Intel Celeron N3050 & Pentium N3700 Review: Entry-level "Braswell" SoCs for small desktops

[Steve]

When Intel announced an accelerated roadmap for its Atom SoC range in May 2011, the company's 22nm Silvermont microarchitecture was scheduled for a 2013 release and would later receive a 14nm die shrink in 2014, codenamed Airmont. However, setbacks in developing the 14nm process delayed the arrival of Broadwell parts, especially the high performance variants.

We first got our hands on the desktop Bay Trail-D Asrock D1800M (Celeron J1800), Q1900M (Celeron J1900) and Q2900 (Pentium J2900) back in August 2014. To recap, we found that the Celeron J1900 was a better value than the J2900 as it offered a similar level of performance at a much lower price, though we'd still go with AMD's AM1 platform for anything 3D related.

Now we have Intel's new Braswell SoCs, which promise to be faster across the board while also consuming less power. The new chips are manufactured using Intel's 14nm process and include the mobile Celeron N3000, N3050, N3150, and Pentium N3700. The last three will be made in versions for desktop systems and today we have the N3050 and N3700 in hand for testing.

Read the complete review.

 

[Evernessince]

So there's really no point to these at all? AMD is promising 2x GPU performance and a hefty CPU performance upgrade with their new APUs coming out very soon.

 

[Cryio]

I still don't understand the utility of these low-end CPUs. You can get the job done just as well with a AMD APU. Better even.

 

[SuperVeloce]

While those things really are weak anyways, it would fare better if cpu would get more % of the TDP to play with. They beefed up the gpu, but it is too weak anyways even for many older games. And for many those older games slow cpu cores could be a problem (like Battle for middle earth, if you can believe it).

Braswell with baytrail-like gpus would be a better bet for those simple office/net machines

 

[bixler]

It would be good to know what kind of ram was used in the benchmarks, since the Braswells could profit a lot of a dual channel ram configuration.
In generall , I do not understand why the review has such a negative vibe. Taken the results of the benchmarks, this is the best HTPC solution available. It requires 7-16w less then the Athlon 5350, is absolutely noiseless and has the more powerfull graphics. Myself, I also expected a bit more, but this is not the N3700s fault. The Braswells should be judged by the facts, that are delivered with the benchmark scores, not by someones expectations.

 

[Steve]

It would be good to know what kind of ram was used in the benchmarks, since the Braswells could profit a lot of a dual channel ram configuration.

We would never test such a board in single-channel mode, these boards only have two memory DIMM’s and they were both populated with 4GB modules for testing.

In generall , I do not understand why the review has such a negative vibe. Taken the results of the benchmarks, this is the best HTPC solution available. It requires 7-16w less then the Athlon 5350, is absolutely noiseless and has the more powerfull graphics. Myself, I also expected a bit more, but this is not the N3700s fault. The Braswells should be judged by the facts, that are delivered with the benchmark scores, not by someones expectations.

If you don’t understand why the review had such a negative vibe then you should carefully read the final page.

I am not sure why you think Braswell wasn’t judged by the facts as you say. The fact is when compared to Bay Trail-D, Braswell was consistently slower and didn’t prove to be any more efficient. The improved graphics performance is fine but at the end of the day does a 20% performance boost really matter? Especially when we are talking about a few extra frames on a platform that can’t really deliver playable performance at 1280x800 even when using the lowest quality in-game settings, ie Tomb Raider.

 

[bixler]

Hi Steve,
I did carefully read the last page. In the review you write, that "the N3700 was almost always at least 10% slower than the J2900". On the other hand it needs at least 25% less power in most scenarios. This being said, it seems that the difference in energy savings is bigger than the difference in performance. According to this, the N3700 is more efficient than the J2900.

Furthermore the percentage in energy savings would be even greater, when you consider, that these kind of computers are usually not driven by 500w power supply units like in your review, but more efficient ones like a pico psu or power supplies like the ones of an Antec ISK 110 case. The lower the total consumption gets, the bigger is the percentage of a difference of lets say 5w is (and the N3700 easily saves 5w in comparison with the J2900).

What I also do not get in your argumentation, is that you claim 20% of improved graphics are not that important, because most games are still unplayable but the 10% the J2900 is faster are important. You use one as an argument against the N3700 but do not accept the even bigger difference in graphics as an argument for the N3700. If 20% approvement in graphics do not make a difference, what difference do 10% in speed matter, especially when it should be clear, that those boards will be used as an HTPC, a firewall or an office-pc.

Also, maybe those 20% in graphics are exactly what is needed to play an older game with acceptable framerates. According to this review League of Legends and FIFA 14 should be easily playable in 768p http://laptopmedia.com/news/what-ga...a-low-voltage-braswell-processors-from-intel/ .

It seems to me, that the lower comsumption, the improved graphics and the extra features this boards brings (I.e. 4K, H.265) are more worth in its predestinated scenario as a HTPC or similar, than a drop of 10% in speed is. And here we are again, when I say, I found that review to have a negative vibe. To me, eventhough I am myself a bit disappointed with the benchmark results in your review, the N3700 seems to be the best buy for an HTPC these days.

P.S. English is not my mother language. I hope I could make myself clear anyway.

 

[Steve]

Hi Steve,
I did carefully read the last page. In the review you write, that "the N3700 was almost always at least 10% slower than the J2900". On the other hand it needs at least 25% less power in most scenarios. This being said, it seems that the difference in energy savings is bigger than the difference in performance. According to this, the N3700 is more efficient than the J2900.

Furthermore the percentage in energy savings would be even greater, when you consider, that these kind of computers are usually not driven by 500w power supply units like in your review, but more efficient ones like a pico psu or power supplies like the ones of an Antec ISK 110 case. The lower the total consumption gets, the bigger is the percentage of a difference of lets say 5w is (and the N3700 easily saves 5w in comparison with the J2900).

If it difficult to accurately compare percentages that way since we are looking at different numbers power vs. performance that aren’t directly comparable.

Again the percentages wouldn’t change with a lower watt power supply as all systems were tested equally. The 500w unit used while overkill is still very efficient anyway.

What I also do not get in your argumentation, is that you claim 20% of improved graphics are not that important, because most games are still unplayable but the 10% the J2900 is faster are important. You use one as an argument against the N3700 but do not accept the even bigger difference in graphics as an argument for the N3700. If 20% approvement in graphics do not make a difference, what difference do 10% in speed matter, especially when it should be clear, that those boards will be used as an HTPC, a firewall or an office-pc.

Also, maybe those 20% in graphics are exactly what is needed to play an older game with acceptable framerates. According to this review League of Legends and FIFA 14 should be easily playable in 768p http://laptopmedia.com/news/what-ga...a-low-voltage-braswell-processors-from-intel/ .

It seems to me, that the lower comsumption, the improved graphics and the extra features this boards brings (I.e. 4K, H.265) are more worth in its predestinated scenario as a HTPC or similar, than a drop of 10% in speed is. And here we are again, when I say, I found that review to have a negative vibe. To me, eventhough I am myself a bit disappointed with the benchmark results in your review, the N3700 seems to be the best buy for an HTPC these days.

P.S. English is not my mother language. I hope I could make myself clear anyway.

Okay there are two separate performance claims here and you seem to be confusing them as being the same.

The N3700 is around 20% faster for 3D work. The problem is you wouldn’t buy the N3700 for gaming, even basic gaming. That 20% boost in rendering performance doesn’t really buy you anything in terms of usable performance, games are still wildly unplayable.

Without question the most important performance aspect of these budget systems is how well they work with office related software such as Word and Excel for example. Giving up performance here is unacceptable and becoming 10% slower on a platform that is already very slow doesn’t make sense.

Gaming shouldn’t be the priority here and that is why we were very disappointed with a 10% reduction in everything non-3D related.

Had Intel matched the clock speeds of Bay Trail-D with Braswell then we feel it would be a much more impressive platform.

 

[bixler]

Hi Steve,

it seems as we just have a different view, as far as the N3700 is concerned.
I can not see, how 10% lower speed, even at this low level, do make a difference in programs like word. For me it does not matter, if I have to wait a few seconds more, both the J2900 and the N3700 are rather slow here, we do agree on this at least. Furthermore I can not remember to have ever waited in word, not even 10 years ago on a single-core cpu.

I also ask myself, why does the N3700 Asrock-board support 4K, H.265 and has an optical out if it is supposed to be an office machine? In my view this is clearly a product haeding for the HTPC market. And this is where light gaming can be an option and 20% increase in graphics may matter. Noone buys the N3700 for gaming, but I would not mind plaing a game of FIFA 14, which is possible here in 720p and low settings.

You say "the percentages wouldn't change with a lower watt power supply as all systems were tested equally". This is not true. In idle the N3700 needs 21w in your review, the J2900 needs 29w. When using a more efficient PSU, the difference in watts between those two systems should stay about the same, whereas the total consumption should decline. The smaller the total usage gets, the greater is the percentage of the difference between those two systems.

I still think the review is too negative and I am curious how other sites will rate the N3700.

 

[Steve]

You say "the percentages wouldn't change with a lower watt power supply as all systems were tested equally". This is not true. In idle the N3700 needs 21w in your review, the J2900 needs 29w. When using a more efficient PSU, the difference in watts between those two systems should stay about the same, whereas the total consumption should decline. The smaller the total usage gets, the greater is the percentage of the difference between those two systems.

I can’t argue with that math or logic, or at least I am not going to

Food for thought though since I am wrong. Hypothetical situation, with a more efficient PSU the N3700 idles at 1 watt which means if the difference remains the same the J2900 would idle at 9 watts using 800% more power…

 

[bixler]

If what I have written about the consumption is based on a logical mistake in my understanding, I am sorry for the confusion. This does not concern the other points I made about the N3700, though.

 

[Steve]

If what I have written about the consumption is based on a logical mistake in my understanding, I am sorry for the confusion. This does not concern the other points I made about the N3700, though.

That’s fine. You are right though ... we do disagree regarding Braswell and how it compares to Bay Trail-D so we might as well leave it at that

 

[Guest]

I thought we'd see Celeron 2957U in the benchmark.

How does it fit in all these tests?

 

[Steve]

I thought we'd see Celeron 2957U in the benchmark.

How does it fit in all these tests?

It's not a desktop CPU...

 

[Guest]

Not on paper
but its very common in many Chromebooks and Small factor desktops.
Is it possible at all to get a comparison between the J2900 and 2957U ? Search online but no luck

 

[Guest]

Hi steve and thenk you for this review : I was waiting for it!
I would like to ask you if you can test h.265/HEVC 1080P video playback.
I will surely buy a N3700 if it's capable to play H265. Is it ?

Thank you

 

[Oneout]

I don't see why the AMD option can't be recommended as the Athlon 5350 and Asus AM1I-A costs the same as Pentium N3050/3150 but tops many of the charts in performance. N3700 is more expensive and performance is lackluster compared to J2900.

The power consumption results are all but useless with such a high wattage and low efficiency PSU. For comparison, the Athlon 5350, Asus AM1I-A, 4gb ram and 240GB ssd idles at 14.8W when used with LC-power 60W PSU that is included in the tiny m-ITX case. I've also built two dual-core Haswell systems with 360W and 400W 80 plus gold efficient power supplies and that idle at 22W.

Other than that nice review, though as you conclude Braswell is quite dissapointing in both performance and efficiency improvements.

 

[smarty2015]

All we have to do is wait for the N3800...N3900 processors
Knowing Intel's strategy I am sure that the Pentium N3900 will be clocked 50% higher...

 

[Guest]

Nevertheless, it does bridge the gap between Intel and AMD for ultra-budget 3D performance, so much so that the Pentium N3700 is faster than the Athlon 5350.

Even tho the Athlon 5350 is faster than the Pentium N3700 in 4 of the 6 games?
Oh and let's not mention the AMD Sempron 3850 beating it in 3 of 6 games, one being equal.
So Intel can only make a more efficient than AMD APU with a die shrink?
It's kind of laughable that an AMD 28nm APU is better than an Intel 22nm APU and requires an Intel 14nm APU to even be on any terms of competition.

Maybe once Intel stops using loss leader tactics with Atom's we can see a more fair market share.

And someones going to come comment on this about Bulldozer and all AMD vs Intel high end, which really isn't relevant to what I'm trying to say here.

 

[Guest]

...and now overclock Athlon 5350 to 2.8 Ghz and Athlon 5150 to 2.2 Ghz and check, if they perform better.

 

[Guest]

Hey there, I would really be interested in AES encryption, as the N3050 and N3700 use AES-Ni, but the older J1900/2900 don't have this feature. Could you provide some data about that ? Would be really usefull for my decission which SoC/CPU I will use for a fully encrpyted system..

 

[Guest]

I'm having trouble with the power efficiency math as well. The power consumption figures are inadequate on their own so I checked the review of the PSU. Since the PSU consumes a fixed 9 watts+, and after that all power is supplied at 90% efficiency at nearly all power draw levels. Thus the N3050-ITX would be consuming 5 watts when idle, whilst e.g. D1800M would consume 14 watts. I.e. the Braswell is in this case would be nearly 3 times more efficient. But why is the N3700-ITX power consumption so high? Surely it's not the cpu idle power draw as they are nearly the same but rather some of the extra circuitry on the motherboard? And not sure which model they are referring to in the article when claiming "That said, the Celeron N3150 still consumed 24% less power than the J1800." since N3150 not even being reviewed.

 

[Guest]

On a second thought, the power consumption makes no sense. Not with those differences that have little to do with the motherboard. Some better power reading benchmarks:
http://www.technikaffe.de/anleitung-307-asrock_n3150_itx_im_test__14nm_braswell_vs._22nm_bay_trail
http://www.hardware.fr/articles/936-6/consommation.html

 

[mcdull]

One thing is definitely missing from the review is comparison of HEVC/h.265 playback. All low power HTPC motherboard, especially fanless, are struggling in playing back 1080p HEVC video as there is no hardware decoding. The new motherboards are supposed to support HEVC which can make a huge difference compared to J2900 for a HTPC setup.

 

[255BB]

You have not mentioned about HEVC hardware decoding at all.These N3050 ,N3700 (via HDMI 1.4a) can play 4K30p HEVC(h265) very smoothly with the right softwares. And they can play 1080p60 with no problem. They are meant to be an HTPC, aren't they? (I have N3050 NUC and I have tried some 4K files)