Intel Z3770 geekbench sighted

[SiliconWars]

This is all adding up to a clear picture. From what I can tell, S800 and Z3370 are basically equal in varied tests and multi-threaded, but the Atom has a big advantage in browser benchmarks or anything that is really single-threaded. S800 has a clear graphics lead.

Might as well throw a blanket over them in performance terms.

 

[mikk]

And here it's much faster: http://pc.watch.impress.co.jp/img/pcw/docs/615/175/html/graph5.gif.html


S800 is still a good performer. The real difference is the efficiency, Silvermont can do the same with a much lower power usage. This is Intels real advantage.

 

[SiliconWars]

S800 is still a good performer. The real difference is the efficiency, Silvermont can do the same with a much lower power usage. This is Intels real advantage.

If a guess, dubious. If a statement, proof please.

 

[mikk]

If a guess, dubious. If a statement, proof please.

S800 is 5-8W SoC, Bay Trail-T apparently 2-4W. A SoC power consumption of 2,xW while running a multithread benchmark like Cinebench 11.5 or 7zip is a different world, ARM based SoC cannot do this with their higher performing one. Perf/watt ARM is lost against this.

 

[USER8000]

And here it's much faster: http://pc.watch.impress.co.jp/img/pcw/docs/615/175/html/graph5.gif.html


S800 is still a good performer. The real difference is the efficiency, Silvermont can do the same with a much lower power usage. This is Intels real advantage.

I am not sure I would trust that benchmark as the XPRT series has Intel involvement(it was even mentioned on the Anandtech website) and here:

http://www.xtremesystems.org/forums...y-in-Tablets&p=5150663&viewfull=1#post5150663

BapCo as you know was behind Sysmark,and Shervin Kheradpir(General Manager of Intel's Platform Evaluation and Analysis) was CEO of that IIRC,and is involved with Principed Technologies which is behind the XPRT series of benchmarks.

Luckily,we have other benchmarks to gauge performance of the various SOCs,but I would not touch any XPRT series benchmark with a bargepole. There are plenty of others we can look at!

 

[Sweepr]

And here it's much faster: http://pc.watch.impress.co.jp/img/pcw/docs/615/175/html/graph5.gif.html

Yup, not very surprising... this chip beats an A4 5000, of course its way faster than a S800 CPU-wise. Thats the overall picture, not some cherry picked benchmarks.

 

[mikk]

Or here: http://pc.watch.impress.co.jp/img/pcw/docs/615/175/html/graph9.gif.html


Physics Score 41% higher. Z3770 is easily faster than S800 from the CPU side no doubt.

 

[SiliconWars]

Yup, not very surprising... this chip beats an A4 5000, of course its way faster than a S800 CPU-wise. Thats the overall picture, not some cherry picked benchmarks.

So this one benchmark somehow negates 3-4 others that show S800 ahead or drawing with the Z3370?

Look - http://www.notebookcheck.net/SoC-Shootout-x86-vs-ARM.99496.0.html

See how Tegra 4 beats or comes close to Kabini in a few benchmarks as well? Tegra 4 is running very close to the Z3770 in a lot of today's benches.

Fact is, these are all very close in performance terms. Battery life surely matters more, as does graphics performance.

 

[Khato]

Fact is, these are all very close in performance terms. Battery life surely matters more, as does graphics performance.

I'll definitely agree that battery life matters more than a marginal increase in absolute CPU performance. Which is a good thing for Baytrail as I just remembered that Qualcomm actually did give us an indication of per-core power consumption for Snapdragon 800 - http://images.anandtech.com/doci/7082/Performance400.png - with the press materials accompanying their MDP benchmark session. I can be reasonably confident that the power listed is per-core since the DMIPS score for Snapdragon 800 on the chart tops out at around 8600 and the original Krait was around 3.3 DMIPS/Mhz per core. (Even 3.3 DMIPS/Mhz at 2.3 GHz would yield 7600 DMIPS, so it's clearly a single core.) Anyway, that chart is basically saying that Snapdragon 800 is using roughly 1335 mW, with all indications being that it's measuring only CPU core power. (Krait 200 using around 700mW CPU core power matches up with the results Anandtech obtained in their ARM vs x86 power analysis article.)

Now here's the important difference - we know that the Z3770 SoC never used more than 2.5W according to any of the reviews I've seen (please correct if I've missed one.) But how much power is the Snapdragon 800 chewing through whenever it comes to a multi-threaded benchmark? It could easily be using four times the above estimated amount, roughly 5.3W. Such would explain why it's suddenly ~35% slower in the 3dmark physics test - can't be using all that power on the CPU cores when the GPU needs to be active as well.

Edit: Forgot that we have the second data point on Snapdragon 800 of Anandtech's LG G2 review. Note that the 3dmark physics test goes from a score of 14828 on the MDP to 8596 in the smartphone, again implying that Snapdragon 800 performance is being limited by its power consumption when under heavy load.

 

[USER8000]

All these arguments about what CPU consumes less,etc might be important from a purely technical angle,but in reality they are mostly academic considering the screen,GPS and wireless components together with battery capacity are the main bottlenecks to decent battery life,in tablets and phones. Even the OS plays an important part in this.

Ultimately,obssessing over one part is not going to indicate,what phone or tablet will give the best battery life - it is the interaction of all components,ie,the balance between them.

Moreover,total product cost is the other factor,ie,the race to the bottom.

 

[Phynaz]

Fact is, these are all very close in performance terms. Battery life surely matters more, as does graphics performance.

An hour extra battery life is basically nothing. 50% faster graphics? Basically nothing.

You're falling all over yourself trying to win your anti-Intel position. You end up just invalidating your own posts.

 

[SiliconWars]

You're falling all over yourself trying to win your anti-Intel position. You end up just invalidating your own posts.

Not at all, it simply means there is a heirarchy in terms of performance, but price is the main factor regardless. All within reason you understand - so long as the subjective performance is on a similar level.

 

[CHADBOGA]

You're falling all over yourself trying to win your anti-Intel position. You end up just invalidating your own posts.

Well he has previously declared that BayTrail has already failed, so . . .

 

[SiliconWars]

It has failed, not because it's not good enough but because it's too late/not in any decent tablet. That's why I said it had already failed - there are no big market share tablets coming in the next year that haven't already been decided, and they all decided against Intel as far as I can tell.

Again, nothing against the performance of the SOC which is more than acceptable, but it doesn't matter how good you are when you are too late to catch the design refreshes.

 

[Phynaz]

Please provide us this list of "big market share tablets coming in the next year" and what CPU is in them.

 

[Phynaz]

price is the main factor regardless.

Apple is the first company that comes to mind that would disagree with the above statement. I'm sure you could list a lot more if you tried

 

[USER8000]

Apple is the first company that comes to mind that would disagree with the above statement. I'm sure you could list a lot more if you tried

Apple uses it own SOCs in its phones and tablets,so probably not such a good choice TBF. I expect price and flexibility are the reasons they have done this. In fact their massive profits are somewhat down to efficient production costs of their tablets and phones,considering the RRP of their designs.

There is even a thread on the A7:

http://forums.anandtech.com/showthread.php?t=2341762

Interestingly they seem to have the first 64 bit ARM chip,months before everyone else.

 

[Imouto]

Apple is the first company that comes to mind that would disagree with the above statement. I'm sure you could list a lot more if you tried

You should see Apple's market share lately in both tablets and phones.

 

[Nothingness]

And here it's much faster: http://pc.watch.impress.co.jp/img/pcw/docs/615/175/html/graph5.gif.html

Oh the benchmark Piednoel has been talking about for months on Twitter? i have a tendency to dismiss benchmarks a company representative keeps on claiming they are the best ones

That being said we need more CPU benchmarks than Geekbench for sure!