costs $82.

 

[lefty2]

according to this site the 1007U costs $86, but you can buy an entire mitx board with one + pch for $90

as for performance I do think Silvermont is not going to beat Kabini, but it might be close for some stuff, like the old Atom was slower than Bobcat overall, but could be faster for some things...

the biggest thing will be power usage and pricing, both seem to be pretty good, but Intel looks like they can go lower, also Intel have a greatly improved IGP I would think.

Yeah, $82 does sound too high.
I don't think IGP is good enough though, 4 EUs clocked at 792MHz is going to be pretty poor.

 

[monstercameron]

Physics score is 2x-3x slower. It's the same with Cinebench 11.5 or Geekbench. The CPU is 2x-3x slower, it's not competitive.





Basically every decent Quad ARM is faster than AMDs tablet CPU.






Lots of false statements here. J1850 is an entry Desktop model. According to Intels price list it costs $82.

i'll give you the physics score but cinebench? didnt think there was arm support and geekbench scores between arm/andriod and x86/windows cant quite be compared. It is slower but "...way too slow..." i don't agree with that.

besides mobile scenarios prefer gpu perf which the a4-1200 is very competitive in.

 

[wlee15]

J1850 CB11.5 score looks comparable to A4-5000. It's a 32bit result and also Cinebench 11.5 is floating point heavy which should favour Jaguar.

I believe Cinebench uses scalar FP so Jaguar's increase in FP width isn't really a benefit in this application.

 

[Sweepr]

An A4 5000 with 4C @ 1.5GHz and 500MHz GPU scores ~23k on Ice Storm. A4 1200 has 2C @ 1GHz and a GPU clocked at only 225MHz (33% CPU & 45% GPU performance of an A4 5000).

 

[mikk]

i'll give you the physics score but cinebench? didnt think there was arm support and geekbench scores between arm/andriod and x86/windows cant quite be compared. It is slower but "...way too slow..." i don't agree with that.

besides mobile scenarios prefer gpu perf which the a4-1200 is very competitive in.

A6-1450 scores ~1.00 in CB11.5 @64bit. It clocks with 1.0-1.1. A4-1250 with only 2 cores and 1.00 Ghz must be in the 0.4-5 range. Compared to Silvermont it's not competitive. Same for Geekbench, here: http://browser.primatelabs.com/geekbench2/compare/2226937/2250328

Jaguar isn't doing good in a Tablet constrained TDP environment.

 

[Abwx]

A4-1250 with only 2 cores and 1.00 Ghz must be in the 0.4-5 range.

Score is 0.51 while a bobcat 1.6ghz is 0.61 , so essentialy
a dual core kabini 1ghz is worth a 1.33ghz bobcat in this
application at least , no doubt that it would suit a tablet.

 

[Khato]

Score is 0.51 while a bobcat 1.6ghz is 0.61 , so essentialy
a dual core kabini 1ghz is worth a 1.33ghz bobcat in this
application at least , no doubt that it would suit a tablet.

In terms of performance, sure. But what does their power usage look like in low load situations? Last I checked they're lacking the necessary power optimizations to work well in the workloads that actually matter on tablets.

 

[monstercameron]

In terms of performance, sure. But what does their power usage look like in low load situations? Last I checked they're lacking the necessary power optimizations to work well in the workloads that actually matter on tablets.

well the a4-1200 is similar to this http://wccftech.com/amd-series-gx-210ja-soc-apu-revealed-tdp-6w/ and they claim 3W average use, so double that for max utilization and gpu [and because amd] and I believe this chip[by itself] will eat 6-7W max then you can add platform power which can be another 5W[display, ram, storage etc.]. So it is possible that it could do 3.5-4.5 hrs underload in a tablet formfactor[40+whr battery a la ipad 4]

 

[Abwx]

Last I checked they're lacking the necessary power optimizations to work well in the workloads that actually matter on tablets.

Did you actualy check ?...
Because according to available TDP numbers it suits
perfectly for a tablet.

1.2W iddle, 1.4W internet browsing , 2.35W for vidéo h.264 1080p and 0.02W for S3 "standby" state.

http://www.forumsducomptoir.com/ind...apu-temash-a4-1200-et-a6-1450-pour-tablettes/

That is for the APU but remember that it s a full SoC.

Edit : the whole plateform numbers :

2.8W iddle, 3.7W web browsing, 5.3W vidéo and 0.07W in standby.

 

[Khato]

Did you actualy check ?...
Because according to available TDP numbers it suits
perfectly for a tablet.

http://www.forumsducomptoir.com/ind...apu-temash-a4-1200-et-a6-1450-pour-tablettes/

That is for the APU but remember that it s a full SoC.

Edit : the whole plateform numbers :

I believe the disconnect here is that we have very different definitions as to what level of power consumption is required for low-load tablet power consumption for the SoC. According to the detailed power consumption data that a few review sites gathered on Intel's Z2760 tablet, even including memory in the numbers you're still only at 0.42 watts idle (0.29 of which is memory), 0.5 watts web browsing, 0.99 watts 1080p video playback. (Note that I'm referencing the Tom's Hardware review as they have a simpler presentation of the data over all.)

Edit: Other point is that I tend to believe the numbers for Kabini from the AMD Tech Day slides more - http://cdn.wccftech.com/wp-content/uploads/2013/05/AMD-Kabini-Low-Power.jpg - 0.77 watts idle, 1.06 watts web browsing, 1.92 watts 720p video playback. Though again, if that's for the SoC only it's still using over twice the power of actual tablet SoCs for these low-load situations.

 

[SiliconWars]

The problem with tablet chips is that they are cheap, and AMD is getting a lot more money with Kabini in entry level notebooks.

Yields and bins will be near 90% for the quad core, 1.5GHz A4-5000. There is simply no motivation for AMD to sell perfectly good Kabini dies in tablets yet.

A second issue is the stats don't look that great. A dual core Temash at 1GHz will probably feel like one of the more responsive tablets you can buy, and graphically it'll be passable...but it's still a 1GHz dual core. That doesn't look good in marketing to the tablet generation, especially to the cheaper end of the market.

 

[SiliconWars]

I believe the disconnect here is that we have very different definitions as to what level of power consumption is required for low-load tablet power consumption for the SoC. According to the detailed power consumption data that a few review sites gathered on Intel's Z2760 tablet, even including memory in the numbers you're still only at 0.42 watts idle (0.29 of which is memory), 0.5 watts web browsing, 0.99 watts 1080p video playback. (Note that I'm referencing the Tom's Hardware review as they have a simpler presentation of the data over all.)

The problem with Clover Trail is that it doesn't have acceptable performance. The Z2760 isn't fast enough to make for an acceptable experience in any format. Yes you get great battery life, but you get nothing for it.

 

[monstercameron]

the old c-50 in a tablet http://www.notebookcheck.net/Review-Acer-Iconia-Tab-W500-Keydock-Notebook.53964.0.html

newer modeled tablet with atom z2760 http://www.notebookcheck.net/Review-Lenovo-IdeaTab-Miix-10-64-GB-Tablet.99494.0.html

if we look at the platform power for the amd c-50 w500 8.8W idle, 15W under load vs 4.7W idle, 7.4W under load for the atom we can see that it isnt looking bad.
if we could some how take off a few W for the 28nm process, remove the fch and account for 1.25V ddr3 ram [the rest being equal] then it is possible we might be looking at better numbers like 4.8W idle, 10W underload...[note all speculation]

 

[Abwx]

I believe the disconnect here is that we have very different definitions as to what level of power consumption is required for low-load tablet power consumption for the SoC. According to the detailed power consumption data that a few review sites gathered on Intel's Z2760 tablet, even including memory in the numbers you're still only at 0.42 watts idle (0.29 of which is memory), 0.5 watts web browsing, 0.99 watts 1080p video playback. (Note that I'm referencing the Tom's Hardware review as they have a simpler presentation of the data over all.)

Edit: Other point is that I tend to believe the numbers for Kabini from the AMD Tech Day slides more - http://cdn.wccftech.com/wp-content/uploads/2013/05/AMD-Kabini-Low-Power.jpg - 0.77 watts idle, 1.06 watts web browsing, 1.92 watts 720p video playback. Though again, if that's for the SoC only it's still using over twice the power of actual tablet SoCs for these low-load situations.

Z2760 numbers according to tom s hardware :

Battery is 30 Wh , so plateform comsumption is between 4 and 5.5W
depending of the usage.

If it is what you call adequate for a tablet then kabini is more than
suited to this task...

 

[SiliconWars]

A6-1450 scores ~1.00 in CB11.5 @64bit. It clocks with 1.0-1.1. A4-1250 with only 2 cores and 1.00 Ghz must be in the 0.4-5 range. Compared to Silvermont it's not competitive. Same for Geekbench, here: http://browser.primatelabs.com/geekbench2/compare/2226937/2250328

Jaguar isn't doing good in a Tablet constrained TDP environment.

The single threaded performance will basically be the same, it's only in multithreaded performance that it will fall behind. A 1GHz dual core Temash will feel very snappy in most tablet loads. That's the benefit of having good IPC.

 

[Khato]

The problem with Clover Trail is that it doesn't have acceptable performance. The Z2760 isn't fast enough to make for an acceptable experience in any format. Yes you get great battery life, but you get nothing for it.

Not going to try and claim otherwise with respect to performance, but Clover Trail tends to be roughly on-par with current ARM based tablets in terms of battery life. How exactly is AMD's offering supposed to be competitive when it's using that much more power in an application where maximum potential performance is typically irrelevant? Sadly there's not a whole lot of independent information regarding Temash system power consumption, else it would be more clear where it actually stands in relation to both the current Atom and Haswell ULT.

 

[Abwx]

Not going to try and claim otherwise with respect to performance, but Clover Trail tends to be roughly on-par with current ARM based tablets in terms of battery life.

Not true , see , the numbers above...

THG review :

The ATIV Smart PC 500T's six-hour battery life
is unremarkable compared to tablets with ARM-based SoCs.

 

[SiliconWars]

Not going to try and claim otherwise with respect to performance, but Clover Trail tends to be roughly on-par with current ARM based tablets in terms of battery life. How exactly is AMD's offering supposed to be competitive when it's using that much more power in an application where maximum potential performance is typically irrelevant? Sadly there's not a whole lot of independent information regarding Temash system power consumption, else it would be more clear where it actually stands in relation to both the current Atom and Haswell ULT.

The single-threaded performance of Temash is good enough that it bridges the performance gap between a decent experience and a bad one. What you get is the feeling of a much snappier chip in everyday tasks.

Vs Clover Trail and last gen ARM, a Temash tablet will feel like a far superior experience overall. Vs the new gen ARM and Bay-Trail, it'll mostly feel the same, while lacking in heavy CPU loads.

Spotted this one earlier, looks nice - http://www.youtube.com/watch?v=nyOE8jUM_5c

This makes me wish AMD had a proper tablet chip. I wouldn't buy a dual core even in a tablet, but if AMD could get a quad into 5W they'd have a huge win.

 

[Khato]

Battery is 30 Wh , so plateform comsumption is between 4 and 5.5W
depending of the usage.

If it is what you call adequate for a tablet then kabini is more than
suited to this task...

I thought I was being clear enough, but I guess I should have provided a link regardless - http://www.tomshardware.com/reviews/atom-z2760-power-consumption-arm,3387.html - aka actual detailed power consumption data.

As well, just to demonstrate what a difference test settings make (aka, how meaningless the platform consumption numbers AMD provided are)... Anandtech's battery life numbers for the Samsung ATIV Smart PC 500T - http://www.anandtech.com/show/6827/samsung-ativ-smart-pc-revisiting-clover-trail-convertibles/5 - show slightly under 3 watts for platform power usage.

 

[monstercameron]

Not going to try and claim otherwise with respect to performance, but Clover Trail tends to be roughly on-par with current ARM based tablets in terms of battery life. How exactly is AMD's offering supposed to be competitive when it's using that much more power in an application where maximum potential performance is typically irrelevant? Sadly there's not a whole lot of independent information regarding Temash system power consumption, else it would be more clear where it actually stands in relation to both the current Atom and Haswell ULT.

haswell ult have good idle power, similar to arm/atom but underload[unless it is clock limited and turbo disabled] it will eat battery like its predecessor and perform similar to temash/kabini

[i3-3229y] http://www.notebookcheck.net/Review-Acer-Aspire-P3-171-3322Y2G06as-Convertible.96583.0.html
[a6-1450] http://www.notebookcheck.net/Review-Acer-Aspire-V5-122P-61454G50NSS-Notebook.93321.0.html
[a4-5000] http://www.notebookcheck.net/Review-Packard-Bell-EasyNote-LE69KB-45004G50Mnsk-Notebook.99642.0.html
[i5-4250u] http://www.notebookcheck.net/Review-Apple-MacBook-Air-13-Mid-2013-MD760D-A-Subnotebook.96311.0.html {11W idle, 37W loaded!}

 

[Abwx]

I thought I was being clear enough, but I guess I should have provided a link regardless - http://www.tomshardware.com/reviews/atom-z2760-power-consumption-arm,3387.html - aka actual detailed power consumption data.

This is the numbers provided by intel , what i linked was
the review that came later with actual products.

Here what THG says about the numbers they
provided in your link :

Right before we published ARM Vs. x86: The Secret Behind Intel Atom's Efficiency, Intel let us know that new firmware and drivers were available for Samsung's tablet that should help its battery life. In fact, representatives mentioned, vendors continue tweaking their software infrastructures to further optimize what they're able to get from available [COLOR=blue !important][FONT=inherit !important][COLOR=blue !important][FONT=inherit !important]battery [/FONT][/FONT][FONT=inherit !important][COLOR=blue !important][FONT=inherit !important]power[/FONT][/COLOR][/FONT][/COLOR][/COLOR]. So, we went back and updated any driver versions that had been released after our first round of numbers was run. H.264-encoded playback went up by 30 minutes, and our Web browsing workload picked up 23 minutes, just about tying Microsoft's Surface.
Now, that's not quite as impressive as what Intel showed us in its lab using Acer's Windows 8-based tablet.

http://www.tomshardware.com/reviews/ativ-smart-pc-500t-windows-8-atom,3360-11.html

 

[blackened23]

Since the bay trail naming schemes for all of the different SKUs are so confusing -

does anyone have a summary of where this one falls into the mix? Is this one a celeron? Pentium? I know there are a ton of Bay Trail SKUs with a variety of performance charateristics.

 

[Khato]

haswell ult have good idle power, similar to arm/atom but underload[unless it is clock limited and turbo disabled] it will eat battery like its predecessor and perform similar to temash/kabini

[i3-3229y] http://www.notebookcheck.net/Review-Acer-Aspire-P3-171-3322Y2G06as-Convertible.96583.0.html
[a6-1450] http://www.notebookcheck.net/Review-Acer-Aspire-V5-122P-61454G50NSS-Notebook.93321.0.html
[a4-5000] http://www.notebookcheck.net/Review-Packard-Bell-EasyNote-LE69KB-45004G50Mnsk-Notebook.99642.0.html
[i5-4250u] http://www.notebookcheck.net/Review-Apple-MacBook-Air-13-Mid-2013-MD760D-A-Subnotebook.96311.0.html {11W idle, 37W loaded!}

Exactly, the only data points we have for Kabini/Temash show minimum idle system power usage (6.9 watts for the A6-1450 system, 6.5 watts for the a4-5000 system) well above what AMD promised. Whereas for the Haswell based Macbook Air the minimum idle power consumption under OSX is a paltry 2.8 watts... and most likely Windows will be similar once 8.1 comes along with proper support for Haswell's power optimizations. I'm not ready to rule out that the high idle power is just due to bad system implementation, but even AMD's marketing materials don't claim levels on par with Haswell.

 

[monstercameron]

Exactly, the only data points we have for Kabini/Temash show minimum idle system power usage (6.9 watts for the A6-1450 system, 6.5 watts for the a4-5000 system) well above what AMD promised. Whereas for the Haswell based Macbook Air the minimum idle power consumption under OSX is a paltry 2.8 watts... and most likely Windows will be similar once 8.1 comes along with proper support for Haswell's power optimizations. I'm not ready to rule out that the high idle power is just due to bad system implementation, but even AMD's marketing materials don't claim levels on par with Haswell.

To be fair the 6.9 and 6.5 numbers are platform idle numbers for quad core chip with usb3 and higher clocks.who is to say that the a4 won't be much lower with tablet optimized components like ssds and 1.25v ram -among others.