A couple of weeks ago at MWC, Qualcomm announced its Snapdragon 801 which was positioned as a speed bump for the next wave of flagship smartphones. Qualcomm touted a 2.5GHz CPU frequency (up from 2.3GHz with Snapdragon 800), as well as increased GPU, ISP and memory interface speeds. Samsung announced immediate support for the new Snapdragon 801 with the Galaxy S 5, as did Sony with the Xperia Z2. Unfortunately this is where confusion set in. The Galaxy S 5 was advertised as a Snapdragon 801 with a 2.5GHz CPU clock, while the Xperia Z2 claimed the same Snapdragon 801 branding but with a 2.3GHz CPU clock - the same frequency as a Snapdragon 800. If it's not CPU frequency that separates a Snapdragon 800 from an 801, what does? The answer, as it turns out, is a little more complex. The table below should help explain it all:

Snapdragon 800/801 Breakdown
  SoC Version Model Max CPU Frequency Max GPU Frequency ISP eMMC DSDA Memory IF
MSM8974VV v2 S800 2.2GHz 450MHz 320MHz 4.5 N 800MHz
MSM8974AA v2 S800 2.3GHz 450MHz 320MHz 4.5 N 800MHz
MSM8974AB v2 S800 2.3GHz 550MHz 320MHz 4.5 N 933MHz
MSM8974AA v3 S801 2.3GHz 450MHz 320MHz 5.0 Y 800MHz
MSM8974AB v3 S801 2.3GHz 578MHz 465MHz 5.0 Y 933MHz
MSM8974AC v3 S801 2.5GHz 578MHz 465MHz 5.0 Y 933MHz

The Snapdragon 800 brand applies to an internal Qualcomm model number of MSM8974. The MSM8974 has four Krait 400 CPU cores, an Adreno 330 GPU, dual-ISP, 64-bit wide memory interface and a 9x25 Cat 4 LTE modem. It turns out there are two different silicon revisions of this SoC: version 2 and 3. The Snapdragon 800 v2 silicon was made available in three different flavors: VV, AA and AB. The difference between all three MSM8974 v2 variants was CPU and GPU frequency. This isn't an unusual practice at all as there's bound to be a distribution of operating frequencies for any design. Better bins end up as higher clocked parts, while others get tested and pass at lower frequencies. OEMs can pay more for the faster bins if they want.

In the case of v2 silicon, you could end up with silicon that ran its CPU cores up to 2.2GHz or 2.3GHz. Only the AB variant saw its max GPU frequency climb to 550MHz. The rest of the specs remain identical between all v2 silicon (-AB does get access to faster DRAM). Note that I'm using the marketing frequencies here for CPUs and not the actual frequencies. For whatever reason the OEMs choose to round up to the nearest 100MHz when quoting CPU speeds (2.2GHz is actually 2.15GHz, 2.3GHz is actually 2.26GHz and 2.5GHz is actually 2.45GHz). That's a battle for another day.

Over time it's possible to squeeze more out of a given process and that's exactly what Qualcomm did with MSM8974 v3. This newer silicon revision used improvements on the process side (process push of 28nm HPm) to push frequencies even higher. The options are now 2.3GHz and 2.5GHz on the CPU side. Just like with v2, v3 silicon offers three different variants. Unlike the situation with v2, v3 sees increases in CPU, GPU and ISP operating frequencies depending on which bin an OEM orders. Note that the increases in ISP frequency are substantial. If my math is correct, the Snapdragon 801 should be able to push almost as many pixels through its ISP as the forthcoming 805.

The other major difference is that v3 silicon enables support for eMMC 5.0.

In short, there are two different versions of MSM8974 silicon. Version 3 adds eMMC 5.0 support and hardware dual-sim, dual-active (DS-DA). All variants of v3 silicon can carry the Snapdragon 801 branding, while all v2 variants are Snapdragon 800s.

Putting it in Perspective

What does all of this mean? Let's first talk about the non-frequency related benefits of MSM8974 v3. eMMC 5.0 adds some new features as well as increases the maximum interface speed from 200MB/s in eMMC 4.5 to 400MB/s. As high end smartphones and tablets start using faster internal storage, having eMMC 5.0 support will be necessary to enable faster transfer speeds. SanDisk's recently announced iNAND Extreme update promises 300MB/s sequential read performance for 32GB+ devices. On a smartphone equipped with USB 3 that means you could feasibly copy large movies or files off of your smartphone at up to 300MB/s. Without eMMC 5.0 support you'd be limited to somewhere south of 200MB/s.

DSDA support matters to specific regions, and for those areas the benefit is obvious.

Supporting faster LPDDR3 means more available memory bandwidth for all of the big consumers on the SoC. I'd expect improvements in high end 3D gaming performance, and potentially certain camera/imaging workloads. Remember that all parts of the SoC have to share that tiny interface to main memory, so more memory bandwidth definitely doesn't hurt.

On the frequency side, the gains are pretty easy to understand. The higher peak CPU speed will translate into faster web page and application loads. Higher GPU frequency will allow for smoother frame rates in 3D games, and the faster ISP frequency can enable quicker processing of camera sensor output. You can also look at the benefits of these things from the perspective of lowering power consumption. Tasks can now complete in less time, allowing these individual IP blocks to quickly move down to lower power states and increase battery life.

POST A COMMENT

30 Comments

View All Comments

  • kenansadhu - Wednesday, March 12, 2014 - link

    You are forgiven. Go and live in peace Reply
  • tipoo - Monday, March 10, 2014 - link

    I wonder if there are also improvements to average power dissipation and thermal hotspotting, so that higher clocks can actually be sustained for longer periods of time. Reply
  • Gigaplex - Monday, March 10, 2014 - link

    "On a smartphone equipped with USB 3 that means you could feasibly copy large movies or files off of your smartphone at up to 300MB/s. Without eMMC 5.0 support you'd be limited to somewhere south of 200MB/s."

    Meanwhile my Galaxy S4 manages around 2MB/s...
    Reply
  • TenshiNo - Tuesday, March 11, 2014 - link

    That's *probably* "burst" transfer rate, not sustained. Even still, you have to remember that transfer rate also very dependent on what the device is plugged into. The 5400RPM in your average laptop is not going to be able to compete with the speeds of your mobile device's NAND memory. Reply
  • madmilk - Tuesday, March 18, 2014 - link

    Even cheap 2.5" 5400rpm drives can easily manage 40MB/s, and the densest 2TB 9.5mm drives can do over 100MB/s. Much better than the <10MB/s you see from most smartphones. Reply
  • HardwareDufus - Tuesday, March 11, 2014 - link

    DSDA will help those of us splitting time equally between Sonora Mexico and Arizona USA. Both states use metric on the highways, english/spanish interchangeably and neither follows daylight savings time. But, I absolutely have to have to cell pone providers for my phone to keep it all affordable (AT&T USA, TelCel Mexico). DSDA will be fantastic for so many of us living along borders! Reply
  • HardwareDufus - Tuesday, March 11, 2014 - link

    sorry, run away auto spell checking.... two cell phone providers (two SIMs). Reply
  • Penti - Tuesday, March 11, 2014 - link

    Given or considering that the US has many regional providers I could guess many people who only cross state lines or just work/live in different cities would have use for dual-sim devices too, maybe at least people who tries to use pre-paid. Not that many would buy unsubsidized devices in the US. The thing with DSDA is that you could receive calls from both numbers/sims. Not terribly useful in my country as all providers has a or shares their national network and plans are pretty equal, if your an expat / immigrant living here that has a card/provider to call your home country with (that is cheaper) it might be useful here. In the developing world there are obviously different needs and conditions were it might make sense. These are high-end chips though, so it's not exactly for the rural Indian. Reply
  • hrrmph - Wednesday, March 12, 2014 - link

    Exactly!

    The entire USA media is afraid to let us in on the advantages of Dual-SIM Dual-Active for domestic consumption.

    Having experienced it overseas, I'm a big fan of it. It is touted as a cheapo feature, but in my mind it is a high-end feature for people who need / can afford to have two providers SIM chips (or in some cases, more than 2 providers, especially on pay-as-you-go plans where you only pay for what you use when you need to use it).

    Since all USA telecoms have some coverage gaps and some areas where weak signals can hamper connectivity, I could really see this as a benefit for traveling business people.

    The first step is to always buy an unlocked device, preferably *not* from your telecom. Being able to do that is still sadly a premium market, although the situation is getting better as the cost of unlocked phones comes down.

    The second step is a bit harder: finding flagship premium devices that have the requisite Dual-SIM slots.

    Once you have a Dual-SIM capable device, then you buy SIM chips (typically AT&T and T-Mobile in the US, and various others for other countries if you need them).

    Then you load the appropriate short-term (for your target travel area) SIM-Chips into the device's Dual-SIM slots and activate them.

    You'll enjoy much better coverage by being able to easily get the telecom that has the best (or in some cases, only) signal without having to do much fussing with changing chips frequently (just leave them in the device - the Dual-SIM Dual Active features means you will very likely always have good signal and you are able to easily choose which signal you would like to use).

    For more complex trips, you simply carry with you, or buy-as-you-go, spare SIM chips for your more distant travel sectors. Just fit them into the device later on in the trip when they are needed.

    When your labor / time is money and / or when you just absolutely have to be connected to the best connection if at all possible, then DS-DA is actually a high-end feature...

    It's about time we start recognizing it as such. You pay more, but you get more capability.

    How much more you pay (if any) varies. The SIM chips and pay-as-you go schemes are fairly reasonable. But the high-end flagship devices that are DS-DA capable are hard to find and typically are built for the uber high-end Asian market.

    These high-end models (like the DS-DA variants of the Samsung S5 and Note 3) aren't to be confused with the low-end Asian and Indian models. The low-end models are economy products often used for families and friends who share a device, but who still want separate billings, and thus want multiple SIM chips.

    The high-end models are for people who want maximum mobility and the ability to seamlessly catch the best signals no matter who the provider is. Some people at the high-end also like having two separate phone numbers that they can easily switch on and off at will through the menu on a DS-DA device.

    A fly in the ointment in the US now is getting the telecom systems built-out with reasonably standardized telecom frequencies and bands, and even more importantly, getting DS-DA devices for sale that have sufficient frequencies and bands to travel widely.
    Reply
  • HardwareDufus - Thursday, March 13, 2014 - link

    Some people at the high-end also like having two separate phone numbers that they can easily switch on and off at will through the menu on a DS-DA device.

    That's me right there. I'll be doing the clumsy SIM Swap at the border on Sunday and Monday. Of course if I'm in the US for more than a few hours... I also have to swap out from time to time to check my email. Would love to just be dual active for the time I need it.

    I'm on a plan on the Mexican side.... and a pay as you go on the USA side.... I don't mind a small amount of Data Roaming to continue to recieve my emails on the USA side, but I like making all of my US calls with my US SIM.
    Reply

Log in

Don't have an account? Sign up now