Tegra 4 benchmarked, clobbers Snapdragon 600.

[Bateluer]

http://www.androidpolice.com/2013/0...rence-phone-at-mwc-with-benchmarks-and-stats/

Definitely some impressive scores. Seems like T4 design wins are few though. Discuss.

 

[GTRagnarok]

Okay, but how does it fare in a phone?

 

[dguy6789]

It's not news that A15 is faster than Snapdragon. It achieves that performance via a greater than 2x power consumption difference vs Snapdragon. Snapdragon completely decimates it in performance per watt.

 

[s44]

Snapdragon completely decimates it in performance per watt.

Not only is "completely decimate" an incoherent word combination (look up "decimate" some time), but there's no evidence of this, either at peak or -- more importantly -- over the mixed-load daily life of a phone/tablet. Most of T4's day is going to be spent on the low-power companion core.

 

[Bateluer]

Okay, but how does it fare in a phone?

Since Phoenix is a reference design for a 5in phone, about like the link.

 

[GTRagnarok]

Since Phoenix is a reference design for a 5in phone, about like the link.

According to the link, that Phoenix is using the Tegra 4i.

 

[ams23]

Snapdragon completely decimates it in performance per watt.

That doesn't appear to be the case with some very common use cases: http://cdn.androidpolice.com/wp-content/uploads/2013/02/nexusae0_tt4_thumb.png . And we would probably need far more data to come to a conclusion one way or another. Keep in mind that quad-core Snapdragon SoC's can be reasonably power hungry too. In fact, Anandtech even thought that the S4 Pro would be better suited to a tablet form factor than a smartphone factor for that very reason.

 

[ams23]

Okay, but how does it fare in a phone?

While raw performance data for a 1080p Tegra 4 reference phone design is not yet available as far as I can tell, see the link in the post above for some power consumption data compiled using a 1080p Tegra 4 (ie. not Tegra 4i) reference phone design.

 

[dagamer34]

Thing is, Snapdragon 800 is Qualcomm's latest and greatest CPU architecture, and Tegra 4 needs to be THAT. And has already been mentioned before, raw CPU performance isn't really the problem, it's always been about managing that performance in a low power situation.

 

[s44]

Thing is, Snapdragon 800 is Qualcomm's latest and greatest CPU architecture, and Tegra 4 needs to be THAT.

Yeah, but it's more like 2011 than 2012. Qualcomm again has the slower CPU design that they're trying to juice with optimizations and higher clocks... Samsung and Nvidia have the more robust base CPU core.

I'm not sure they have to "beat" each other from a consumer perspective, though (obviously design wins will matter a lot) -- we've really entered the "good enough" zone for phones/tablets, and the hybrid space is about beating Intel.

 

[sontin]

In the tablet market we still away from the "good enough" zone. I want more performance. Using a WinRT tablet with a monitor is very nice experience.

 

[gmaster456]

Won't matter much in everyday use until apps become a lot more demanding.

 

[tviceman]

Won't matter much in everyday use until apps become a lot more demanding.

Build a fast processor, and the demanding applications will come. I personally envision the day when everyone's cell phone entirely supplants desktops and consoles. Plug the cell phone into a dock, and a keyboard/mouse/monitor fed from the dock becomes a full powered computing station. Or, in the case of consoles, easy HDMI out with bluetooth controllers playing GTA7. SIGN ME UP.

 

[gmaster456]

Build a fast processor, and the demanding applications will come. I personally envision the day when everyone's cell phone entirely supplants desktops and consoles. Plug the cell phone into a dock, and a keyboard/mouse/monitor fed from the dock becomes a full powered computing station. Or, in the case of consoles, easy HDMI out with bluetooth controllers playing GTA7. SIGN ME UP.

We already have CPU's that have yet to be fully utilized. A5X, A6X, Tegra 3 and the Dual Core A15 in the Nexus 10 all have yet to be maxed. I'm all for the advancement of mobile technology. I'm just saying we're not quite seeing the benefits due to software having to catch up.

 

[Bateluer]

We already have CPU's that have yet to be fully utilized. A5X, A6X, Tegra 3 and the Dual Core A15 in the Nexus 10 all have yet to be maxed. I'm all for the advancement of mobile technology. I'm just saying we're not quite seeing the benefits due to software having to catch up.

Those are the high end. Need to look at the low end, thats where the base line is. Developers develop for the lowest common denominator.

 

[podspi]

We already have CPU's that have yet to be fully utilized. A5X, A6X, Tegra 3 and the Dual Core A15 in the Nexus 10 all have yet to be maxed. I'm all for the advancement of mobile technology. I'm just saying we're not quite seeing the benefits due to software having to catch up.

I'm not so sure of that. There is still a noticeable difference just browsing the web on these devices.

 

[djgandy]

Note how all the tests that Nvidia quotes power consumption are basically CPU and GPU idle tasks.

Video decode and encode is done by dedicated hardware.
Web / book reading. How long is this measured over? Do they actually render the page?
Audio playback, more dedicated hardware, display off etc.
Standby, 10mW? haha yeah let's see someone who isn't Nvidia get that eh 😉

There are 4 A15's on this chip, you'd think Nvidia would be interested in telling us how long the battery lasts when they are stretching their legs.

 

[PingviN]

So, comparing the Tegra 4 - flagship chip to be released in late summer/early autumn with the mid-end Snapdragon 600 available this quarter? Really? If it didn't beat the S600, it'd be a freaking joke. It's not going to compete with the S600 though, it's going up against the S800. Herpdiderp, Nvidia marketing team at it's finest.

 

[MrX8503]

"At the same performance level, Tegra 4 operates at 40% lower power than Tegra 3. The comparison is unfortunately not quite apples to apples as we’re artificially limiting Tegra 4’s peak clock speed, while running Tegra 3 at its highest, most power hungry state. The clocks in question are 1.6GHz for Tegra 3 and 825MHz for Tegra 4. Running at lower clocks allows you to run at a lower voltage, which results in much lower power consumption. In other words, NVIDIA’s comparison is useful but skewed in favor of Tegra 4.

What this data does tell us however is exactly how NVIDIA plans on getting Tegra 4 into a phone: by aggressively limiting frequency. If a Cortex A15 at 825MHz delivers identical performance at a lower power compared to a 40nm Cortex A9 at 1.6GHz, it’s likely possible to deliver a marginal performance boost without breaking the power bank." -Anand

Looks like to fit it in a phone, it'll be down clocked. I'll wait for Anand's full review.

 

[tviceman]

"At the same performance level, Tegra 4 operates at 40% lower power than Tegra 3. The comparison is unfortunately not quite apples to apples as we’re artificially limiting Tegra 4’s peak clock speed, while running Tegra 3 at its highest, most power hungry state. The clocks in question are 1.6GHz for Tegra 3 and 825MHz for Tegra 4. Running at lower clocks allows you to run at a lower voltage, which results in much lower power consumption. In other words, NVIDIA’s comparison is useful but skewed in favor of Tegra 4.

What this data does tell us however is exactly how NVIDIA plans on getting Tegra 4 into a phone: by aggressively limiting frequency. If a Cortex A15 at 825MHz delivers identical performance at a lower power compared to a 40nm Cortex A9 at 1.6GHz, it’s likely possible to deliver a marginal performance boost without breaking the power bank." -Anand

Looks like to fit it in a phone, it'll be down clocked. I'll wait for Anand's full review.

I think Nvidia is/was fully aware that Tegra 4 is a stop gap for phones until they can get 4i out. I'd much rather have something less powerful that gets closer to 2x the battery life than the soul-sucking chips that are out now. Tegra 4's obvious strength will be in tablets.

 

[tviceman]

Herpdiderp, Nvidia marketing team at it's finest.

Qualcomm has said the s800 will be 25% faster than the s600. T4's benchmarks are showing it to be well more than 25% faster than the S600. There are currently no benchmarks of the S800, soooooooooooooo beyond the information qualcomm has given I don't know how Nvidia could make any direct comparisons to the S800.

Herpdiderp????

 

[ChronoReverse]

If we actually see this in phones with reasonable battery life then the claims that this "clobbers" Snapdragon can be brought out.

Also, video playback, mp3 playback and eReading, seriously? How about trying something that actually uses the CPU?


Plus we know power usage doesn't scale linearly with frequency. Why didn't they show the T4 running at its native frequency (or even at whatever frequency it has equal performance to the S4)? Because then the Watt by Performance doesn't look as good.

 

[Eug]

Actually, personally I'm most interested in APU battery life for lower demand usage, like web surfing, MP3 playback, making phone calls, and even video playback, and stuff like that. That's what is most important for a phone.

 

[ChronoReverse]

Sure but that sort of thing is dependent on the dedicated decoders, the radio and the screen. It's VERY important for a cell phone but less relevant when discussing the CPU performance (the thread title is Tegra 4 Benchmarked after all).

In any case, it'll be interesting to see how it fares when it's put in a real-life cell phone.

 

[PingviN]

Qualcomm has said the s800 will be 25% faster than the s600. T4's benchmarks are showing it to be well more than 25% faster than the S600. There are currently no benchmarks of the S800, soooooooooooooo beyond the information qualcomm has given I don't know how Nvidia could make any direct comparisons to the S800.

A tablet with Tegra 4 is "well more than 25% faster" than a phone with S600. Why do you think Nvidia kept the phone as eyes only while demonstrating the silicon in a tablet format? The power envelope is my guess.

Going from the numbers provided by Anand, Tegra 4 is at 60% the power of Tegra 3 at the same level of performance. Tegra 3 didn't make it into many phones, again, probably because of the power consumption. If the correlation between clock speed and power is linear, Tegra 4 could theoretically offer 140% the performance of Tegra 3 without breaking the power envelope (which Tegra 3 must've done seeing how few phones used the SoC).

Maybe my logic is just off, but it does seem pretty ridiculous saying Tegra 4 clobbers Snapdragon 600 when comparing a tablet to a phone.