Unless there's more to Huang's quote to provide context it could also mean that it will cover most, but not all, frequency/protocols (eg GSM, CDMA, and LTE but not Wimax); or that nVidia expects most tablets to remain wifi only, in which case the basebandless wayne variant would be the appropriate SoC for them, with Grey being used almost exclusively in phones.
I haven't read any other articles about the upcoming Tegra roadmap so this is new to me. I like how their codenames are all after superheroes! Jean GREY (Phoenix), Kal-el (Superman), Wayne (Batman), Logan (Wolverine), Stark (Iron Man).
That is one seriously optimistic roadmap, optimistic and wrong. Nvidia is now stating Kal-El as having 2x the power of Tegra 2. Kal-El is just 4 A9 cores with a gpu upgrade. Arm anything will never (or at least for a decade or two) have a processing power multiplier of 30x over C2D and that's what this "roadmap" is trying to say. They put the "Stark" soc at 90x more powerful than Tegra 2. I have one word for that...RIDICULOUS.
I have a prediction that I'm pretty damn sure will come true. Passively cooled Arm processors will hit the "thermal wall" before 2014 and max out at 2ghz or less. I wouldn't be a bit surprised if the A15 generation of soc's in 2012 and 2013 is the last architecture to have truly large jumps in processing power over previous gen soc's. After A15 you are going to have the same problem that Intel and AMD have been facing for many years now. The thermal wall will make developing more powerful soc's very expensive and dependent on additional cores to keep up with Moore's law and that will make programing for performance for these future soc's much more difficult and costly.
A single core of a Core i7-2600k (Sandy Bridge) -vs- a single core of a Core2Duo E6850 (Conroe) with both cores running at 3ghz only shows a 10%-30% power advantage for the much newer Core i7.
To fairly compare different generations of processors you need to at least compare equal clockspeeds and the 6850 is a 3.0Ghz while the i7-2600k is a 3.4 that turbo's up to 3.8Ghz so that immediately skews your (facts) 10-30% power savings at a 13-23% clock increase, all while powering 2 more cores is very impressive, not to mention the clock-for-clock improvements in work accomplished allowing the new generation to return to idle state sooner, and the gate improvements reducing idle leakage.
And where has Nvidia changed their story on Kal-El's performance? link please, I thought the initial reports of 5x GPU performance was made after the Alpha Silicon was already spun and working processors were already demoing?
That benchmark DOES take clock speed into account and they are both ran at 3.0GHz.
However, it does NOT take power consumption into account. Total watt/hours for a sandy bridge to do the same task will ALSO be ~30% lower. So its 30% faster when using 30% less power.
When these other guys start making 2-5w chips, and if Intel pulls off a good Atom re-roll in 2013 I think it will be a heck of a battle and will def be fun to watch :)
Currently a Core 2 Duo is magnitudes more powerful than ARM SoC's, even assuming their 5x speed increase is true I don't think it would quite reach that level.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
13 Comments
Back to Article
DanNeely - Friday, September 9, 2011 - link
Unless there's more to Huang's quote to provide context it could also mean that it will cover most, but not all, frequency/protocols (eg GSM, CDMA, and LTE but not Wimax); or that nVidia expects most tablets to remain wifi only, in which case the basebandless wayne variant would be the appropriate SoC for them, with Grey being used almost exclusively in phones.tviceman - Friday, September 9, 2011 - link
This is my best guess...jjj - Friday, September 9, 2011 - link
Grey was first mentioned 3 days ago and again yesterday.Saltticus - Friday, September 9, 2011 - link
I haven't read any other articles about the upcoming Tegra roadmap so this is new to me. I like how their codenames are all after superheroes! Jean GREY (Phoenix), Kal-el (Superman), Wayne (Batman), Logan (Wolverine), Stark (Iron Man).Red Storm - Friday, September 9, 2011 - link
Someone educate me, which hero is Tegra named after? XDJarredWalton - Friday, September 9, 2011 - link
Tegra is the shipping name, not the code name.jmcb - Friday, September 9, 2011 - link
Probably somebody from Thundercats..or He Man...lolWell was trying to be funny but look what a quick Google search found:
http://pdsh.wikia.com/wiki/Tegra_Jungle_Empress
Wow...lol
LauRoman - Friday, September 9, 2011 - link
I think Grey is named after Jean Grey, if you know your X-men or in this case woman.salimbest83 - Friday, September 9, 2011 - link
thats make sensesKamen75 - Friday, September 9, 2011 - link
That is one seriously optimistic roadmap, optimistic and wrong. Nvidia is now stating Kal-El as having 2x the power of Tegra 2. Kal-El is just 4 A9 cores with a gpu upgrade. Arm anything will never (or at least for a decade or two) have a processing power multiplier of 30x over C2D and that's what this "roadmap" is trying to say. They put the "Stark" soc at 90x more powerful than Tegra 2. I have one word for that...RIDICULOUS.I have a prediction that I'm pretty damn sure will come true. Passively cooled Arm processors will hit the "thermal wall" before 2014 and max out at 2ghz or less. I wouldn't be a bit surprised if the A15 generation of soc's in 2012 and 2013 is the last architecture to have truly large jumps in processing power over previous gen soc's. After A15 you are going to have the same problem that Intel and AMD have been facing for many years now. The thermal wall will make developing more powerful soc's very expensive and dependent on additional cores to keep up with Moore's law and that will make programing for performance for these future soc's much more difficult and costly.
A single core of a Core i7-2600k (Sandy Bridge) -vs- a single core of a Core2Duo E6850 (Conroe) with both cores running at 3ghz only shows a 10%-30% power advantage for the much newer Core i7.
http://www.tomshardware.com/charts/x86-core-perfor...
Black1969ta - Saturday, September 10, 2011 - link
To fairly compare different generations of processors you need to at least compare equal clockspeeds and the 6850 is a 3.0Ghz while the i7-2600k is a 3.4 that turbo's up to 3.8Ghz so that immediately skews your (facts) 10-30% power savings at a 13-23% clock increase, all while powering 2 more cores is very impressive, not to mention the clock-for-clock improvements in work accomplished allowing the new generation to return to idle state sooner, and the gate improvements reducing idle leakage.And where has Nvidia changed their story on Kal-El's performance?
link please, I thought the initial reports of 5x GPU performance was made after the Alpha Silicon was already spun and working processors were already demoing?
extide - Sunday, September 11, 2011 - link
That benchmark DOES take clock speed into account and they are both ran at 3.0GHz.However, it does NOT take power consumption into account. Total watt/hours for a sandy bridge to do the same task will ALSO be ~30% lower. So its 30% faster when using 30% less power.
When these other guys start making 2-5w chips, and if Intel pulls off a good Atom re-roll in 2013 I think it will be a heck of a battle and will def be fun to watch :)
tipoo - Monday, September 12, 2011 - link
*Puts on sceptical face*Currently a Core 2 Duo is magnitudes more powerful than ARM SoC's, even assuming their 5x speed increase is true I don't think it would quite reach that level.