Kishonti GFXbench

[french toast]

I'm not all over the place, I'm consistent with my conclusions.

Valid point about the here and now, ok then ignore the msm flavour and just take the snapdragon plus discrete baseband.
That has been rumoured/demoed and benchmarked in various degrees since MWC, actual products have been announced and I'll launch within a month or so, going off that it's going to take few months or so from product announcements to sale of smartphones containing omap5/tegra 4/A6/nova Thor/exynos 5250 etc....of those exynos 5250 could turn up in the galaxy note, which would match or exceed snapdragon in certain areas, others like power consumption, multithreading and lte wouldn't.

Other than exynos only omap 5 is anywhere near production, and no products have leaked/benched anything, so it's reasonable to assume it's going to be 3 or so months from now.
Even then it runs into the same problems as exynos, better single thread performance, but likely worse power consumption, a current generation gpu that would likely get out performed, and likely no access to the same discrete baseband due to reasons already outlined.

So my whole statement right from the beginning about snapdragon being the all round best smartphone platform for at least 4 months seems a very logical conclusion.

Adreno 320-Halti plus d3d 11 (efficiency only features?) And if it outperforms power vr 544 mp 2 @500mhz, for the same power consumption category, then yes it makes it more advanced Rhys and the one I'd want in my smartphone.

 

[tangey]

I see the first smartphone with 4470 hit glbenchmark 2.5. Looks like they are clocking it the same as the archos tablet, given the very similar results.
http://www.glbenchmark.com/compare....ed_only=1&D1=Kobo Arc (zeus)&D2=Archos 101G10


Actually might not be a smartphone but another tablet

 

[Rys]

That has been rumoured/demoed and benchmarked in various degrees since MWC, actual products have been announced and I'll launch within a month or so, going off that it's going to take few months or so from product announcements to sale of smartphones containing omap5/tegra 4/A6/nova Thor/exynos 5250 etc....of those exynos 5250 could turn up in the galaxy note, which would match or exceed snapdragon in certain areas, others like power consumption, multithreading and lte wouldn't.

Without proper testing, how do you know APQ8064 + MDM9xxx uses more or less power than Exynos 5250 (or any other SoC really), especially without knowing what LTE baseband is going to be paired with the other SoCs? I'll give you multithreaded performance though, that's a lot more obvious.

Other than exynos only omap 5 is anywhere near production, and no products have leaked/benched anything, so it's reasonable to assume it's going to be 3 or so months from now.

You can't trace an accurate line from leaks and press events to product shipments. Those time deltas are all over the place.

Even then it runs into the same problems as exynos, better single thread performance, but likely worse power consumption, a current generation gpu that would likely get out performed, and likely no access to the same discrete baseband due to reasons already outlined.

Qualcomm will sell those basebands to anybody. I'll come back to the gap in GPU performance you think exists between APQ8064 and OMAP5.

So my whole statement right from the beginning about snapdragon being the all round best smartphone platform for at least 4 months seems a very logical conclusion.

You've still not said when you're counting 4 months from From right now, the 29th August 2012, or 4 months from when you can buy the first APQ8064-powered devices?

Adreno 320-Halti plus d3d 11 (efficiency only features?) And if it outperforms power vr 544 mp 2 @500mhz, for the same power consumption category, then yes it makes it more advanced Rhys and the one I'd want in my smartphone.

Adreno 320 does not support any D3D11 Feature Level that has anything you could call "efficiency only features" that give it a performance/watt advantage. Given how close A5X gets in graphics performance to APQ8064 (Anand's graphics benchmarks), wouldn't you say that OMAP5 is also going to be very close to it in the same benchmarks? I'm trying to figure out where it's going to outperform OMAP5 in any large way.

That's before we get to GPU architectural efficiency.

Small but important point: my name is Rys.

 

[Lazy8s]

The performance TI is getting with one GPU core in the 4470 is impressive. The two cores and higher clock rate of the 5430 should be very competitive.

The more intensive benchmarks for graphics like GLBenchmark 2.5 and Basemark GUI, as limited in representation of real-world performance potential as they may be, are still the best tools for comparing different devices for modern workloads currently, and I don't expect APQ8064 to go unchallenged in those. Qualcomm did hit their window of opportunity well, though, and should be a top competitor for performamce in these upcoming months.

On the topic of CPU performance, though, what real-world mobile application/workload is heavily multi-threaded such that a quad core Krait would beat a dual core A15, as implemented in, for example, APQ8064 versus OMAP5430? Anyway, Krait is here and now, so that's what matters for now.

 

[Ailuros]

By the way aren't the fillrate results of the SGX544 somewhat low? http://www.glbenchmark.com/compare....ified_only=1&D1=Archos 101G10&D2=Apple iPad 2

2 TMUs * 384MHz = 768 MTexels/s
4 TMUs * 250MHz = 1000 MTexels/s

I can't imagine the 544 with its overall results to run at less than 384MHz, so it might have less system level cache than the MP2 or the SoC has less bandwidth or both?

 

[french toast]

I know your name and your profession rys ...you have my respect.

However I will just clarify what I have already said further.
I mean from now.
Qualcomm doesn't sell lte basebands to it's rivals that I'm aware of...tats why we get the absurd situation of different SOC vendors operating in the the same phone in different markets...ie...galaxy s3 international vs US variants.
...you don't think Samsung would rather bung lte baseband on exynos rather then give sales to a rival?...
-same with tegra.

Your correct about me not knowing power consumption numbers of the two said products, however neither do you, and it's logical to assume 4 kraits (smaller) at a lower frequency will be able to spend the load and consume less power than 2 gorillas operating at a higher frequency don't you think??..especially on the same process.

Cross linking what I mentioned above, it's not certain that omap 5 will have access to LTE is it??..therefore my comment about that being a possible advantage for Qualcomm makes sense.

I never said adreno 320 would out class 5xt mp2 @500mhz in open gl es 2.0 performance...but it already beats the lower clocked mp4 ...and as I have said it offers higher api which makes it more advanced an be able to do some things that 5xt can not.

The fact that snapdragon pro has been formally announced in a SMARTPHONE and be on sale in one month or so, and no sniff of an omap 5 device..tablet or other wise .(galaxy note 2 won't carry exynos 5250)..lends credence to my assumption that snapdragon will be the best bet up to 4 months or so..Maybe more?

 

[Ailuros]

Your correct about me not knowing power consumption numbers of the two said products, however neither do you, ......

I apologize for the interference, but let's just say that any insider has times more chances to actually know these details than we "mortals" have.

I never said adreno 320 would out class 5xt mp2 @500mhz in open gl es 2.0 performance...but it already beats the lower clocked mp4 ...and as I have said it offers higher api which makes it more advanced an be able to do some things that 5xt can not.

At the moment the difference between the highest Adreno320 score and the SGX534MP4 is at a "huge" 2% or else 25,4 fps for the first vs. 24,8 fps for the latter. The MP4 in iPad3 is clocked at 250MHz, while the MP2 in OMAP at 532MHz and not 500. Do the math.

 

[Arun]

and as I have said it offers higher api which makes it more advanced an be able to do some things that 5xt can not.

To be clear, SGX544 supports the same DirectX 11 Feature Level 9.3 as Adreno 320. However it doesn't support OpenGL ES 3.0 because of a few missing features. That means it could in theory support a fair bit of ES3 functionality via ES2 extensions. As for performance, obviously a lot depends on driver improvements on both sides...

 

[Laurent06]

Qualcomm doesn't sell lte basebands to it's rivals that I'm aware of...

Isn't the new iPad 4G using Qualcomm LTE?

 

[Rys]

I know your name and your profession rys ...you have my respect.

The profession doesn't matter, just that you don't think I'm Welsh

I mean from now.

You should have picked the other option IMHO, since I don't think it's at clear that the LG will ship as soon as you think it will in September. I hope it does since I want to get my hands on one or ten, but I don't think it's that close .

Qualcomm doesn't sell lte basebands to it's rivals that I'm aware of...tats why we get the absurd situation of different SOC vendors operating in the the same phone in different markets...ie...galaxy s3 international vs US variants.
...you don't think Samsung would rather bung lte baseband on exynos rather then give sales to a rival?...
-same with tegra.

Qualcomm sell MDMs to rivals. Apple are probably a bigger customer for that than anyone else on the planet. As for Samsung, Samsung Mobile is a completely different beast than Samsung LSI who make the chips. It will source from whoever it wants and doesn't exclusively use Samsung LSI at all, SoC or modem.

Your correct about me not knowing power consumption numbers of the two said products, however neither do you, and it's logical to assume 4 kraits (smaller) at a lower frequency will be able to spend the load and consume less power than 2 gorillas operating at a higher frequency don't you think??..especially on the same process.

That depends a lot on the integration but all signs point to Qualcomm doing a good job there.

Cross linking what I mentioned above, it's not certain that omap 5 will have access to LTE is it??..therefore my comment about that being a possible advantage for Qualcomm makes sense.

See above. OMAP5 customers could use Qualcomm modems.

The fact that snapdragon pro has been formally announced in a SMARTPHONE and be on sale in one month or so, and no sniff of an omap 5 device..tablet or other wise .(galaxy note 2 won't carry exynos 5250)..lends credence to my assumption that snapdragon will be the best bet up to 4 months or so..Maybe more?

Let's see how the timescales all pan out.

 

[ltcommander.data]

To be clear, SGX544 supports the same DirectX 11 Feature Level 9.3 as Adreno 320. However it doesn't support OpenGL ES 3.0 because of a few missing features. That means it could in theory support a fair bit of ES3 functionality via ES2 extensions. As for performance, obviously a lot depends on driver improvements on both sides...

Would the SGX545 be OpenGL ES 3.0 capable with the right drivers seeing it supports desktop OpenGL 3.2 and DX10.1?

 

[Rys]

Nope, no SGX is GLES3 compliant.

 

[metafor]

Isn't the new iPad 4G using Qualcomm LTE?

Also the Galaxy Nexus LTE.

Tegra/Exynos still used SDIO for its modem communications; which made it unsuitable to work with the MDM9xxx series.

I would imagine Exynos 5xxx resolves this. OMAP has never had this problem.

 

[tangey]

In the tablet space, the Linley group consider clovetrail (Z2760)and omap 5432 as the highest performing socs to ship this year.( will 5432 ship this year ?, I doubt it)
http://www.linleygroup.com/newsletters/newsletter_detail.php?num=4872

Separately, there is also some suggestions that the Motorola announcement in a couple of weeks time might be for the dual core medfield, the Z2580 with dual core 544 graphics. Now an announcement and a phone are two different things, but such a phone should end up towards the top end of the performance tables, assuming a reasonable release timeframe.

 

[tangey]

I see a samsung GT-I9105 has showed up with 544MP.
http://glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Samsung+GT-I9105&testgroup=gl

Nothing startling on the score, slightly less than the 4470 scores on the archos and kobo.I know there is such a thing as MP1,however it's interesting that the above 2 omap4 scores refer to the renderer as "544" (no MP). So is this a test of omap5, or is this a first citing of enginnerring testing of Samsung's recently acquired 5XT license?

 

[Ailuros]

I see a samsung GT-I9105 has showed up with 544MP.
http://glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Samsung+GT-I9105&testgroup=gl

Nothing startling on the score, slightly less than the 4470 scores on the archos and kobo.I know there is such a thing as MP1,however it's interesting that the above 2 omap4 scores refer to the renderer as "544" (no MP). So is this a test of omap5, or is this a first citing of enginnerring testing of Samsung's recently acquired 5XT license?

IMHO some sort of early results of Samsung's own upcoming single core 544 SoC.

 

[french toast]

RYS: I thought you were Welsh

All I was referring to was that adreno 320 is the top dog on very early drivers, and no lpddr3 that (I think) it can have access to, although some educated people above point out SGX 544 has nearly the same api feature set, the fact is as of press releases- SGX 544 is not halti compliant...unless I read RYS incorrectly :/...making my statement of likely best in class smartphone gpu logical..or at least equal best.

It seems likely omap 5 would have access to Qualcomm lte ip, but again going off public information that is not certain, neither is Snyder kind of release date for smartphones containing that chip.

Bringing apple into the equation doesn't count as they don't use competing SOC's...Qualcomm would gain nothing by cock blocking apple as they would by stopping nvidia/TI..

4 smaller krait cores independently volted and at a lower frequency would seem to have the power efficiency lead over a higher clocked larger A15 on same process...again I don't know that, but it seems a logical conclusion to draw, regardless of performance discussions.

I wasn't aware of the technicalities of modem integration that metaphor mentioned, I was just basing my assumption off the multi SOC phone variants and the fact Samsung was moaning about this very fact sometime around MWC..(Qualcomm not being lte friendly ) also nvidia have gone on record saying tegra can be fitted to lte..but made no mention of reasons.

 

[Nebuchadnezzar]

IMHO some sort of early results of Samsung's own upcoming single core 544 SoC.

Given the processor frequencies, I doubt it's anything Samsung.

 

[Ailuros]

Given the processor frequencies, I doubt it's anything Samsung.

If it's some sort of engineering sample it wouldn't run at final frequencies at least parts of the SoC. Given the GPU results its probably running at somewhere around 350MHz which is quite a bit apart from the target frequency I'd expect for it. The latter has obviously nothing to do with the CPU frequency. We'll see.

 

[Deleted member 13524]

If it's some sort of engineering sample it wouldn't run at final frequencies at least parts of the SoC. Given the GPU results its probably running at somewhere around 350MHz which is quite a bit apart from the target frequency I'd expect for it. The latter has obviously nothing to do with the CPU frequency. We'll see.

Why not just a lower-clocked OMAP4470? It looks like the performance that a ~300MHz SGX544 would have.

Or maybe a Nova A9540? Samsung has a couple of handsets using SoCs from ST-Ericsson (Galaxy Ace 2, Galaxy S Advance).