Intel Z3770 geekbench sighted

[wlee15]

Dresdenboy found a benchmark on geekbench of Z3770 which is 2.4 Ghz Turbo 2.5W SDP Bay Trail-T SOC according to a Intel slide.
http://www.anandtech.com/show/7048/...celeron-and-pentium-brands-24ghz-z3770-leaked

A comparison to a A6-1450
http://browser.primatelabs.com/geekbench2/compare/2157740/2068301

A better performing A6-1450
http://browser.primatelabs.com/geekbench2/compare/2119445/2068301

And the ARM A15 version of the Galaxy S4.
http://browser.primatelabs.com/geekbench2/compare/2166084/2068301

 

[monstercameron]

very competitive with both jaguar and a15 quads...
power use should go a15-->atom-->jaguar
cpu performance should go jaguar-->a15-->atom
gpu performance should go jaguar-->[toss up for exynos and atom]

 

[SPBHM]

interesting thanks,
since it's also going to replace ivy bridge celeron/pentium

vs the ULV celeron 1007u
http://browser.primatelabs.com/geekbench2/compare/2161898/2068301

 

[erunion]

I suppose the use of the number 3770 depicts the top of line baytrail sku

 

[monstercameron]

interesting thanks,
since it's also going to replace ivy bridge celeron/pentium

vs the ULV celeron 1007u
http://browser.primatelabs.com/geekbench2/compare/2161898/2068301

I guess it is obvious but it performs lower in singlethreaded performance, but what about comparing quad high burst atom to quad low clocked haswell?

 

[pelov]

What's the TDP? I don't care for marketing branded acronyms with no meaning.

If the chip is 5-10W TDP then that doesn't look all that impressive. If, on the other hand, the chip is <5W TDP, then that's a neat little Atom.

The TDP would also determine what it's going up against. If it's 5-10W for the SoC, then it's pitted against the A15/A57, Snapdragon 800, and Temash SoCs, but if it's under that (and presumably smartphone territory) then we're looking at the A53, 600 (and lower) and a bevy of other ARM SoCs

 

[monstercameron]

What's the TDP? I don't care for marketing branded acronyms with no meaning.

If the chip is 5-10W TDP then that doesn't look all that impressive. If, on the other hand, the chip is <5W TDP, then that's a neat little Atom.

The TDP would also determine what it's going up against. If it's 5-10W for the SoC, then it's pitted against the A15/A57, Snapdragon 800, and Temash SoCs, but if it's under that (and presumably smartphone territory) then we're looking at the A53.

i thought baytrail was tablet and up clovertrail+ for tablets and below?

 

[pelov]

Well, if the TDP is ~2W range then it's a candidate for smartphones.

That really brings to light why something like 'SDP' is so stupid and annoying.

 

[monstercameron]

Well, if the TDP is ~2W range then it's a candidate for smartphones.

That really brings to light why something like 'SDP' is so stupid and annoying.

none of the baytrail parts have a 2W tdp http://liliputing.com/2013/07/leake...eron-and-pentium-chips-use-less-than-10w.html

and spd seems to be int the range of ~20-50% lower than tdp based off this chart http://www.anandtech.com/show/6655/...-rating-to-get-there-yseries-skus-demystified

 

[kimmel]

What's the TDP? I don't care for marketing branded acronyms with no meaning.

What's the TDP for Exynos Dual/Octa? From what I can tell ARM and its vendors do not give public TDP numbers for anything. You honestly expect Intel to quote a TDP number that will get branded as power consumption by ARM marketing when TDP isn't power consumption?

Expect SDP in areas where Intel competes with ARM and expect TDP for laptop/desktops.

 

[erunion]

i thought baytrail was tablet and up clovertrail+ for tablets and below?

baytrail replaces clovertrail. Its a shrink and a new arch. Think sandybridge to haswell.

i should have added: the silvermont phone chips will be dual core and called merrifield( to replace 32nm Medfield)

 

[Exophase]

The biggest problem with Geekbench is how much variation there is in the scores, although if the power management of the device is causing this that's not necessarily the benchmark's fault. In Exynos 5410's case there are a bunch (like, dozens, see here http://browser.primatelabs.com/geekbench2/search?q=i9500) of results that score ~3600 w/o overclocking the device so the one selected here (2769) is not a good representation. Here's a much better one:

http://browser.primatelabs.com/geekbench2/compare/2068301/2165156

 

[sushiwarrior]

Am I misreading something or are the A15 and A6 (especially the A6) absolutely demolishing the Atom on clock for clock performance..... Wow. Guess the real story lies in power consumption.

 

[monstercameron]

Am I misreading something or are the A15 and A6 (especially the A6) absolutely demolishing the Atom on clock for clock performance..... Wow. Guess the real story lies in power consumption.

arm a9 is 2.5DMIP/MHz
arm a15 is 3.5DMIP/MHz
apple swift is around 3-3.5DMIP/MHz
clovertrail atom is around 2.4-2.7DMIP/MHz
jaguar is around 3.6DMIP/MHz

silvermont/baytrail ?

 

[Exophase]

No not absolutely demolishing, for Exynos 5410 the scores have a perf/MHz ratio of:

Blowfish ST              1.1138552296
Blowfish MT              1.0733206164
Text Compress ST        0.9524279616
Text Compress MT        0.8327415181
Text Decompress ST      0.9627377377
Text Decompress MT      0.4665091426
Image Compress ST        1.1180747126
Image Compress MT        1.1309331966
Image Decompress ST      1.6360354223
Image Decompress MT      1.4935654571
Lua ST                  1.3717582418
Lua MT                  1.2252378774
Mandelbrot ST            1.608053539
Mandelbrot MT            1.5648141892
Dot Product scalar ST    2.4652708333
Dot Product scalar MT    2.4511165246
LU Decomp ST            0.9799537988
LU Decomp MT            0.9224439103
Primality Test ST        2.0105562201
Primality Test MT        1.3333826938
Sharpen Image ST        9.9204572447
Sharpen Image MT        8.9950280733
Blur Image ST            6.190316092
Blur Image MT            6.133104436

At the very least the last four are probably totally broken and should be ignored, some of the other FP ones might be suspect (and they're not representative of anything to begin with)

For some reason the Exynos 5410 device has really bad scaling with text decompress MT, I doubt it's a problem with the uarch but who knows.. I wouldn't look too heavily into that either. Other than that you see a result that's not that surprising.. Cortex-A15 wins some due to having deeper OoO (particularly for FP), wider decode, and more back-end parallelism (particularly simultaneous load + store), while Silvermont wins some due to having lower latency L2 cache and a lower branch misprediction penalty. Silvermont should make up in peak perf with higher MHz in phones and tablets, I suspect their peak performance won't be hugely far off. Not expecting the curb-stomp a lot of people are (that Intel is also implying..)

 

[wlee15]

The biggest problem with Geekbench is how much variation there is in the scores, although if the power management of the device is causing this that's not necessarily the benchmark's fault. In Exynos 5410's case there are a bunch (like, dozens, see here http://browser.primatelabs.com/geekbench2/search?q=i9500) of results that score ~3600 w/o overclocking the device so the one selected here (2769) is not a good representation. Here's a much better one:

http://browser.primatelabs.com/geekbench2/compare/2068301/2165156

That 2769 score is actually for a Samsung ATIV Book 9 Lite notebook containing a AMD A6-1450.

 

[Exophase]

Wow, I feel dumb now. I have no idea how I clicked on one and thought I was looking at the other! Thanks for the correction >_> (I guess I jumped to it because in the past I very commonly saw people post Geekbench scores that were poor representations)

 

[Khato]

none of the baytrail parts have a 2W tdp http://liliputing.com/2013/07/leake...eron-and-pentium-chips-use-less-than-10w.html

Correct, none of the Baytrail I, M, or D SKUs that are referenced in that article have a 2W TDP. The Baytrail T Z3770 has a 2W SDP as per the link the OP provided... which is more applicable in the mobile realm than some would like to believe.

Am I misreading something or are the A15 and A6 (especially the A6) absolutely demolishing the Atom on clock for clock performance..... Wow. Guess the real story lies in power consumption.

Depends primarily upon whether the reported 1.46 GHz frequency for the Z3770 is correct or not. If it is then the Atom is actually performing quite a bit closer to A15 per clock than expected on integer. Especially considering the fact that the Z3770 would almost certainly be using less power at that frequency.

 

[mikk]

Way too low scores for the Z3770 imho. This CPU should easily crush a 2 core 1,0 Ghz Temash.

 

[Nothingness]

Correct, none of the Baytrail I, M, or D SKUs that are referenced in that article have a 2W TDP. The Baytrail T Z3770 has a 2W SDP as per the link the OP provided... which is more applicable in the mobile realm than some would like to believe.

It's a tablet chip as far as Intel nomenclature goes. I guess smartphone chips will have a lower SDP.

 

[pelov]

What's the TDP for Exynos Dual/Octa? From what I can tell ARM and its vendors do not give public TDP numbers for anything. You honestly expect Intel to quote a TDP number that will get branded as power consumption by ARM marketing when TDP isn't power consumption?

Expect SDP in areas where Intel competes with ARM and expect TDP for laptop/desktops.

Well, yes.

Temash
http://www.cpu-world.com/CPUs/Jaguar/AMD-A6-Series A6-1450.html

And you can bet that its 'SDP' is lower still.

Regarding Qualcomm:

When asked about the TDP, Qualcomm stated, "While QCOM doesn&#8217;t quote or use TDP values in describing mobile device thermal characteristics; we believe that typical operation in a smartphone form factor needs to stay below 2.5W while typical operation in a tablet form factor needs to say below 5W. Operations in this context would include power consumption from the entire AP along with any other connectivity subsystem including cellular modem, Wi-Fi, etc. The specific power requirement can vary from device to device based on form factor, material properties, and any passive cooling solutions that can impact peak processor power consumption, but the above averages are generally accepted for smartphones and tablets and all Qualcomm processors are designed to perform within these power limits so they are able to achieve peak performance while minimizing any thermal impact. Designing for optimal mobile thermal efficiency is one of the many reasons Snapdragon processors are able to fit comfortably within a wide variety of highly innovative customer form factors and design."

http://www.brightsideofnews.com/news/2013/6/22/qualcomm-snapdragon-800-benchmarking-day.aspx

So they set certain targets for both smartphones and tablets. Intel's 'SDP' is still very much a marketing tool, and at best a very rough metric set for devices with respect to power consumption and consequently heat dissipation.

So the question is still valid: what's the TDP? or at least what's the typical power usage and what are they aiming for? If Intel's 'SDP' claims 2.5W but the cooling will require 5W dissipated, this thing's not going to end up in smartphones.

 

[kimmel]

So the question is still valid: what's the TDP?

Indeed the question is still valid. What are the TDP values following Intel's TDP definition for all of Qualcomm's products? You quoted a very squishy article about "approximate power" then complain about a squishy SDP number?

http://www.anandtech.com/show/6655/...-rating-to-get-there-yseries-skus-demystified

Scenario Design Power (SDP), on the other hand, is specific to Intel's Y-series SKUs. Here Intel takes a portion of a benchmark that stresses both the CPU and GPU (Intel wouldn't specify which one, my guess would be something 3DMark Vantage-like) and measures average power over a thermally significant period of time (like TDP, you're allowed to violate SDP so long as the average is within spec). Intel then compares its SDP rating to other, typical touch based workloads (think web browsing, email, gaming, video playback, multitasking, etc...) and makes sure that average power in those workloads is still below SDP. That's how a processor's SDP rating is born.

Sounds roughly as squishy as the following statement that you quoted:
http://www.brightsideofnews.com/news/2013/6/22/qualcomm-snapdragon-800-benchmarking-day.aspx

we believe that typical operation in a smartphone form factor needs to stay below 2.5W while typical operation in a tablet form factor needs to say below 5W.

Double standards abound.

 

[pelov]

So they set certain targets for both smartphones and tablets.

By 'they' I mean Qualcomm. We don't know what Intel's targets are, and it's safe to presume their tablet goals, from both a power and heat dissipation perspective, are both significantly higher, especially given their Ivy Bridge-Y TDPs are over 10W.

The bolded part accounts for that

So the question is still valid: what's the TDP? or at least what's the typical power usage and what are they aiming for? If Intel's 'SDP' claims 2.5W but the cooling will require 5W dissipated, this thing's not going to end up in smartphones.

It's not a double standard if one asks the same thing of both parties.

 

[USER8000]

Whats with the single threaded FP scores?? At around 50% higher clockspeed it is lower than Kabini in most of the tests!!

 

[AndroAsc]

The Antutu benchmarks comparing Baytrail clocked at 1.4GHz (45,000) to Snapdragon 800 (30,000) i.e. the best SoC from ARM camp this year, shows an impressive lead of Baytrail over ARM, even when using an UNDERCLOCKED Baytrail

Later this year, we'll see the brainless tech blogging sheeps telling the exact opposite story from last year's "Intel is dead, ARM is king" nonsense to "Intel makes a comeback, challenge the ARM's throne". Go Intel!!!