AMD Beema/Mullins Launch Thread

[NostaSeronx]

Jaguar/Puma Core/FPU Resources:

Core:
1 ALU + 1 ALU/Mul/Div + 1 Load AGU + 1 Store AGU
FPU:
1 VALU/ViMUL/FPADD + 1 VALU/StoreConv/FPMUL

Bulldozer Core/FPU Resources:

Core:
1 ALU + 1 ALU/Mul + 2 AGLUs

FPU:
1 VFMA/Fconv/VI + 1 VFMA/Flogic + 1 VALU + 1 VALU/Fstore

Piledriver Core/FPU Resources:

Core:
1 ALU/Div + 1 ALU/Mul + 2 AGLUs/Mov

FPU:
1 VFMA/Fconv/VI + 1 VFMA/Flogic + 1 VALU + 1 VALU/Fstore

SteamrollerB Core/FPU:

Core:
1 ALU/Div + 1 ALU/Mul + 2 ALUs + 4 AGLUs/Mov

FPU:
1 VFMA/Fconv/VI/VALU + 1 VFMA/FLL + 1 VALU/Fstore/Fshuf

(KV-A1 has the 2 simple ALUs units and 2 of the four AGLU/Mov units disabled)
---
IPC doesn't care about bit size capability so I am ignoring implementing those.

128b Vector vs 256b Vector = 1 instruction, 4 GFlops vs 8 GFlops, 2x GFlops if FMA.

---
@IDC, the ML-A0 samples came out within the RHS Box for 28nm FDSOI. I noticed the same behavior for KV-A0 samples that were in the RHS Box for 28nm SHP. Beema and Mullins are most likely on the 28nm FDSOI node for easy implementation of per core Wide DVFS.

ML-A1, A10 Micro-6700T, 4W;
Idle/Gated: 0.25 volts, 1 GHz
Base: 0.575 volts, 1.6 GHz
Boost: 0.925 volts, 2.2 GHz
^-- pretty wide range.

 

[Khato]

This is completely contradicted by this :

AMD claims to have achieved a 19% leakage reduction across the CPU cores and a 38% leakage reduction in the integrated GPU.

http://techreport.com/review/26377/a-first-look-at-amd-mullins-mobile-apu

One must take account that it s no more the same process and that either 30% higher absolute frequency or 40% lower power drain at the same frequencies is what apparently differ from the two iterations, not negating that there was some rework that helped extract more perfs from the new silicon.

Completely contradicted? That's a pretty confident statement especially when your quote is only a claim. (Especially since reducing the base frequency at which those leakage figures can be taken would result in those manner of leakage reduction because, ya know, lower voltage results in less leakage when all else is held constant.)

I will admit that I was intentionally going a bit far in the opposite direction - if AMD's PR materials are to be believed they did make minor improvements with respect to power efficiency. But the massive improvement that many are extrapolating from the demonstrated performance combined with the reduction in TDP? Well, feel free to believe it if you want, but all indications are that the TDP is when running at base frequencies, not the turbo frequencies that result in its admirable benchmark performance.

Also I'll freely admit that there's a chance I'm wrong - we don't yet have the necessary data to come to a firm conclusion, can only make educated guesses.

 

[blackened23]

That's the confusing thing to me. Of all the benchmarks, from what I saw, literally none of them have performance per watt or battery life benchmarks. That means that all of the benchmarks, in a mobile context, are meaningless. Battery life is the #1 metric for mobile devices. If this performance is achieved with 4 hours of battery life, guess what? Worthless mobile device.

You can literally take nearly any architecture from the past year and make it perform well, say damned with battery life, and have it perform incredibly well. You scale the clockspeeds and turbo up and down dependent on what battery life you're aiming for. Most competitive SOCs do 8+ hours fairly easily in typical workloads. So the real question is at what performance per watt does it do so at? What battery life? We don't have the answers to these questions, and while the benchmarks are impressive. Like I said. Meaningless without the context of battery life in actual real world devices. They could have a SKU that does full turbo with 4 hours of battery life which of course is worthless. The real question is, does it achieve competitive battery life in similarly sized form factors compared to other SOCs from qualcomm, nvidia, intel, etc.

So again we're in a situation where AMD is intentionally hiding that information, which leads one to believe......if it walks like a duck, talks like a duck, it's probably a duck. Probably not good.

The other problem is lack of android compatibility, but that's minor compared to the lack of battery life context. If there's a REPUTABLE website that isn't pure AMD marketing that has this information, do let me know. I didn't see battery life at toms, LR, or here. So. Yeah. If AMD is intentionally hiding it. What's the conclusion? The turbos are scaled up to unrealistic levels that result in poor battery life, or good battery life levels (8 hours+) would require lower turbo, no turbo, or lower clockspeeds. That's just the facts. These benchmarks are meaningless without battery life context. Because performance can go WAY DOWN in actual devices depending on what battery life is desired; if performance per watt sucks, then the performance will go down for more battery life.

I mean, if this performance is attainable with great battery life, i'd say good on AMD. It would appear to be a very good SOC in that case. But, they didn't provide it - they're hiding that information, so we don't know and can only assume the worst since the information is intentionally being with-held.

 

[Arachnotronic]

That's the confusing thing to me. Of all the benchmarks, from what I saw, literally none of them have performance per watt or battery life benchmarks. That means that all of the benchmarks, in a mobile context, are meaningless. Battery life is the #1 metric for mobile devices. If this performance is achieved with 4 hours of battery life, guess what? Worthless mobile device.

You can literally take nearly any architecture from the past year and make it perform well, say damned with battery life, and have it perform incredibly well. You scale the clockspeeds and turbo up and down dependent on what battery life you're aiming for. Most competitive SOCs do 8+ hours fairly easily in typical workloads. So the real question is at what performance per watt does it do so at? What battery life? We don't have the answers to these questions, and while the benchmarks are impressive. Like I said. Meaningless without the context of battery life in actual real world devices. They could have a SKU that does full turbo with 4 hours of battery life which of course is worthless. The real question is, does it achieve competitive battery life in similarly sized form factors compared to other SOCs from qualcomm, nvidia, intel, etc.

So again we're in a situation where AMD is intentionally hiding that information, which leads one to believe......if it walks like a duck, talks like a duck, it's probably a duck. Probably not good.

The other problem is lack of android compatibility, but that's minor compared to the lack of battery life context. If there's a REPUTABLE website that isn't pure AMD marketing that has this information, do let me know. I didn't see battery life at toms, LR, or here. So. Yeah. If AMD is intentionally hiding it. What's the conclusion? The turbos are scaled up to unrealistic levels that result in poor battery life, or good battery life levels (8 hours+) would require lower turbo, no turbo, or lower clockspeeds. That's just the facts. These benchmarks are meaningless without battery life context. Because performance can go WAY DOWN in actual devices depending on what battery life is desired; if performance per watt sucks, then the performance will go down for more battery life.

I mean, if this performance is attainable with great battery life, i'd say good on AMD. It would appear to be a very good SOC in that case. But, they didn't provide it - they're hiding that information, so we don't know and can only assume the worst since the information is intentionally being with-held.

Shhh!!! AMD won the benchmarks! That's all that matters! ;-)

Don't thread crap.
-ViRGE

 

[Enigmoid]

That's the confusing thing to me. Of all the benchmarks, from what I saw, literally none of them have performance per watt or battery life benchmarks. That means that all of the benchmarks, in a mobile context, are meaningless. Battery life is the #1 metric for mobile devices. If this performance is achieved with 4 hours of battery life, guess what? Worthless mobile device.

You can literally take nearly any architecture from the past year and make it perform well, say damned with battery life, and have it perform incredibly well. You scale the clockspeeds and turbo up and down dependent on what battery life you're aiming for. Most competitive SOCs do 8+ hours fairly easily in typical workloads. So the real question is at what performance per watt does it do so at? What battery life? We don't have the answers to these questions, and while the benchmarks are impressive. Like I said. Meaningless without the context of battery life in actual real world devices. They could have a SKU that does full turbo with 4 hours of battery life which of course is worthless. The real question is, does it achieve competitive battery life in similarly sized form factors compared to other SOCs from qualcomm, nvidia, intel, etc.

So again we're in a situation where AMD is intentionally hiding that information, which leads one to believe......if it walks like a duck, talks like a duck, it's probably a duck. Probably not good.

The other problem is lack of android compatibility, but that's minor compared to the lack of battery life context. If there's a REPUTABLE website that isn't pure AMD marketing that has this information, do let me know. I didn't see battery life at toms, LR, or here. So. Yeah. If AMD is intentionally hiding it. What's the conclusion? The turbos are scaled up to unrealistic levels that result in poor battery life, or good battery life levels (8 hours+) would require lower turbo, no turbo, or lower clockspeeds. That's just the facts. These benchmarks are meaningless without battery life context. Because performance can go WAY DOWN in actual devices depending on what battery life is desired; if performance per watt sucks, then the performance will go down for more battery life.

I mean, if this performance is attainable with great battery life, i'd say good on AMD. It would appear to be a very good SOC in that case. But, they didn't provide it - they're hiding that information, so we don't know and can only assume the worst since the information is intentionally being with-held.

This is 100% true.

Its easy to see that under heavy loads the chip tested is running base clocks (1.2 ghz). I have no idea how this relates to power consumption but the turbo is extremely aggressive (good).

I was talking about the Quad Core A10 Micro-6700T. I dont have Data for the Dual Core.

Except the 2.8 SDP Mullins is FASTER than the Baytrail 3740.

Yep thats just "miserably weak".

Lets see:

web browseing tests:
Sunspider 1.0 = Mullins faster than 3740
Mozilla Kracken benchmarkk = mullins faster than 3740
WebXPRT = mullins faster tahn 3740

Cpu tests:
Cinebench 11,5 single thread = mullins faster than 3740.
Cinebench 11,5 multi thread = mullins faster than 3740.

You make it sound like this new chip from AMD is weak compaired to the Baytrail,
when infact its faster both in CPU tasks and in GPU ones.

Check out the review Anandtech did of it, or any of the other review sites that benchmarked it.

My context was in reference to the tablet mentioned which runs a dual kabini downclocked to 600 mhz. Even AMD's own marketing slides don't say that a quad core with 2.2 ghz turbo is going to fit in that thermal envelope as that more than a quadrupling at the same (estimated) power use.

Seeing that a 600 mhz dual kabini fit in the required thermal and power envelope for the tablet a doubling of speed or performance is a dual kabini at 1.2 ghz which is weaker than a quad baytrail.

I have no idea how much power this chip is using (no one does at this point) but baytrail is likely ahead in terms of perf/W. 22nm trigate vs. 28 nm planar is hard to overcome.

 

[cbn]

But for amd moving this to gf is a huge gain for the wsa agreement and their economy. Add this product give them a stronghold for the low end laptop and desktop and it looks like a solid win. No way are they going into tablets with this because of Intel paying their way no matter the product and we still dont know battery life for this. My guess is Intel still holds a very comfortable lead here but amd gained the performance crown at the laptop and desktop level.

I have no idea how well the lowest end Mullins APU (E1-Micro 6200T, 3.95 watt TDP dual core @ 1.4 Ghz) will do for tablet wins, but it would have been very interesting for Mini-ITX NAS (if only it had more SATA ports)

Instead I see Asrock is releasing the C70M1 (C70 Brazos APU with four native SATA ports)........ in 2014;

http://www.asrock.com/mb/index.asp?s=AMD CPU

http://yournewsticker.com/2014/03/mini-itx-board-dual-asrock-c70m1-generation-apu-amd-ontario.html

A new Brazos product fron a major OEM in 2014 is pretty surprising. Now granted it is just the existing Asrock E350M1 with the C70 replacing the E350 but who would have guessed we'd see Brazos still being introduced this late several years ago?

P.S. As far as I can tell the C70M1 appeared on the Asrock website in March and has made it to several retailers starting this month.

 

[VirtualLarry]

Instead I see Asrock is releasing the C70M1 (C70 Brazos APU with four native SATA ports)........ in 2014;

http://www.asrock.com/mb/index.asp?s=AMD CPU

http://yournewsticker.com/2014/03/mini-itx-board-dual-asrock-c70m1-generation-apu-amd-ontario.html

A new Brazos product fron a major OEM in 2014 is pretty surprising. Now granted it is just the existing Asrock E350M1 with the C70 replacing the E350 but who would have guessed we'd see Brazos still being introduced this late several years ago?

P.S. As far as I can tell the C70M1 appeared on the Asrock website in March and has made it to several retailers starting this month.

Since the E-350 and C-70 are essentially the same silicon, only the C-70 is downbinned, could this basically be a model of motherboard, designed to absorb the leftover CPUs that AMD had lying around?

 

[cbn]

Since the E-350 and C-70 are essentially the same silicon, only the C-70 is downbinned, could this basically be a model of motherboard, designed to absorb the leftover CPUs that AMD had lying around?

Yes, I am thinking they must be leftover chips.

P.S. Notice C70M1 has some rear I/O stripped compared to the E350M. Otherwise, the rest of the layout is looks exactly the same between the two boards.

C70M1

E350M1

 

[OCGuy]

Anyone buying?

 

[Enigmoid]

Anyone buying?

If there are no power consumption/battery caveats I will buy a tablet. Probably when 'back to school' hits and there is availability.

 

[Vesku]

Come winter sales I'll be interested if it manages to find its way into cheaper tablet mode capable 12-14 inch 1080P notebooks.

 

[Abwx]

Completely contradicted? That's a pretty confident statement especially when your quote is only a claim. (Especially since reducing the base frequency at which those leakage figures can be taken would result in those manner of leakage reduction because, ya know, lower voltage results in less leakage when all else is held constant.)

I will admit that I was intentionally going a bit far in the opposite direction - if AMD's PR materials are to be believed they did make minor improvements with respect to power efficiency. But the massive improvement that many are extrapolating from the demonstrated performance combined with the reduction in TDP? Well, feel free to believe it if you want, but all indications are that the TDP is when running at base frequencies, not the turbo frequencies that result in its admirable benchmark performance.

Also I'll freely admit that there's a chance I'm wrong - we don't yet have the necessary data to come to a firm conclusion, can only make educated guesses.

Leakage is proportional to voltage, to reduce CPU leakage by 19% and GPU leakage by 40% they sould had reduced voltage by as much following your logic wich is of course completely wrong given thoses numbers.

So far AMD has been extremely honnest by specyfing TDP instead of being stuck on SDP, you know the metrics that allow some kind of J1900 to be claimed as 5W when running with 4 cores at 2.2GHz, i guess that then everything has to be believed despite all evidence saying otherwise.

 

[Abwx]

baytrail is likely ahead in terms of perf/W. 22nm trigate vs. 28 nm planar is hard to overcome.

"Likely" is neither a number nor any metric that would be a valuable argument,
wich render the rest of your sentence without any sense.

 

[blackened23]

Yeah, it's just too bad AMD is intentionally hiding battery life. That would give us the answer, but with AMD hiding it, i'm sure THAT's a "valuable argument". If AMD is hiding it, that means it isn't good, period. That means real devices will have to lower turbo or do away altogether for proper battery life in mobile devices. It seems AMD prohibited reviewers from even testing the MOST VALUABLE mobile metrics of all.

 

[el etro]

One thing to look at is that beema top bin is stated as "Up to 2.4Ghz" clock, but don't turbo to this frequency the whole time.


Anyway, seems AMD made a big progress in efciency matter without an architectural change and this is awesome. Even if Mullins fails on mobile front is a big win for AMD, once is their first strike on this market.

 

[Abwx]

Yeah, it's just too bad AMD is intentionally hiding battery life. That would give us the answer, but with AMD hiding it, i'm sure THAT's a "valuable argument". If AMD is hiding it, that means it isn't good, period. That means real devices will have to lower turbo or do away altogether for proper battery life in mobile devices. It seems AMD prohibited reviewers from even testing the MOST VALUABLE mobile metrics of all.

This post should be saved as a blatant exemple of thread crapping, i see nothing else that rambling and unsubstancied accusations.

We'll decide what thread crapping is, thank you very much. And to be honest you're closer than blackened right now
-ViRGE

Did Intel release battery life numbers the first day they released BT, btw.?..

Anyway, seems AMD made a big progress in efciency matter without an architectural change and this is awesome. Even if Mullins fails on mobile front is a big win for AMD, once is their first strike on this market.

A few months ago i pointed that AMDs CTO Mark Papermaster, who is a very high level enginer, is not a marketing man and that as a purely technical guy he should be listened to seriously when making claims such as this one :

“AMD is establishing excellent momentum this year in the low-power, mobile computing market and with ‘Mullins’ and ‘Beema’ coming in 2014 we are not standing still,” said Mark Papermaster, AMD’s chief technology officer and senior vice president, during his closing keynote at APU13. “AMD aims to deliver a set of platforms in the first half of next year that will outperform the competition in graphics and total compute performance in fanless tablets, 2-in-1s and ultrathin notebooks.

http://wccftech.com/amd-unveils-beema-mullin-rival-bay-trail-2014-apu-roadmap-revealed/

Guess that many people do not pay attention when relevant people make
interesting claims, so far he s proven right.

 

[Vesku]

Yeah, it's just too bad AMD is intentionally hiding battery life. That would give us the answer, but with AMD hiding it, i'm sure THAT's a "valuable argument". If AMD is hiding it, that means it isn't good, period. That means real devices will have to lower turbo or do away altogether for proper battery life in mobile devices. It seems AMD prohibited reviewers from even testing the MOST VALUABLE mobile metrics of all.

TBH, it's not as if AMD's reference platform would matter much for helping figure out what purchasable devices will be like. Recall Acer's launch A6-1450 was a power hog compared to other A6-1450 products. The preview shows us performance is decent, real power analysis of shipping products will fill in the rest of the picture.

 

[frozentundra123456]

"Aims to" is a pretty ambiguous statement. I never heard an executive say they "aimed to" produce a product inferior to the competition. Nothing is "proven" yet either, since we have no tests of power usage or battery life in a more realistic form factor.

AMD may in fact have a great product, but we need objective tests of battery life, temperatures, and power consumption in real world form factors to know that, not the statement of some company executive, whether he is an engineer or not.

 

[Phynaz]

Papermaster completely skirted CPU performance.

 

[cbn]

I wonder when/if these are coming to desktop?

A mini ITX platform with one of these would be pretty sweet. Will the stricter memory standards mean it is impossible for these to come out for desktop now?

http://www.anandtech.com/show/7974/...hitecture-a10-micro-6700t-performance-preview

AMD also went in and tweaked the SoC’s memory interface. Kabini/Temash had a standard PC-like DDR3 memory interface. All of the complexity required for broad memory module compatibility and variations in trace routing was handed by the controller itself. This not only added complexity to the DDR3 interface but power as well. With Beema and Mullins, AMD took a page from the smartphone SoC design guide and traded flexibility for power. These platforms now ship with more strict guidelines as to what sort of memory can be used on board and how traces must be routed. The result is a memory interface that shaves off more than 500mW when in this more strict, low power mode. OEMs looking to ship a design with socketed DRAM can still run the memory interface in a higher power mode to ensure memory compatibility.

These SoCs won’t be available in a PoP configuration unfortunately - OEMs will have to rely on discrete DRAM packages rather than a fully integrated solution. Beema/Mullins also show up to a 200mW reduction in power consumed by the display interface compared to Kabini/Temash.

 

[Abwx]

"Aims to" is a pretty ambiguous statement. I never heard an executive say they "aimed to" produce a product inferior to the competition. Nothing is "proven" yet either, since we have no tests of power usage or battery life in a more realistic form factor.

AMD may in fact have a great product, but we need objective tests of battery life, temperatures, and power consumption in real world form factors to know that, not the statement of some company executive, whether he is an engineer or not.

Dont expect an enginer to use selected words...

They have specified the TDPs so far, the SDPs are meaningless in respect of Intel s own SDP since we dont know in both cases what are the tasks that yield those numbers.

 

[piesquared]

Papermaster completely skirted CPU performance.

Which AMD wins in there also. bay trail just got kicked into the low end budget, low quality category.

 

[frozentundra123456]

You mean amd territory?

 

[PPB]

You mean amd territory?

You make it sound bad, to have a territory where you actually do good. That is actually quite better than trying to penetrate someone else's territory year after year, and fail miserably, even with contra-revenue.

 

[piesquared]

You mean amd territory?

Nope it's now known as intel's territory, AMD has the premier product for premier devices. intel is selling theirs for bargain bin products and bargain bin prices.