Intel’s Braswell Should Be Extremely Competitive

At the high end of the PC processor market, Intel (NASDAQ: INTC  ) is the undisputed champion as far as performance, power, and cost structure go. Long-time rival Advanced Micro Devices (NYSE: AMD  ) , despite its repeated and valiant efforts over the years, has yet to produce a competitive part for higher-end PCs. That said, where Advanced Micro Devices has traditionally excelled is at the low-end of the market, particularly as Intel's offerings there hadn't been all that aggressive until the fairly recent Bay Trail-M products (which have proven very competitive). The next generation part, known as Braswell, should eliminate Intel's final deficiency in this space.

Bay Trail-M/D CPU performance and power are superb – graphics performance could be better
Intel's low-cost Bay Trail-M and Bay Trail-D parts for low-cost notebooks and desktops, respectively, have proven to be very competitive on general purpose CPU performance and power consumption against the competition from AMD. Where Intel's Bay Trail isn't as competitive is in graphics performance, where AMD's IP is generally much more powerful.

That said, it appears that Intel recognizes this deficiency and with its next generation product for this space, known as Braswell, it should pretty dramatically improve graphics performance at the low end. Indeed, this product should be based on the same basic design as the company's upcoming Cherry Trail system-on-chip for higher-end Windows and Android tablets (although modified to include the various I/O required for PC functionality). This implies a brand new graphics architecture and roughly four times the graphics "cores" as the current Bay Trail-M/D products.

Source: Intel

Illustrating the competitive situation as it stands today
So far, the discussion has centered on vague notions of graphics performance comparisons. But we can actually quantify these numbers using the performance tests run on AMD's Discovery Tablet reference design based on the top-shelf Mullins silicon. In particular, let's look at the 3D Mark Unlimited graphics subscore to get a sense of how far ahead on graphics AMD is today.

Source: AnandTech

We can see that the AMD part offers a tad under twice the graphics performance of the current Intel Bay Trail processor that is ubiquitous in Windows tablets today (note that the AMD results are from a bulky reference design while Intel's chips have delivered this performance in smaller, shipping products). Of course, note that these are tablet platforms (there do not appear to be any public benchmarks of AMD's latest notebook part, so this was the best way to approximate the performance differential), but should be fairly representative of the competitive picture in the low-end of the PC/convertible market.

With Braswell offering a new GPU architecture (the Gen. 7 in Bay Trail is quite outdated), and given that Intel will be packing four times the number of graphics cores onto that die (thanks to design optimizations and a shrink to 14-nanometer), it stands to reason that Braswell-based chips should offer graphics performance well in excess of what AMD's Beema (notebook variant of the Mullins chip tested above) does.

Further, according to recently leaked roadmaps, Braswell is slated to be available in volume during Q1 2015. This should be well ahead of AMD's "Project Skybridge," which should be AMD's next shot at providing meaningfully improved graphics/processor performance. While it would stand to reason that if Intel does its job correctly its process lead should allow it to zoom past whatever AMD offers in terms of performance/watt, it is probably safer to say (in light of Intel's historical weakness in graphics) that Intel's competitive positioning with Braswell against Project Skybridge should be better than Bay Trail-M/D are against Beema today.

Foolish bottom line
Intel is apparently very serious and very enthusiastic about the low end of the PC market following the success of Bay Trail-M/D in helping to stabilize/grow PC unit volumes. If Bay Trail-M/D, which had a serious graphics performance deficiency against the AMD parts, could gain the share that it has against AMD, then Braswell should only improve that situation for Intel.

Your credit card may soon be completely worthless
The plastic in your wallet is about to go the way of the typewriter, the VCR, and the 8-track tape player. When it does, a handful of investors could stand to get very rich. You can join them -- but you must act now. An eye-opening new presentation reveals the full story on why your credit card is about to be worthless -- and highlights one little-known company sitting at the epicenter of an earth-shaking movement that could hand early investors the kind of profits we haven't seen since the dot-com days. Click here to watch this stunning video.


Read/Post Comments (7) | Recommend This Article (3)

Comments from our Foolish Readers

Help us keep this a respectfully Foolish area! This is a place for our readers to discuss, debate, and learn more about the Foolish investing topic you read about above. Help us keep it clean and safe. If you believe a comment is abusive or otherwise violates our Fool's Rules, please report it via the Report this Comment Report this Comment icon found on every comment.

  • Report this Comment On June 10, 2014, at 11:03 PM, ericcinvest wrote:

    Maybe you can tell me whats going on here. Intel is using it's best technology something that R & D spends billions on per year to produce a chip with the lowest margins on intel's balance sheet? I understand that its going to kick the crap out of amd in the low end on tablets (and anything else it puts a 14nm chip in)but why does Intel care about this really low margin business? There's only so much 14nm factory power that intel can bring to bear and the question remains is this the best use of that power?

  • Report this Comment On June 11, 2014, at 12:21 AM, ta152h wrote:

    Where are you getting that Intel has stabilized or is growing PC units? Also, what makes you think it's at the low-end?

    Also, the supposition that Bay Trail did well in CPU is unfounded. It did at very low wattages, until Puma, but the 2.05 GHz AMD beat a 2.5 GHz Intel in most benchmarks. This despite them being the same size, and Intel having a good lead in lithography technology. So, roughly, the IPC of the AMD product is about 25% higher, or, if you want to be very conservative 20% higher. That's strange when you consider the sparse GPU resources on the chip.

    AMD has been losing market share in laptops mainly, and gaining share in desktops. My guess is Intel is doing better with the Haswell, since it's so much better than Kaveri or Richland, and the power use has been lowered so it can do better in laptops.

    We'll see this quarter going into the next whether AMD's desktop initiative with Jaguar will gain back a few percent.

    But, I think fundamentally Intel is working with a failed design with the Atom line. How else can you explain the chip is the same size as the Jaguar/Puma, despite superior lithography, performs about 20% less per cycle in CPU, and gets destroyed in GPU? That's a bad design, my friend. Take it from an AMD investor that's had to witness Bulldozer, Piledriver, and Steamroller. It is what it is, and it's not good. But, there's no sense in denying what everyone can see.

    Even so, I agree that with enough transistors, and a two node lead, Intel should be able to catch or pass AMD in GPU, as it appears to be their primary focus with this design.

    But, they need to get this out soon, because Puma's successor is coming out next year, and it's not clear if they'll extend the CPU lead over Atom. I do agree, the GPU situation will probably not be as favorable now, but the CPU could be even better for AMD after both next generation parts.

  • Report this Comment On June 11, 2014, at 9:47 AM, maarten12100 wrote:

    Braswell is said to be over 40% better in terms of gpu which is good but probably nothing game changing. That would still put them behind, gpu cores consume the most power so AMD actually does better flops/W. Also Braswell and baytrail are double channel products so they have more bandwidth yet they are slower. If AMD was to make it double channel the gap would widen.

    You preach Intel's node advantage which is shrinking rapidly. The GF28SHP node showcased with Mullins/Beema that it is a better node than Intel 22.

    Kaveri also showcased the same thing the GF28SHP is vastly superior to Intel's 22nm. But there are also some disadvantages which don't plague the Intel 22 node which are that the GF28SHP node needs extremely high voltage if it want to attain high frequencies.

    The AMD 35W kaveri FX parts deliver 3/4th of the performance of their 95W desktop brother at less than 1/3 of the power. (35W is the TDP which for AMD means that the cpu and gpu combined will actually be using a tad less)

    Not only does AMD have a chip here that is better that Iris pro chips(for gaming) it does so with less power than the Iris pro monsters and without fast ram. (a GDDR5 version would obliterate Intel's Iris pro)

    Not to mention that Intel's graphics don't even work in most games and are stuttery/glitchy(Techreport Iris Pro review).

    A mid-end prototype AMD built with Kaveri pro has a 60Wh battery and delivers 12 hours of idle with the screen on and 9,5 hours of web browsing. Those are amazing numbers in terms of power which high end laptops can't even match on that same battery power with Intel. (my i7 2630qm based laptop idles 12W and that is with an SSD and 1 stick of ram and a dvd drive less)

    High end Beema laptops should be even better but I'm rather sure that those come to a point where it doesn't matter how efficient they are as the screen is gulping up 2W and the board is gulping another 3W for perphs and onboard stuf.

    The reason AMD mullins isn't in shipping tablets is because it just launched, Intel is providing contra revenue(bribery) and of course no real android.

    "it stands to reason that Braswell-based chips should offer graphics performance well in excess of what AMD's Beema (notebook variant of the Mullins chip tested above) does."

    NO it does not and furthermore the A10 micro 6700t is a Mullins chip not a Beema one do your research. (essentially the same silicon but very different markets)

    AMD is in a long term (dis)agreement with GF ever since they spun it of. Something which first was a major problem for AMD since GF was lagging behind and couldn't deliver AMD with the performance it needed. At the same time the Bulldozer design failed.(which engineers already spoke of back then)

    However now tides are changing GF's fab is showing engineering excelence and AMD is actually taping out 14nm prototypes this year. With Samsung and GF working on 14nm it is quite certain that those fabs will be better than TSMC and also compete with Intel in both wafer pricing as actually performance and power.

    Sure I'm biased since I actually own some shares of AMD but at least I'm not blind writing non factual stories with no backing.

  • Report this Comment On June 11, 2014, at 9:55 AM, maarten12100 wrote:

    misread that part about the A10 micro 6700t still it is highly doubtfull that braswell sub-4W tdp or sub-2W SDP SoCs will be able to match 128 GCN SPs running 800MHz. But if they do that is applaudable but probably too little to cut it from what we have seen from Nvidia's redesings moving to smaller cores with double cache on the 28nm node (the maxwell 750 and 750ti) we can deduct that a major increase in graphics performance will be coming with Nolan around this time next year.

  • Report this Comment On June 11, 2014, at 11:40 AM, Uconfan wrote:

    There is nothing more transient, in the transistor world than technological advantage. AMD is moving to 20 and 14 nanometer as well and Intel will have 14 nanometer bragging rites for only a short time.

    From a technical standpoint AMD just broke out of a wedge formation. Higher lows meeting a stagnant resistance level compressing the trading range. This is bullish indicator of a trend reversal. For traders on the side line this is a big buy sign. For a value investor sitting on a stock with a price to sales ratio one fifth of its competitors (myself), this justifies patience.

  • Report this Comment On June 11, 2014, at 11:46 AM, binartech wrote:

    Paying revenue(bribery) make any Intel product

    "Extremely Competitive"

  • Report this Comment On June 11, 2014, at 2:05 PM, raghu78 wrote:

    Ashraf

    Its amazing how you can turn around whats a miserable situation in graphics for Intel to a positive one. Mullins thrashes the Baytrail-T parts in GPU perf by 50 - 75%. Beema A6-6310 crushes Baytrail-D and runs at 2x the perf.

    http://techreport.com/review/26377/a-first-look-at-amd-mulli...

    http://www.tomshardware.com/reviews/athlon-5350-am1-platform...

    http://www.techspot.com/article/810-amd-beema-and-mullins-pr...

    http://www.computerbase.de/forum/showthread.php?t=1353344

    3DMark Icestorm

    A6-6310 (800 Mhz GPU) - 37617 / 3591 (CG)

    A6-5200 (600 Mhz GPU) - 31765 / 2674 (CG)

    A10-6700T(500 Mhz GPU) - 24775

    Baytrail-T - 16500

    Beema A6-6310 is 20 - 35% faster than Kabini A6-5200 due to a 33% higher GPU clock and support for faster DDR3 1866 Mhz which helps in bandwidth limited situations. Beema can play the latest AAA titles at low to medium settings at 720p. Intel's Baytrail is wholly unplayable. btw AMD's drivers are leagues ahead of Intel which still sucks. Also you forget that Mantle shows best results with low end CPUs. So the AMD SOCs will perform better on major engines like Frostbite 3, Cryengine 3 and the latest game titles which use them.

    by Q2 2015 the next gen Puma+ based 20nm Nolan will be shipping. You are looking at 256 - 384 GCN cores running with a dual channel DDR3 or LPDDR4 memory controller at 1600Mhz. At 20nm you are looking at a 1.6 - 1.7 billion transistor SOC. The clocks will be lower than Beema/Mullins so as to stay within the TDP limits. 20nm planar has half node like power efficiency gains. But given the extra memory bandwidth and massive increase in cores a 2x GPU perf increase is not to be ruled out. GCN scales almost perfectly with more cores (R9 290X is 2x the perf of R9 270X).The best part is due to full HSA support the SOC will gain power efficiency due to a unified address space and bandwidth usage will be more efficient. Also the Skybridge version running low power Cortex A57 cores on Android will be more efficient running native Android apps/games while the Intel x86 chips struggle with binary translation overheads.

    But the biggest problem is you are still thinking that its a pure silicon race. Not really. Its the overall solution and there the software (drivers, Mantle API) make a massive difference. If anything the improvements in architectural efficiency with HSA and software drivers/API advantages should widen the gap even further next gen.

Add your comment.

Sponsored Links

Leaked: Apple's Next Smart Device
(Warning, it may shock you)
The secret is out... experts are predicting 458 million of these types of devices will be sold per year. 1 hyper-growth company stands to rake in maximum profit - and it's NOT Apple. Show me Apple's new smart gizmo!

DocumentId: 2989260, ~/Articles/ArticleHandler.aspx, 10/30/2014 9:49:09 AM

Report This Comment

Use this area to report a comment that you believe is in violation of the community guidelines. Our team will review the entry and take any appropriate action.

Sending report...


Advertisement