Bay Trail's not so bad... (N2830)

[Shivansps]

If you go by this and this, the Haswell Celeron U has about 2x the ST IPC of Bay Trail (172.143 vs. 88.333). I'm surprised the Haswell Celeron U isn't higher than 2x of Bay Trail though. If you have more comparisons, feel free to post them.

mm no really, from what ive see from my Z3735D and other internet sources the NotebookCheck numbers looks legit to me.

http://www.notebookcheck.net/Intel-Pentium-N3540-Notebook-Processor.117416.0.html

http://www.notebookcheck.net/Intel-Celeron-2957U-Notebook-Processor.104913.0.html

 

[Vesku]

On the other hand they have made it possible to get a 200.00 laptop, which was previously unheard of.

Well they've replaced almost all the big core Celerons in the low end so they've actually reduced what the customer gets for the money.

Now instead of $250 for an IvyBridge or Haswell Celeron Chromebook it's pretty much completely replaced with dual core Atom offerings. It's not much better with Windows laptops but at least they'll have some quad Atom models mixed in.

A bit annoying, just imagine if the Toshiba Chromebook 2 with its IPS 1080P screen had a 2955U or better in it instead of a N2840.

 

[waltchan]

Haswell's low 1.40 GHz frequency speed and without any turbo can still easily outperform Bay Trail's higher 2.16 GHz plus 2.41 GHz turbo number. It's all based on architectural design, not frequency speed. The Silvermont tablet processors need to be overclocked at 3.50 GHz in order to match Celeron 2957U 1.40 GHz Haswell processor speed.

 

[Shivansps]

Haswell's low 1.40 GHz frequency speed and without any turbo can still easily outperform Bay Trail's higher 2.16 GHz plus 2.41 GHz turbo number. It's all based on architectural design, not frequency speed. The Silvermont tablet processors need to be overclocked at 3.50 GHz in order to match Celeron 2957U 1.40 GHz Haswell processor speed.

You can see the numbers by yourselft, Ill agree that 3.5Ghz its about what is needed to match 2957U ST perf, but N3540 2.66Ghz turbos can already close that enoght to take a considerable lead in MT.

 

[ninaholic37]

mm no really, from what ive see from my Z3735D and other internet sources the NotebookCheck numbers looks legit to me.

http://www.notebookcheck.net/Intel-Pentium-N3540-Notebook-Processor.117416.0.html

http://www.notebookcheck.net/Intel-Celeron-2957U-Notebook-Processor.104913.0.html

Your ST numbers actually seem make Bay Trail look worse than mine did.

Celeron 2957U 1.4GHz vs. Pentium N3540 2.66GHz

Cinebench R15 Single Thread
55 vs. 43 (39.29 vs. 16.17... 2.45x faster)

Geekbench 3 Single Thread
1283 vs. 982 (916.43 vs. 369.17... 2.48x faster)

Cinebench R11.5 Single Thread
0.6 vs. 0.5 (0.43 vs. 0.19... 2.26x faster)

Cinebench R10 Single Thread 64-bit
2714 vs. 1959 (1938.57 vs. 736.47... 2.63x faster)

Cinebench R10 Single Thread 32-bit
2077 vs. 1385 (1483.57 vs. 520.68... 2.85x faster)

 

[waltchan]

Your ST numbers actually seem make Bay Trail look worse than mine did.

Celeron 2957U 1.4GHz vs. Pentium N3540 2.66GHz

Cinebench R15 Single Thread
55 vs. 43 (39.29 vs. 16.17... 2.45x faster)

Geekbench 3 Single Thread
1283 vs. 982 (916.43 vs. 369.17... 2.48x faster)

Cinebench R11.5 Single Thread
0.6 vs. 0.5 (0.43 vs. 0.19... 2.26x faster)

Cinebench R10 Single Thread 64-bit
2714 vs. 1959 (1938.57 vs. 736.47... 2.63x faster)

Cinebench R10 Single Thread 32-bit
2077 vs. 1385 (1483.57 vs. 520.68... 2.85x faster)

Welcome to the world of Bay-Trail tablet processors. Based on your numbers, looks like they need to be overclocked up to 5 GHz in order to match Haswell's 1.4 GHz speed (ouch).

 

[Shivansps]

Your ST numbers actually seem make Bay Trail look worse than mine did.

Celeron 2957U 1.4GHz vs. Pentium N3540 2.66GHz

Cinebench R15 Single Thread
55 vs. 43 (39.29 vs. 16.17... 2.45x faster)

Geekbench 3 Single Thread
1283 vs. 982 (916.43 vs. 369.17... 2.48x faster)

Cinebench R11.5 Single Thread
0.6 vs. 0.5 (0.43 vs. 0.19... 2.26x faster)

Cinebench R10 Single Thread 64-bit
2714 vs. 1959 (1938.57 vs. 736.47... 2.63x faster)

Cinebench R10 Single Thread 32-bit
2077 vs. 1385 (1483.57 vs. 520.68... 2.85x faster)

I have no idea what you are doing. The IPC is slower in BT, thats no news and who cares?

 

[ninaholic37]

I have no idea what you are doing.

It means you need a Bay Trail of 4GHz to match a Haswell Celeron U at 1.4GHz in ST (1.4 * 2.85) from the results in your link, basically.

 

[Shivansps]

It means you need a Bay Trail of 4GHz to match a Haswell Celeron U at 1.4GHz in ST (1.4 * 2.85) from the results in your link, basically.

Thats taking the worst case scenario, using the older software avalible, a more real scenario is about 3.45, and 2.66 is close enoght for MT perf to be considerable higher. Cinebench MT and 3dmark 2013 Physics score shows that.

 

[waltchan]

It means you need a Bay Trail of 4GHz to match a Haswell Celeron U at 1.4GHz in ST (1.4 * 2.85) from the results in your link, basically.

That's a very interesting and steep gap difference between the two to match its performance. Bay Trail tablet processors are not really suitable for PCs in the first place if they need to be overclocked at 5 GHz.

 

[seitur]

@Shivanps

In everyday multimedia and office usage you won't be using potential of 4 cores.

Same goes for kind of games you might run on big-core Celeron or Baytrail.

In those scenarios single thread matter most by far.

So I have no idea what 'real scenario' you have in mind.

 

[Shivansps]

@Shivanps

In everyday multimedia and office usage you won't be using potential of 4 cores.

Same goes for kind of games you might run on big-core Celeron or Baytrail.

In those scenarios single thread matter most by far.

So I have no idea what 'real scenario' you have in mind.

You are HORRIBLE WRONG, dont forget i have a Z3735D tablet and a E-350 notebook, and you know what? they both have almost the same ST perf, the Z3735D stumps my E-350 netbook for everyday usage. And in games it depends, some run better in the E-350 because of the igp, some run better in the Z3735D because of the cpu. And thats even considering that the netbook has 3Gb of ram and the tablet only 2GB.

 

[waltchan]

Update...

Processor quality and physical appearance are divided into several groups by Intel, ranked from slowest to fastest:

1. Embedded nano (Intel Quark)
2. Embedded tablet (Intel Silvermont)
3. Embedded notebook (Intel Haswell-ULT/ULX)
4. Socketed notebook (Intel Haswell-MB)
5. Embedded desktop (Intel Haswell-D)
6. Socketed desktop (intel Haswell-DT)
7. Socketed desktop performance (Intel Haswell-E)

 

[Shivansps]

Embedded nano is Silvermont too (E Series), quark is for wareables.

 

[waltchan]

If you look at Wikipedia, Intel Silvermont aims at tablets, hybrid devices, netbooks, nettops, and embedded/automotive systems.

http://en.wikipedia.org/wiki/Silvermont

It does NOT say desktops or notebooks in the introduction, which means Silvermont is NOT intended for desktop and notebook, so I call it "tablet" processor in desktop and notebook as correct word.

 

[ninaholic37]

mm no really, from what ive see from my Z3735D and other internet sources the NotebookCheck numbers looks legit to me.

http://www.notebookcheck.net/Intel-Pentium-N3540-Notebook-Processor.117416.0.html

http://www.notebookcheck.net/Intel-Celeron-2957U-Notebook-Processor.104913.0.html

Sorry, I think I misinterpreted your post here. When you said "mm no really" I thought you were disagreeing and giving links as to why, but you were probably saying "mm no really" meaning you didn't have many comparisons easily ready either. My mistake.

Most of the stuff I do is still single-threaded (I think Firefox still is, mostly?), but considering that Intel themselves label the Haswell as a Celeron and the Atom as a Pentium, they may also feel that the Atom Pentium is better (or at least just as good as the Haswell Celeron) in general, as you say. I wouldn't mind a Quark laptop, I still have MS-DOS and some games that can run on 386. :awe:

 

[Seba]

You are HORRIBLE WRONG, dont forget i have a Z3735D tablet and a E-350 notebook, and you know what? they both have almost the same ST perf, the Z3735D stumps my E-350 netbook for everyday usage. And in games it depends, some run better in the E-350 because of the igp, some run better in the Z3735D because of the cpu. And thats even considering that the netbook has 3Gb of ram and the tablet only 2GB.

So you say that because Z3735D Bay Trail performs better than AMD E-350, he is wrong about Haswell outperforming Bay Trail (by a large margin)?

 

[krumme]

Op didnt we tell you to stop buying that crap? Lol

 

[monstercameron]

Op didnt we tell you to stop buying that crap? Lol

He can't help it...

 

[coercitiv]

So you say that because Z3735D Bay Trail performs better than AMD E-350, he is wrong about Haswell outperforming Bay Trail (by a large margin)?

He says 4 Threaded CPUs will perform better than 2T CPUs in Windows as long as they have somewhat similar ST performance for the moments when that matters.

 

[coercitiv]

Op didnt we tell you to stop buying that crap? Lol

The thread was resurrected as a starting point for a forum debut.

 

[dahorns]

Your ST numbers actually seem make Bay Trail look worse than mine did.

Celeron 2957U 1.4GHz vs. Pentium N3540 2.66GHz

Cinebench R15 Single Thread
55 vs. 43 (39.29 vs. 16.17... 2.45x faster)

Geekbench 3 Single Thread
1283 vs. 982 (916.43 vs. 369.17... 2.48x faster)

Cinebench R11.5 Single Thread
0.6 vs. 0.5 (0.43 vs. 0.19... 2.26x faster)

Cinebench R10 Single Thread 64-bit
2714 vs. 1959 (1938.57 vs. 736.47... 2.63x faster)

Cinebench R10 Single Thread 32-bit
2077 vs. 1385 (1483.57 vs. 520.68... 2.85x faster)

I'm confused by this post. None of those numbers in parenthesis appear anywhere in the notebookcheck links. What are you looking at?

 

[Seba]

He says 4 Threaded CPUs will perform better than 2T CPUs in Windows as long as they have somewhat similar ST performance for the moments when that matters.

And since Haswell and Bay Trail ST performance is far from similar, his comparison is irrelevant.

 

[coercitiv]

And since Haswell and Bay Trail ST performance is far from similar, his comparison is irrelevant.

Comparison is not irrelevant, being the 20% slower in ST means BT is still 50% faster in MT loads.

 

[Seba]

20% slower? See the benchmarks from this page.