Given how NV counts so far SoC performance compared to predecessing SoCs, that sounds rather like quite a dissapointment for Wayne/T4.
NVIDIA Tegra Architecture
- Thread starter french toast
- Start date
-
- Tags
- nvidia
Given how NV counts so far SoC performance compared to predecessing SoCs, that sounds rather like quite a dissapointment for Wayne/T4.
anexanhume
Veteran
My guess is that Stark is getting a huge GPU bump and they're quoting ideal peak FLOPS in their metric. What do we know about Wayne GPU?
french toast
Veteran
Well seeing as they used the cpu cores to come to the conclusion tegra 3 was "5 times" tegra 2...then 2 x tegra 3 is going to be a massive disappointment if measured in any metric.
french toast
Veteran
How hell nvidia hopes to compete against exynos 5 series, omap 5 series, snapdragon s4 prime/s5?, and rogue gpu soc s this time next year is going to be puzzling...especially as they will all carry on board lte and be on 28nm....
ltcommander.data
Regular
If the performance isn't there to speak for itself, I suspect we'll be hearing a lot about Tegra Zone and platform exclusives.
The Tegra 4 SoC to be announced December 2012/January 2013 will be at least 2x more powerful than Tegra 3 in terms of CPU and GPU performance. So at that time, Tegra 4 should be very competitive overall. The real question is, at what point in 2013 will Batman make an appearance?
Problem is that Nvidia seems to always calculate performance for the whole soc itself. We already know Cortex A15 will bring substantial improvement over Cortex A9 so how much will the 2x better than Tegra 3 improvement be from the processor and how much better is the GPU compared to T3
2X Tegra 3 GPU should put it roughly equal to A5X, OMAP 5 and Adreno 320.
2X Tegra 3 GPU should put it roughly equal to A5X, OMAP 5 and Adreno 320.
french toast
Veteran
Ha ha. Yea but I'm thinking the new gpu s will be unified shader designed..with open cl, halti and dx 3d 11 9.3.
-Mali t604 is quoted as being 4-5x the performance of Mali 400..
-Adreno 320 with decent drivers, high clockspeed and lpddr3 will be hitting A5X levels in gl benchmark 2.5...
-sgx 544 mp2 high clocked will equal that..
- rogue will obliterate the lot of them.
So would twice the performance of tegra 3 really be enough in that company? More over if they were switching to a new uarch wouldnt the performance be much greater than 2x??
-we would be looking at more like Mali 400 > T604 jump wouldn't we?.
I have some serious doubts we will be seeing a new uarch with Wayne...I think we will be seeing twice the execution resources of tegra 3 with the same architecture...just my opinion.
-Mali t604 is quoted as being 4-5x the performance of Mali 400..
-Adreno 320 with decent drivers, high clockspeed and lpddr3 will be hitting A5X levels in gl benchmark 2.5...
-sgx 544 mp2 high clocked will equal that..
- rogue will obliterate the lot of them.
So would twice the performance of tegra 3 really be enough in that company? More over if they were switching to a new uarch wouldnt the performance be much greater than 2x??
-we would be looking at more like Mali 400 > T604 jump wouldn't we?.
I have some serious doubts we will be seeing a new uarch with Wayne...I think we will be seeing twice the execution resources of tegra 3 with the same architecture...just my opinion.
french toast
Veteran
Well as I've explained..surely a new uarch would be seeing much greater gains than 2x....hell even going from tegra 2 > 3 yielded better than that!...are you seriously telling me a whole new advanced unified shader architecture on 28nm and dual channel lpddr3 memory will be only 2x tegra 3 on 40nm and single channel lpddr2!?....either he's telling some pork pies to keep the hype down or we are seeing more of the same current generation.
Compared to upcoming SGX544MP2, Adreno 320, or anything else with A5X-like performance, yes that would be enough raw performance to be competitive. Compared to upcoming Mali 6xx/Rogue 6 series/etc., it may be a different story. But that is what Batman is for, right? :smile:
Can you blame Nvidia though? What would Rogue level performance give them in terms of revenue? If they spend a lot of R&D money on a big chip they need a partner that can move a lot of units to make up for the money spent. Apple can afford to do this because they know they will sell 20 million units in a quarter easy. Who is backing Nvidia? HTC is invested in Qualcomm, Motorola is tight with TI, Samsung prefers to license from ARM. Nvidias current strategy with time to market on a smaller budget atleast keeps them in the game.
Windows 8 RTs potential success could make a difference for Nvidias status though
Windows 8 RTs potential success could make a difference for Nvidias status though
2nd time you mention Batman. Do you have some inside info or just messing around
Last i checked Logan is the next chip after Wayne
There were rumors posted a long time ago on the internet that Wayne consists of two SoC variants, Batman and Robin. Robin is the variant speculated to be 2x GPU performance improvement vs. Tegra 3.
Don't forget that between now and end of next year, Tegra 5 [Logan] will be 50x faster (!) than Tegra 2 and 10x faster than Tegra 3. The biggest jump in performance occurs sometime between Wayne and Logan, not between Logan and Stark, and not between Kal-El and Wayne! Just some food for thought.
True, the exact performance differences as quoted by NVIDIA are not exactly reliable without knowing how they define performance, but the important thing is that the performance increase is much higher between Wayne and Logan than it is between Logan and Stark and between Kal-El and Wayne.
Similar threads
