CES 2014: more details on Mullins/Beema

[Abwx]

Sorry about the typo. I meant Mullins and Temash.

Just saying that comparisons must be at the same power level (approximation with TDP is much less accurate but doable). Compare Intel mobile SKU's. Dual 35W part vs. Quad 35W part vs. Quad 45W part. Where do you think the greatest efficiency is even though they are all on the arch?

(Offtopic the relatively high power use of 128 GCN cores at 600 mhz is why I'd argue that nvidia has no chance of getting 192 kepler cores @ 950 mhz in a 4W window).

AMD is comparing against TDP (though the same core config). Its a poor way to do unless you are looking at the same TDP levels (really CPU power draw should be used).

Just compare the A10-6800k to the a10-6700. Nowhere near the same efficiency.

Curious that Intel Bay trail didnt require a comparison at same TDP
than their predecessors , not to bring the intel/amd debate but
then no one talked of dishonnesty , it was , hey it perform better
and consume less...

Here it s the same , a 15W part replace a 25W one
and a 4.5W replace a 8W.

 

[Sweepr]

They are telling this every time. When AMD launches a mobile product it needs months until we really can buy the notebooks or tablets.

Tablet oriented A4-1200 (3.9W TDP, lower than Mullins's 4.5W TDP) was released 8 months ago and the smallest/lightest tablet based on it right now is an unimpressive 11.6'' MSI W20-A421 weighting 674 grams that cant even run its 4 cores at the advertised clock (capped @ mere 600MHz!) to get reasonable battery life numbers (7,5 hours @ WiFi surfing with 34 Wh battery). There's certainly a lot of room for improvement even on the same node, but I feel like Mullins should have been a 2013 chip and I'm not holding my breath for impressive design wins even this time around.

 

[inf64]

Like I've said, pure performance:

Footnotes #10 and #11. You even have BT numbers so you can search for reviews and see if they are not gimped .

 

[ams23]

Do you even read what you re confidently writing.??.

They re saying that the performance is measured at max TDP ,
how can it be otherwise since they are specifying the MAX PERFS.?

So read that the scores are for 15 watts thermal dissipation , actualy
it s very clear , dont brand something silly because you didnt understand
the obvious.

Considering the grammar and spelling errors above, you are one to talk Anyway, they never said PEAK power consumed or AVERAGE power consumed while running the application. They arbitrarily assigned a TDP value.

There is no need to be obtuse about it. Power consumption varies by application. If they don't list performance AND power consumed using the exact same application, then any notion of obtaining an accurate perf. per watt is gone.

 

[Abwx]

Considering the grammar and spelling errors above, you are one to talk Anyway, they never said PEAK power consumed or AVERAGE power consumed while running the application. They arbitrarily assigned a TDP value.

There is no need to be obtuse about it. Power consumption varies by application. If they don't list performance AND power consumed using the exact same application, then any notion of obtaining an accurate perf. per watt is gone.

If one sell a 25W chip he better specify its perfs measured
at 25W TDP ; why would he publish perfs measured at a 10
W power comsumption and thus publish sub par specs..?

Where would be the advantage.?

 

[ams23]

If one sell a 25W chip he better specify its perfs measured
at 25W TDP ; why would he publish perfs measured at a 10
W power comsumption and thus publish sub par specs..?

Where would be the advantage.?

Power consumption varies for each application due to the way that each application utilizes each application processor in the system. Without measuring both performance and power consumed while running an application, there is no accurate way to determine perf. per watt for that said application, simple as that.

 

[tviceman]

x86 was eclipsed by ARM long ago in sheer usage. There are over a million apps in the App store. I have no idea how many apps are on the Play store, but probably a million give or take. x86 may be the powerhouse king when it comes to all-out performance, but ARM has a stranglehold on the mobile market and even if x86 makes dents, ARM isn't going anywhere. In fact, the only reason x86 can be touted as being "better" is simply because of Intel's lead in processing. x86, when compared to similar process nodes, is inferior in performance per watt.


Get used to it.

Android works on x86 just fine.

Why are you telling me this? Where did I say Android does not work on x86?

 

[Enigmoid]

Curious that Intel Bay trail didnt require a comparison at same TDP
than their predecessors , not to bring the intel/amd debate but
then no one talked of dishonnesty , it was , hey it perform better
and consume less...

Here it s the same , a 15W part replace a 25W one
and a 4.5W replace a 8W.

Baytrail 3770 uses more power than its predecessor at max load. However, idle loads are up in battery life and efficiency is up as well. I never mentioned or saw intel comparing the two 'per tdp' as well or commented on it.

 

[AtenRa]

Just to clear this, AMD states the following.

Mullins 4-Core APU is 21% faster at PCmark 8 than BayTrail Z3770
Mullins 4-Core APU is %150 faster in 3D Mark 11 than BayTrail Z3770

Beema 4-Core APU is 28% faster at PCmark 8 than BayTrail N3510
Beema 4-Core APU is 257% faster in 3D Mark 11 than BayTrail N3510

The Performance per Watt was used by dividing the TDP values.

Mullins 4-Core APU 4.5W TDP (Home Score 1809) scores higher than Temash 8W TDP (Home Score 1343) in PCmark 8.

Beema 4-Core APU 15W TDP (Home Score 2312) scores higher than Kabini 25W TDP (Home Score 1861) in PCmark 8.

From the above we arrive at the following,

Mullins 4-Core 2W SDP is 21% faster than BayTrail Z3770 2W SDP in PC mark 8
Beema 4-Core 15W TDP is 28% faster than BayTrail N3510 7.5W TDP in PC mark 8

Mullins 4-Core 2W SDP is 150% faster than BayTrail Z3770 2W SDP in 3D Mark 11
Beema 4-Core 15W TDP is 257% faster than BayTrail N3510 7.5W TDP in 3D Mark 11

Edit: there will also be a 25W Beema

 

[lefty2]

They are telling this every time. When AMD launches a mobile product it needs months until we really can buy the notebooks or tablets.

This is the exactly same for the same for Bay Trail products

 

[ShintaiDK]

This is the exactly same for the same for Bay Trail products

When was BT launched and when could you buy the first tablet?

 

[mikk]

This is the exactly same for the same for Bay Trail products

It is not. Intel seems to have a much longer timeframe between shipment and launch hence why the gap between launch and availability is much narrower. This is nothing new and everyone who denies it is delusional. Temash launched May 2013 and it really showed off at end of 2013 in a couple of devices.

 

[inf64]

I see you guys gave up with perf./watt claims for the 1st slide in topic now that we proven you wrong with some hard facts ? Good.
I'm curious to know how intel can bridge the GPU gap with CherryTrail? We can see the gap is humongous. Even in CPU department they will be solidly behind but that is not that important. What is left to see is how well Beema and Mullins do with regards to battery life and product designs, if that is also a win then they will do well in the market.

 

[mrmt]

Even in CPU department they will be solidly behind but that is not that important. What is left to see is how well Beema and Mullins do with regards to battery life and product designs, if that is also a win then they will do well in the market.

Are you really buying AMD's ridiculous claims on performance per watt? Not even AMD themselves believe these claims, otherwise they wouldn't have to come with ridiculous metrics like performance/TDP on a disclaimer.

But even if AMD's ridiculous claims were true, it wouldn't solve Beema/Mulins fundamental problem:

- If you want more performance, you can go Core.

- If you want more battery life/lower power consumption, you can go Atom.

AMD will still be squeezed in a small niche, and I'm not even mentioning Broadwell/Cherry Trail, that comes with another 30% reduction in power consumption and probably extra power gating features.

 

[Abwx]

I see you guys gave up with perf./watt claims for the 1st slide in topic now that we proven you wrong with some hard facts ? Good.
I'm curious to know how intel can bridge the GPU gap with CherryTrail? We can see the gap is humongous. Even in CPU department they will be solidly behind but that is not that important. What is left to see is how well Beema and Mullins do with regards to battery life and product designs, if that is also a win then they will do well in the market.

I m quite curious to see how they will achieve thoses numbers
technicaly speaking , from the performances numbers provided
by the slides they ll have to increase frequency notably for both
GPU and CPU to get thoses numbers , assuming they didnt increase
the SP count of the GPU as we know that they incorporated an ARM
core so they could have as well slightly enlarged said part.

 

[Sweepr]

I see you guys gave up with perf./watt claims for the 1st slide in topic now that we proven you wrong with some hard facts ? Good.
I'm curious to know how intel can bridge the GPU gap with CherryTrail? We can see the gap is humongous. Even in CPU department they will be solidly behind but that is not that important. What is left to see is how well Beema and Mullins do with regards to battery life and product designs, if that is also a win then they will do well in the market.

Problem is, I can buy lots of Bay Trail-T based products right now while PowerPoint slides champion Mullins might not appear in the market much sooner than Cherry Trail-T based stuff and 8 months old A4 1200 is still a no show in an actual decent 8-10'' tablet.

They can easily close the 20% CPU gap just by increasing clockspeeds. Also we don't know if Cherry Trail is just a shrink or if there are architectural improvements. Perhaps its just like BT, perhaps there will be a typical ''up to 5%'' Tick perf/clock increase (perhaps more?).

Regarding the iGPU, too early to tell, but I bet that with 4x BT's EU number and a new gfx architecture they can easily match AMD's 2.5x number and maybe beat them by a fair margin. And they need to, cause if NVIDIA's claims for K1 are real then they will have 4x faster graphics performance than Bay Trail-T and totally smoke Mullins in the graphics department too (A4-5000 scores 37 FPS @ GFXBench 2.7, Mullins scores 570 pts @ 3DMark 11 according to AMD which is sligtly slower than A4-5000, Tegra K1 scores 60 FPS at GFXBench 2.7 according to NVIDIA).

 

[AtenRa]

I m quite curious to see how they will achieve thoses numbers
technicaly speaking , from the performances numbers provided
by the slides they ll have to increase frequency notably for both
GPU and CPU to get thoses numbers , assuming they didnt increase
the SP count of the GPU as we know that they incorporated an ARM
core so they could have as well slightly enlarged said part.

Turbo with better thermal/power management and perhaps better process (GloFo 28nm) suited for lower power/lower frequency APUs.

 

[lefty2]

When was BT launched and when could you buy the first tablet?

Bay Trail is launched September 11 and only now are products starting to appear - that's pretty much the same for Kabini. Now, I'm talking about laptops, not tablets. There's no point in comparing Temash to Bay Trail T, because Temash didn't really appear in any tablets.

 

[ShintaiDK]

Bay Trail is launched September 11 and only now are products starting to appear - that's pretty much the same for Kabini. Now, I'm talking about laptops, not tablets. There's no point in comparing Temash to Bay Trail T, because Temash didn't really appear in any tablets.

Wrong, you could buy BT tablets since october.

 

[lefty2]

Wrong, you could buy BT tablets since october.

You haven't even read my post properly. I'm talking about laptops and Bay Trail laptops haven't really appeared on the market until now.

 

[ShintaiDK]

You haven't even read my post properly. I'm talking about laptops and Bay Trail laptops haven't really appeared on the market until now.

You do know they didnt get launched in september like the tablet versions, right?

 

[ShintaiDK]

One a note, why do AMD yet again pick a lower binned BT in their benchmarks to screw results in their favour? Using the 2Ghz no turbo N3510 with 750Mhz GPU instead of the 2.16Ghz, 2.4Ghz turbo N3520 with a 864Mhz GPU.

 

[witeken]

I'm curious to know how intel can bridge the GPU gap with CherryTrail?

With x4 the amount of EUs (16), with a major redesigned gen8 architecture and 14nm 3D Tri-Gate transistors .

Even in CPU department they will be solidly behind but that is not that important.

At this moment, Silvermont is the fastest cpu core on the market, both in performance and perf/watt. When everyone else catches up, Intel will already have Cherry and Willow Trail.

 

[lefty2]

You do know they didnt get launched in september like the tablet versions, right?

Sorry, my mistake, indeed they were launched in November. And laptops with Bay Trail M started appearing now - which much the same lead time as Kabini laptops.

 

[frozentundra123456]

Bay Trail is launched September 11 and only now are products starting to appear - that's pretty much the same for Kabini. Now, I'm talking about laptops, not tablets. There's no point in comparing Temash to Bay Trail T, because Temash didn't really appear in any tablets.

I know the convertible Asus A300 I think it is has been on the market a couple of months. The dell venue 8 has also been. Honestly, I would not be interested in a laptop with either atom or kabini. Just too much sacrifice in performance. So I actually consider it a plus that atom has not been crammed into crap tops yet.