That's a massive design win for powerVR, Samsung semi has been exclusively using Mali for 3 cycles of exynos. Question is why switch.
In terms of gflops its slower that ipad4, but perhaps Samsung feels the filtrate of implementations of hi screen res is lacking with Mali. The setup suggested by anand would provide x1.5 the filtrate of the latest iPad. The latest iPad already has 50% more filtrate that the T604 in the nexus10, giving an overall doubling of fillrate for the suggested 543mp3@533 over the Current T604 implemenation. This would align with the x2 improvement that Samsung suggested in their CES presentation.
Is there also a power reason, anand's conclusions about power draw of the 5250 might indicate so.
In terms of gflops its slower that ipad4, but perhaps Samsung feels the filtrate of implementations of hi screen res is lacking with Mali. The setup suggested by anand would provide x1.5 the filtrate of the latest iPad. The latest iPad already has 50% more filtrate that the T604 in the nexus10, giving an overall doubling of fillrate for the suggested 543mp3@533 over the Current T604 implemenation. This would align with the x2 improvement that Samsung suggested in their CES presentation.
Is there also a power reason, anand's conclusions about power draw of the 5250 might indicate so.
Hence my reasoning above. In any case if it should be a 544MP3@533MHz it might roughly be on iPad4 GPU level.
Would Rogue have been available for a april/may launch?
clearly it could have been, in that LG showed a TV @ CES with GC6200.
But there is no evidence in the public realm to suggest that samsung has a rogue license.
ltcommander.data
Regular
So is the ARM Mali-T604 still getting OES3.0 support? Because, if it does, it looks kind of strange that their latest SoC goes backwards in graphics standard support even if performance on what it does support is better. I guess this might be why IMGTech emphasized the Series 5XT supporting many OES3.0 features through OES2.0 extensions.
I'm guessing the 2x faster triangle setup is actually due to the frequency difference rather than something inherent in the SGX544 vs SGX543? I thought they were supposed to have similar performance just different feature support, but the way Anand wrote that sentence makes it seem like a hardware performance difference.
Also, by quoting a 71.6 GFLOPS rating at shipping frequency for the A6X, Anand's indirectly confirming a 280MHz GPU frequency for the SGX554MP4 in the iPad 4 although no articles have actually outright said it.
The A6 is a 543MP3 and the A6X is a 544MP4, right? A 544MP3@533MHz is probably faster than what's in the A6X then.
LG has integrated a dual cluster Rogue already in a smart TV which they've demonstrated at CES; so the answer is yes.
SGX554MP4@280MHz:
4 * 8Vec4 * 2 FLOPs * 0.28GHz = 71.7 GFLOPs/s
4 * 2 TMUs * 280MHz = 2.24 GTexels/s
4 * 16 z * 280MHz = 17.92 GPixels/s
186 M Tris/s
SGX544MP3@533MHz:
3 * 4Vec4 * 2 FLOPs * 0.53GHz = 51.2 GFLOPs/s
3 * 2 TMUs * 533MHz = 3.19 GTexels/s
3 * 16 z * 533MHz = 25.58 GPixels/s
266 M Tris/s
In GLBenchmark2.5 at least I'd expect that the two above solutions could roughly be give or take in the same performance ballpark. Remember the A6 contains a SGX543MP3@325MHz and yields 3290 frames in 2.5 vs. 5868 frames of the A6X/iPad4. If you scale in theory the A6/iPhone5 score up to 533MHz you get 5396 frames.
You're right, my bad :S. I should pay more attention.....
Interesting. Altough im only speculating, there is nothing concrete showing this chip will be used in Galaxy S4, Samsung has more Exynos 5 chips in the pipeline
In some ways its an odd decision. Sticking to 1 GPU vendor should probably be easier on Android OS development for their phones. But maybe Mali-600 series simply are too big and consume too much power for a phone
OS development is obviously a Google only affair. Besides that Samsung is continiously using SoCs from third parties all the time; they've used since the Exynos 4 introduction: Tegra2, Qualcomm S3 and S4 SoCs and a number of Texas Instruments OMAP SoCs in their mobile products.
Sorry what i meant was releasing OS updates. While they do use various vendors, in my opinion the Exynos chips often get the most support from Sammy and the fastest updates.
It's not that it's the first time that they've had two different GPUs at the same time in their SoCs; when they started out with ARM Mali400MP they still had products with SGX540. I'm even surprised Samsung hasn't made the whole development cycle even more complicated and develop custom CPU cores based on ARM instruction sets. If NVIDIA should use project denver CPU cores after 20nm kicks in also for the small form factor space it'll be three SoC manufacturers with custom cores and Samsung isn't the type of giant that cannot afford any such stunts; rather the contrary.
Applied Micro and its X-Gene 64-bit core.
Perhaps T624 quoted 50% increase in performance over T604 was not seen as being enough, and T658 is not yet ready for production. Driver development takes as much time and resources as hardware development (if not more), and SGX drivers, particularly for opengles2.0 are very mature. Remember that multiple vendors have taken 543/544 (apple,ti,renesas,ST,mediatek etc etc), its drivers have been worked on for a long time.
Similar threads
- Deleted member 13524
- Mobile Devices and SoCs