Nvidia going for the jugular... Tegra 3 quad core this year?

[shabby]

http://www.engadget.com/2011/01/24/nvidia-tegra-3-equipped-with-1-5ghz-quad-core-madness-teased-b/

Can't believe how quickly nvidia is moving with tegra.

 

[yh125d]

Even if they're at 32nm, I can't imagine how a part like that could be feasible power wise in a smartphone, even if its speed cut down to like 800mHz. *Maybe* they could get away with it in a tablet, but even then I don't know.

 

[foghorn67]

Even if they're at 32nm, I can't imagine how a part like that could be feasible power wise in a smartphone, even if its speed cut down to like 800mHz. *Maybe* they could get away with it in a tablet, but even then I don't know.

Why do people say this before even knowing anything about the chips besides the roadmap? Honestly.

 

[destrekor]

Why do people say this before even knowing anything about the chips besides the roadmap? Honestly.

Exactly.

We don't know the thermal envelope, and we are equally clueless as to the voltages used in the chips.

 

[Bateluer]

Execution is going to be critical. Getting a working product into consumer hands in a timely fashion. I don't give a damn if Nvidia ships Tegra 3 quad core silicon to HTC/Motorola in Q4 2011 if I can't buy a device with it until Q3 2012.

 

[tfinch2]

Given how slow manufacturers have been able to execute with this new hardware, I frankly find myself not being able to get excited about this stuff anymore.

The only tablet currently available with the Tegra 2 SoC is the Viewsonic G Tablet, which probably has the worst display on any device I have seen in some time.

 

[akugami]

http://www.engadget.com/2011/01/24/nvidia-tegra-3-equipped-with-1-5ghz-quad-core-madness-teased-b/


Can't believe how quickly nvidia is moving with tegra.

I can. They are going to be effectively locked out of the low end GPU business in a couple of years. CPU's with on die GPU's from AMD and Intel will be as good as any $50 and under GPU chipset.

Keep in mind that you likely won't be seeing consumer Tegra 3 devices until 2012. We are just now seeing Tegra 2 devices released and that debuted in Fall 2010.

Execution is going to be critical. Getting a working product into consumer hands in a timely fashion. I don't give a damn if Nvidia ships Tegra 3 quad core silicon to HTC/Motorola in Q4 2011 if I can't buy a device with it until Q3 2012.

I don't think Tegra 3 will be quite that late but it likely won't hit consumer hands until Q1 2012 much like Tegra 2. I also think 28nm production from TSMC is going to play a huge role in how fast Tegra 3 goes from being officially unveiled and how fast it gets into consumer hands. TSMC screwing up was what caused a huge delay in nVidia's last flagship product Fermi. Granted the process node wasn't mentioned but I just don't see how nVidia plans to hit decent battery life without a process shrink.

 

[yh125d]

Why do people say this before even knowing anything about the chips besides the roadmap? Honestly.

Are you serious? People already complain about battery life on current 800mhz-1ghz phones, what makes you think Nv might hold the secret to putting a quad core SoC in a smartphone as soon as this fall?

NV is suggesting they can put a quad core SoC in a smartphone by the end of this year. I'm suggesting that even with a die shrink and lower frequency, a quad core ARM SoC in a smartphone with a common 1300-1600mAh battery is a pipedream.

Unless you have some insider info saying that a quad core SoC in 9 months will use as much or less power than current SoCs, which in the opinion of many are barely cutting it for battery life, troll someone else.

 

[cbn]

Hopefully this Tegra has power gating:

http://www.anandtech.com/show/4103/qualcomm-demos-dualcore-snapdragon

Unlike Tegra 2, which can clock gate cores but cannot power gate the second core, scorpion cores can be independently turned off to conserve power as well as be clock gated.

 

[MrX8503]

Are you serious? People already complain about battery life on current 800mhz-1ghz phones, what makes you think Nv might hold the secret to putting a quad core SoC in a smartphone as soon as this fall?

NV is suggesting they can put a quad core SoC in a smartphone by the end of this year. I'm suggesting that even with a die shrink and lower frequency, a quad core ARM SoC in a smartphone with a common 1300-1600mAh battery is a pipedream.

Unless you have some insider info saying that a quad core SoC in 9 months will use as much or less power than current SoCs, which in the opinion of many are barely cutting it for battery life, troll someone else.

Tegra 3 is Q4 2011, expect Tegra 3 Devices in Q1/12 2012.

This is a long time so you can't make claims either way about battery life. There's more to battery life than just the SoC. You have the SoC, OS, and the battery itself.

All three of these components can advance within that time. Looking at the history of smartphones, battery life has been slowly increasing so I don't see a reason why new smartphones of the future will have worse battery life. Overall, I think we'll expect similar battery life.

Look at the PC industry, You think we'd continue on a thermal envelope that the Pentium 4's had?

 

[yh125d]

Tegra 3 is Q4 2011, expect Tegra 3 Devices in Q1/12 2012.

This is a long time so you can't make claims either way about battery life. There's more to battery life than just the SoC. You have the SoC, OS, and the battery itself.

All three of these components can advance within that time. Looking at the history of smartphones, battery life has been slowly increasing so I don't see a reason why new smartphones of the future will have worse battery life. Overall, I think we'll expect similar battery life.

The roadmap says devices this fall, and as I have no other info that's what I'm basing my opinion on.

The reason phones keep getting faster without battery life changing too much is die shrinks, which allow phones to get slowly faster while staying in the same power envelope. Current flagship phones are in the ~1ghz single core SoC area. Nv seemingly plans to quadruple that in 9-12 months, and I don't see how that's even close to likely, unless they plan on releasing phones with 2 hour usable times.

Look at the PC industry, You think we'd continue on a thermal envelope that the Pentium 4's had?

I most certainly do. In fact, we're higher. Prescotts were only like 115w. Typical quad cores today are 95-140w TDPs. But they can afford to do that, because they don't run on battery

 

[zerocool84]

Given how slow manufacturers have been able to execute with this new hardware, I frankly find myself not being able to get excited about this stuff anymore.

The only tablet currently available with the Tegra 2 SoC is the Viewsonic G Tablet, which probably has the worst display on any device I have seen in some time.

Slow about executing new hardware? Are you new to electronics??? Things always get announces months or even years before we get them in our hands.

 

[Nebor]

What jugular? I'm failing to see the violence implied by your title.

 

[MrX8503]

The roadmap says devices this fall, and as I have no other info that's what I'm basing my opinion on.

The reason phones keep getting faster without battery life changing too much is die shrinks, which allow phones to get slowly faster while staying in the same power envelope. Current flagship phones are in the ~1ghz single core SoC area. Nv seemingly plans to quadruple that in 9-12 months, and I don't see how that's even close to likely, unless they plan on releasing phones with 2 hour usable times.

Expect delays. Tegra 2 was Q1 2011, but its more like Q2/Q3 2011.

Most likely it'll be similar battery life and probably not worse. We'll have dual core shortly, doubling that is not uncommon. Its been done on the desktop. You're also forgetting OS optimizations, Google could have a much more efficient running Android by then.

How else do you think Apple increased both battery life and performance in their iPhone4? These things are possible.

I most certainly do. In fact, we're higher. Prescotts were only like 115w. Typical quad cores today are 95-140w TDPs. But they can afford to do that, because they don't run on battery

...And a 4 core Sandy Bridge is 65W-95 TDP and how many several times more powerful is Sandy Bridge compared to a Prescott?

4 Cores does not mean current Single Core TDP times four. This is why I brought up the Pentium 4 example. Just because we get a quad core CPU doesn't mean that Sandy Bridge is going to be a 400W CPU!

Desktops don't run on battery, but power draw still matters because electricity isn't free.

 

[tfinch2]

Slow about executing new hardware? Are you new to electronics??? Things always get announces months or even years before we get them in our hands.

I do not really want to get into an argument about this, but I am not really new to these sorts of things. A large part of my job entails doing everything we can to make sure our partners are successful when we announce a product.

Basically what I am saying is that Nvidia needs to do a better job of these things if they want to be successful in the SoC market. Team up with a partner and put out a real product.

Edit: I understand that Nvidia is not announcing anything here.

 

[LostPassword]

Maybe in tablets late next year. Don't it will be in phones anytime soon

 

[OBLAMA2009]

I can. They are going to be effectively locked out of the low end GPU business in a couple of years. CPU's with on die GPU's from AMD and Intel will be as good as any $50 and under GPU chipset.

Keep in mind that you likely won't be seeing consumer Tegra 3 devices until 2012. We are just now seeing Tegra 2 devices released and that debuted in Fall 2010.



I don't think Tegra 3 will be quite that late but it likely won't hit consumer hands until Q1 2012 much like Tegra 2. I also think 28nm production from TSMC is going to play a huge role in how fast Tegra 3 goes from being officially unveiled and how fast it gets into consumer hands. TSMC screwing up was what caused a huge delay in nVidia's last flagship product Fermi. Granted the process node wasn't mentioned but I just don't see how nVidia plans to hit decent battery life without a process shrink.

yes i read the same article. they debut the tegra 3 next month but it doesnt appear in any products until the NEXT YEAR, MAYBE????? sounds like vaporware to me. everyone else already has dual core and this gives them plenty of time to come out with quad cores as well. jensen is such a clown

 

[akugami]

yes i read the same article. they debut the tegra 3 next month but it doesnt appear in any products until the NEXT YEAR, MAYBE????? sounds like vaporware to me. everyone else already has dual core and this gives them plenty of time to come out with quad cores as well. jensen is such a clown

I wouldn't go so far as to call it vaporware. I have full confidence that Tegra 3 will be used in many shipping products. However, much like the debut of Tegra 2 and it's "hundreds of design wins" it's going to take some time to get shipping products out. Tegra 2 showed a fall 2009 debut but we are just now seeing shipping products. I don't see Tegra 3 products being actually available until the 2012 Q1 time frame.

It's also interesting to note that for all intents and purposes, Tegra 2 has been a "paper launch." A term that, if not coined by nVidia, was certainly popularized by nVidia in describing rival ATI/AMD.

Not everyone has dual core ARM CPU's yet but enough do have it that nVidia is going to see some very tough competition. There's also enough time from now to the release of Tegra 3 that it's quite possible another company will have a quad core ARM CPU out as you say.

A word of caution for those dead set on jumping to a quad core ARM CPU is that the move from single to dual cores should yield much better performance gains than the move from dual to quad cores. I really don't see any mobile phone software benefiting in any great way with a quad core ARM CPU in the next couple of years. The screen size limits what you can do with the available power. The quad core ARM CPU's make more sense in a tablet device IMHO.

 

[yh125d]

Expect delays. Tegra 2 was Q1 2011, but its more like Q2/Q3 2011.

Most likely it'll be similar battery life and probably not worse. We'll have dual core shortly, doubling that is not uncommon. Its been done on the desktop. You're also forgetting OS optimizations, Google could have a much more efficient running Android by then.

How else do you think Apple increased both battery life and performance in their iPhone4? These things are possible.

Apple increased performance by a decent margin while retaining batt life but not a drastic performance boost that would come by going from 1 -> 4 cores. They did so with a SoC die shrink (among other factors). This makes sense, right?

But NV trying to quadruple the core count with only what, one die shrink? And expect to retain usable batt life? That doesn't make sense. And I'm chided for being skeptical.

...And a 4 core Sandy Bridge is 65W-95 TDP and how many several times more powerful is Sandy Bridge compared to a Prescott?

4c SB is 95w right now, which is on the low end of quads if you don't count e models. Most quads in use (Q6600/Q9550/i5 750/i7 920/PHII 965) are more like 95-140w, right around what high-end processors have been for years now.

You know how SB can be 10x faster than prescott and use the same power? because there's been like 5 process shrinks! Duh, that's how it works. What NV is trying to do is make a huge step forward in core count WITHOUT the huge advantages in process. It took Intel 7 years to go from introducing 90nm prescott to 32nm sandy bridge, but NV is wanting to go from 65nm 1gHz single cores (EVO, which was much lambasted for poor battery life) and 45nm 1gHz cores (hummingbird/galaxy S, which seems to be alright on batt life) to 4 core SoCs at presumably 32nm in a year? You'll have to forgive me for being less than optimistic about their chances of reaching those goals.

And it doesn't help that it's NV that's setting these goals. They ate a lot of crow for releasing GTX280 at 65nm when it clearly needed to be on 55nm, and further proved their eyes were too big for their process node with fermi. Is this gonna be a third?

4 Cores does not mean current Single Core TDP times four. This is why I brought up the Pentium 4 example. Just because we get a quad core CPU doesn't mean that Sandy Bridge is going to be a 400W CPU!

It does if you don't go through the necessary die shrinks. There's a reason SB is 32nm, because it has to be. Because SB at 90nm wouldn't work...

 

[MrX8503]

Serious?

I'm not gonna repeat what I said. You really think Intel/AMD/NVDIA dont do die shrinks?

Get back to me when that fable Tegra 3 with 4X the TDP of single core ARM comes to existence because it seems that is what you're suggesting.

NVIDA isn't stupid to release an inefficient quadcore SoC. Twist my words into whatever you want, your claims of poorer battery life is not only unlikely but has no basis. Expect similar battery life.

 

[yh125d]

Serious?

I'm not gonna repeat what I said. You really think Intel/AMD/NVDIA dont do die shrinks?

Get back to me when that fable Tegra 3 with 4X the TDP of single core ARM comes to existence because it seems that is what you're suggesting.

NVIDA isn't stupid to release an inefficient quadcore SoC. Twist my words into whatever you want, your claims of poorer battery life is not only unlikely but has no basis. Expect similar battery life.

Reading comprehension much? I don't see where I suggested they dont to process shrinks...


I'm saying that intel has been able to do what they've done over many process shrinks, and couldn't have without. What I see here, is that NV is trying to make a huge leap ahead in performance with only a small step ahead in process. For a huge leap ahead in performance, you need a correspondingly huge step ahead in process. Hell, even if it's actually 28nm, that's only 1.5 nodes ahead of current SoCs

 

[cbn]

There's more to battery life than just the SoC. You have the SoC, OS, and the battery itself.

Apparently the LCD screen size also has major impact on the power draw.

 

[MrX8503]

Yh125D apparently works for NVIDIA and knows their entire roadmap.

Apparently the LCD screen size also has major impact on the power draw.

yeah the screen is a factor too.

 

[Raghu]

TSMC has no 32nm. So Im guessing this is gonna be 28nm.

Also thats ~2x shrink from 40nm - transistor area is (40/28)^2.

 

[s44]

I can't believe how few people can read a graphic... It indicates that Tegra3 samples are 4Q*10* -- which means they already happened. Late this year is quite possible.

As the original article points out, this jives with Nvidia's September statement that it was already almost done.