Perhaps. But Toshiba is quite good at making laptops, and this isn't hugely different. What other reason could there be for a tablet running hot than excessive power consumption?
NVIDIA Tegra Architecture
- Thread starter french toast
- Start date
-
- Tags
- nvidia
Perhaps. But Toshiba is quite good at making laptops, and this isn't hugely different. What other reason could there be for a tablet running hot than excessive power consumption?
One candidate could be Adreno320 since it is OGL_ES3.0 compliant. On another note I'm impressed with the Island demo; NV for sure knows (for years now) how to create pretty techdemos. Others should really take an example from that since techdemos can be technically interesting and pleasing to the eye at the same time.
D
Deleted member 13524
Guest
So are we making a motion to force Ailuros to eat a hat or what?
Maybe it makes more sense to have the same TMUs/ALU ratio as GK107, after looking at the typical workload in Android games?
Or maybe the GK208 was developed really fast, after the "half-GK107" design had already been submitted to Logan..
I thought there were no Rogues with DX11 compliance.
That's too bad, the new Toshiba tablets looked really good on paper.
I think the illumination LED panel for the ipad3/4 is responsible for a good chunk of the heating that happens in those tablets.
Maybe it makes more sense to have the same TMUs/ALU ratio as GK107, after looking at the typical workload in Android games?
Or maybe the GK208 was developed really fast, after the "half-GK107" design had already been submitted to Logan..
Anandtech only compares with ipad 4..
I thought there were no Rogues with DX11 compliance.
That's too bad, the new Toshiba tablets looked really good on paper.
I think the illumination LED panel for the ipad3/4 is responsible for a good chunk of the heating that happens in those tablets.
As I said before I've never had a problem taking things back or apologizing in public if I had been wrong. Aren't you a bit in a hurry to wear your party hat?
Not yet announced variants; all so far announced afaik are DX10.0 while future variants will be DX11.1. Neither Android nor iOS really needs more than DX10.0 for the time being. For anything windows obviously in order to be competitive DX11 is more of a necessity.
I don't know. I just wanted to point out that a few comments on a web site is too little to draw a conclusion
D
Deleted member 13524
Guest
But eating a virtual hat is so much more fun than apologizing in public!
Please don't apologize for having a wrong prediction, this is just a forum
BTW, I am always in a hurry to wear my party hat. Especially when I get to be wrong so many times in this forum..
But will there be DX11.1 variants for Rogue? Is the compliance with windows APIs something that IMG wants, now that Intel will use their own graphics IP for future mobile SoCs?
But eating a virtual hat is so much more fun than apologizing in public!
Please don't apologize for having a wrong prediction, this is just a forum
BTW, I am always in a hurry to wear my party hat. Especially when I get to be wrong so many times in this forum..
It's too early to wish anyone bon apetit yet; that's all I meant. You know we old farts have the benefit of higher patience albeit it's an oxymoron.
No idea what they'd want it for now that ST_E amongst other is out of the picture. Intel will use their GenX GPUs for tablet SoCs only afaik; of course Intel wouldn't theoretically need anything >DX10 either for its smartphone SoCs
Yes we need more Ruby vs. Dawn again
.Adreno3xx won't really be a competitor as Adreno4xx is going to be in chips way before Logan appears. I have no idea though what features it's going to have.
And I'm wondering if that's really 28nm? Anandtech figures it can't be 20nm if they got silicon back 3 weeks ago, which may be true.
Last edited by a moderator:
Yes we need more Ruby vs. Dawn again
Adreno3xx won't really be a competitor as Adreno4xx is going to be in chips way before Logan appears. I have no idea though what features it's going to have.
Adreno420 is according to QCOM's roadmap DX11.1 if memory serves well.
It could be many things above 20nm. One funky idea would be the GPU not being 28LP but something else like HPM f.e. By the way if it needs almost 1GHz to reach its peak GFLOP rate, my estimate would be that it shouldn't need more than 200-250MHz to reach those 18 fps in GLB2.7.
If I take the current best case scenario of a GK110 I have for the QuadroK6000 roughly 23 GFLOPs/W; for a worst case scenario for Logan my speculative math gives me for the above around 75 GFLOPs/W if not more in the end.
A1xLLcqAgt0qc2RyMz0y
Veteran
Is Logan sample on 28nm or 20nm?
Anand seems to believe 28nm (not verified with anyone, just a belief).
Isn't 11 weeks more than enough time for that to happen?
Anand seems to believe 28nm (not verified with anyone, just a belief).
Whereas a poster in the comments points out that TSMC has had 20nm risk production available since April 14, 2013.
If 20nm risk production was available since April 14, 2013 and the samples of Logan came back 3 weeks ago (around July 1st) that would leave 11 weeks between 20nm risk production availability and the Logan sample being physically available.
Isn't 11 weeks more than enough time for that to happen?
"Risk production" generally starts a long time before you start seeing complex logic device doing "risk production"; they often start with more simple, easily repairable, structures like memory or FPGA's.
A1xLLcqAgt0qc2RyMz0y
Veteran
Define "long time" please.
There's other news regarding 20nm production, for instance xilinx claiming their fpga chip taped out 2 weeks ago was the first on 20nm: http://press.xilinx.com/2013-07-09-...traScale-ASIC-class-Programmable-Architecture - so that indeed seems to indicate 20nm isn't ready for samples (unless nvidia found some other magic foundry?).
A1xLLcqAgt0qc2RyMz0y
Veteran
So the consensus here is that the Tegra 5 (Logan) sample is on 28nm.
Since Nvidia seems to announce new Tegra's at the CES show in January and have availability 3-5 months later.
Will that time frame allow for 20nm production or will the Tegra 5 be on the trailing 28nm process?
Since Nvidia seems to announce new Tegra's at the CES show in January and have availability 3-5 months later.
Will that time frame allow for 20nm production or will the Tegra 5 be on the trailing 28nm process?
A1xLLcqAgt0qc2RyMz0y
Veteran
Tegra hasn't turned a profit yet, so basically, yes. Of course, NVIDIA is betting that Tegra will eventually grow enough to be profitable.
But if it doesn't happen within a couple of generations, I guess they may reconsider. Bleeding money is only fun for so long.
TSMC started risk production of High Performance 28nm at the end of Q2 2010, the first product was Tahiti in very low volume at the tail end of Q4 2011.
Blazkowicz
Legend
You need a mature process for low power device?, if trying to push the boundary at least.
And expectations are distorted by Intel's lead.
As a point of comparison, the AMD chips which do about the same thing as Logan but a category above in TDP : APU SoC with a desktop GPU. Bobcat came out in the beginning of 2011, on 40nm. Kabini on 28nm just launched, 2013. Temash, dunno where is it but it exists (lowest power version basically competes with Tegra 4, Logan and Atom old and new).
Tegra 4 itself, announced on January 6 2013.
And expectations are distorted by Intel's lead.
As a point of comparison, the AMD chips which do about the same thing as Logan but a category above in TDP : APU SoC with a desktop GPU. Bobcat came out in the beginning of 2011, on 40nm. Kabini on 28nm just launched, 2013. Temash, dunno where is it but it exists (lowest power version basically competes with Tegra 4, Logan and Atom old and new).
Tegra 4 itself, announced on January 6 2013.
Similar threads
