Review Index:

AMD Unveils Beema and Mullins: A Greater than Expected Refresh of Kabini

Subject: Processors
Manufacturer: AMD

AMD Makes some Lemonade...

I guess we could say that AMD has been rather busy lately.  It seems that a significant amount of the content on PC Perspective this month revolved around the AMD AM1 platform.  Before that we had the Kaveri products and the R7 265.  AMD also reported some fairly solid growth over the past year with their graphics and APU lines.  Things are not as grim and dire as they once were for the company.  This is good news for consumers as they will continue to be offered competing solutions that will vie for that hard earned dollar.

View Full Size

AMD is continuing their releases for 2014 with the announcement of their latest low-power and mainstream mobile APUs.  These are codenamed “Beema” and “Mullins”, but they are based on the year old Kabini chip.  This may cause a few people to roll their eyes as AMD has had some fairly unimpressive refreshes in the past.  We saw the rather meager increases in clockspeed and power consumption with Brazos 2.0 a couple of years back, and it looked like this would be the case again for Beema and Mullins.

It isn’t.

I was again expecting said meager improvements in power consumption and clockspeeds that we had received all those years ago with Brazos 2.0.  Turns out I was wrong.  This is a fairly major refresh which does a few things that I did not think were entirely possible, and I’m a rather optimistic person.   So why is this release surprising?  Let us take a good look under the hood.

Click here to read the entire Beema/Mullins introduction!

Process and Design

The previous generation of low power APUs is based on the Kabini and Temash designs.  Kabini/Temash is comprised of Jaguar CPU cores and GCN graphics/compute cores.  These products spanned the 4W to 25W TDP range, and did so very well.  It competed with the latest Intel Bay Trail parts fairly effectively (beating Bay Trail out in graphics, but pulling more power overall).  These products had a fair amount of success in the marketplace, and it culminated with the aforementioned AM1 platform.  While these products are good, they just could not overcome the power and thermal advantages that Intel brought to the table with Bay Trail.

View Full Size

So AMD went back to the drawing board.  Instead of doing a clean sheet design that would take years to complete, they went over Kabini/Temash with a fine-toothed comb and extracted every little bit of performance and efficiency out of the part.  They also introduced new software technology that helps the APU address multiple workloads in the fastest and most efficient manner it could.  AMD also appears to have worked very closely with the foundry people to extract the greatest amount of improvements possible from the 28 nm HKMG node.

Currently Intel is the only semiconductor manufacturer that is using a sub-28 nm process in any kind of volume.  Yes, there are a few 20 nm products from other manufacturers in production, but they have not reached the market so far.  20 nm production and below has been problematic for the industry (except for Intel) due to a variety of reasons.  We do not expect 20 nm products from pure-play foundries and their partners until 2H 2014, and those look to be primarily low power parts and SOCs.  So to have any refresh product in 2014, these designers have to use 28 nm HKMG and its derivatives.  For quite a while TSMC was the only pure-play foundry offering 28 nm, but GLOBALFOUNDRIES has finally gotten their process running at a decent pace.  GLOBALFOUNDRIES appears to be the manufacturer of the 25 watt Kabini AM1 products for AMD.

View Full Size

AMD has not confirmed that GF is producing the latest Beema and Mullins chips, but there is a very good chance that they are.  One thing that GF does not have that TSMC does is a lot of customers vying for space on process lines.  This does not allow designers a lot of engineering time with foundry engineers who could work with individual designs to improve thermals and clockspeeds if there is great demand for line space.  There is a certain amount of work being done by foundry engineers to improve yields and bins, but demand does not allow as in-depth work as is desired to be done.  There just is not enough manpower, money, and time to effectively do this.

If GF is in fact producing these parts, it looks like they have worked extensively with AMD to more adequately optimize Beema/Mullins for their 28 nm process.  Now, this is complete speculation based on factors such as line space availability, GF opening up their 28 nm process, and the pretty significant improvements found with the latest Beema/Mullins APUs.  Needless to say, even if TSMC is the primary foundry for this APU release, a lot of work has been done at 28 nm to get these improvements.

View Full Size

AMD engineers have also played a key role in the improvements we see.  The Beema/Mullins parts are based on Puma+ x86 cores.  This core is functionally identical to the earlier Jaguar, but it is again highly optimized to improve clock speeds and power consumption.  There was some work done on the GPU side, but I do not think nearly as extensively as what we see from the CPU side.  All indications point to process improvements and the lower TDP Puma+ cores allowing more thermal headroom for the GPU portion to be clocked significantly higher.

So how much faster are these new parts?  Well, they are quite a bit faster clocked than the previous generation.  What is more significant is that they can clock higher while achieving much LOWER TDPs.  The results so far look almost like a half-node jump in thermals and clockspeeds, but these chips are still produced on the 28 nm HKMG that was introduced almost two and a half years ago.  I must reiterate, these chips are functionally identical (except for one major feature) to the previous Kabini/Temash parts.  The basic design is a quad core CPU with 2 x GCN compute cores (128 total stream units).

View Full Size

The top-end SKU is the 15 watt A6-6310.  This number has a top clock speed of 2.4 GHz for the four CPU cores, which is up 400 MHz from the 25 watt TDP Kabini parts.  The GPU portion runs at 800 MHz, a full 200 MHz faster than the aforementioned Kabini.  So, we have the same process node, but the TDP goes from 25 watts down to 15 watts, and the clockspeeds for the CPU and GPU portions rise pretty dramatically.  Progress!  That 15 watts also allows the memory controller to address memory at speeds up to DDR-3 1866.

The top “efficiency” SKU is the A10 Micro-6700T.  This is still a quad core part with 2 x GCN compute cores (128 stream units) with impressive numbers.  It has a max clock of 2.2 GHz for the CPU and 500 MHz for the GPU.  This part in particular has a rated TDP of 4.5 watts.  It also supports DDR3L-1333 memory (L stands for low-power).  This product has a SDP of 2.8 watts (Scenario Design Power).  Temash did not get anywhere near close to these numbers, though they were in the 4 watt TDP ballpark.

April 29, 2014 | 03:04 AM - Posted by JohnGR

Bravo AMD! :D

I only hope to see these parts to be compatible with existing AM1 boards. A scenario in which only a BIOS update will be enough would be great, but there seems to be many upgrades to the cores, especially that PSP(which unfortunately is not a hardware PSP emulator :p).
I am almost sold to the idea of buying a Sempron 3850 to become my next HTPC, but Beema's announcement and all the question marks around AM1 platform and Beema, is holding me back temporarily.

April 29, 2014 | 03:16 AM - Posted by Anonymous (not verified)

man ddr3 again, this sux, these apu will never make it if they dont change the memory.

April 29, 2014 | 03:23 AM - Posted by JohnGR

Do you have the money to give them, so they can push DDR4 first into the market? If you don't then AMD is doing the right thing, waiting for Intel to open the road and then just walk on it, free of charge.

April 29, 2014 | 07:12 AM - Posted by Anonymous (not verified)

yea but the point of the APU is the graphics, and gddr3 limits it alot, at least for desktop, too bad that mddr thingy didnt work could have made APU way more attractive on desktop and maybe higher end.
but am guessing with the lack of good memory AMD is focused on laptops and mobiles anyway and low end desktop, hope AMD brings something good on 2015 for desktop

April 29, 2014 | 08:52 AM - Posted by JohnGR

128 stream processors aren't going to play Crysis any time soon. So, more performance in the GPU part would have been good, but maybe not enough to convince people that this could be a low power gaming platform. It would gather much more positive opinions, but maybe not so many buyers.
In the case of cpu performance, dual channel usually gives an extra 3-5% over single channel. So faster memory there would also help, but not much.

In the end you just look at these facts and consider the expenses from your side to push a brand new memory or the extra cost that will be added to the platform by using a new memory standard or implementing dual channel. That could eliminate maybe it's biggest advantage which is, enough performance at ridiculously low price.

AMD did all the right choices with this platform in my opinion.

April 29, 2014 | 09:25 AM - Posted by Anonymous (not verified)

DDR3 is not the same as GDDR3. GDDR3 is based on DDR2 while DDR3 is a the base for GDDR5. DDR3 = DDR3 even when used for graphics.

April 29, 2014 | 03:53 AM - Posted by Silver Sparrow

Wow I am amazed by what has been written, love to see this new iteration in real world reviews soon. AMD could potentially be powering my new HTPC and tablet.

But part of me after reading all this is saying: Please apply this level of detail/work to your enthusiast level APU aspirations (should they exist)! I would love to see a 12/16 core high end consumer APU!

April 29, 2014 | 12:13 PM - Posted by Anonymous (not verified)

2 to 16 cores would sell to people wishing to do rendering workloads, but the Beema SKUs coud be great for a render farm, the A6 6310 on a low cost motherboard would be an affordable low power SKU with 4 or 5 boards(AM1 style LANed up into a small low cost cluster configuration). Can't wait to see what the Puma cores can do as far as raytracing and rendering, with their refinements over the previous Jaguar cores. I only hope that the included ARM security core is not used to permanently lock any Tablet hardware to a closed ecosystem! Hopefully the low power Mullins cores are not restricted to OEMs only, people are designing working smartphones around the Raspberry PI, with the off the shelf parts that are available for that platform. Hopefully there will be future SOCs with more cores, once the 14nm process becomes the standard.

April 29, 2014 | 12:17 PM - Posted by john22 (not verified)

The claim that Intel has a 'smaller' process is hilarious. There are ZERO industry standards for defining a 'process'- a fact Intel exploits when describing its chips as 22nm.

Ask yourself a question. Why is Intel's transistor density (transistors per mm2) LOWER on its 22nm process than TSMC's 28nm half-node process? And Intel uses FinFETS, which are smaller than traditional transistors anyway.

Intel's process is good for exactly one thing- making power efficient x86 parts. Other types of chip on Intel's process have proven to be terrible, which is why no major ARM SoC company has partnered with Intel to use its incredible amounts of spare foundry space. You think Nvidia, for instance, wouldn't jump ship from TSMC to Intel, if Intel could actually fabricate a decent non-x86 chip?

April 29, 2014 | 02:52 PM - Posted by Anonymous (not verified)

Nvidia is partnering with IBM, and IBM's technology foundations, including the IBM's OpenPower consortium. I would expect if Nvidia Licenses the Power IP/ISA from IBM that Nvidia could have a more powerfull CPU competitor and chip fab process by working with Samsung and GlobalFoundrys(Both part of IBM's technology shareing consortium/foundations). NvLink being derived from IBM's CAPI (Coherent Accelerator Processor Initiative). IBM is spreading a lot of its chip fab IP, and power ISA IP around these days, in preperation of IBM getting out of the fab business(except for IBM's research FAB unit). IBM's chip fab process and technology shareing has helped both Samsung and GlobalFoundrys, and will provide IBM with a steady supply of power based CPU parts at the 14nm node, this IBM tech sharing will have side benifits for both AMD and Nvidia, and anyone that uses Samsung or GlobalFoundrys for their foundry work. Hopefully AMD will bring HSA and hUMA to their next iteration after the Beema and Mullins SKUs.

April 29, 2014 | 04:34 PM - Posted by Anonymous (not verified)

I hope these new chips will make my linux gaming rig fly! Can't wait for the release, been wanting a linux based laptop for a long time.

April 30, 2014 | 05:09 AM - Posted by Anonymous (not verified)

I had an AMD CPU BACK IN 1999-early 2000`s

May 8, 2014 | 10:30 PM - Posted by ai5000 (not verified)

Any word on availability or pricing?