News Raspberry Pi 5 challenger LattePanda Mu uses Intel N100 CPU to take on industry-leading single-board computer

[Admin]

The gloves are off and the N100 totting LattePanda Mu takes aim at the flagship Raspberry Pi 5 hoping to best its stellar performance

Raspberry Pi 5 challenger LattePanda Mu uses Intel N100 CPU to take on industry-leading single-board computer : Read more

 

[peachpuff]

Comparing a raspberry pi5 to basically a mini pc? You can sometimes get minipc's between $100-$150 on amazon when on sale.

 

[HideOut]

This "article" is writen by an idiot. I mean come on... The "We've yet to confirm these details but in Geekbench 6 tests, LattePanda claims that the LattePanda Mu is twice as powerful as the Raspberry Pi 5 for multi-core operations, and just under twice as powerful for single core." is fun. I mean EVERYONE has benched both for months. A quick google search would get you some verifiable results. Its just lazyness.

 

[JamesJones44]

I'm curious about the power draw. I've seen several posts were the n100 locked to 6w TDP was still drawing 14+ watts of power for light workloads. The Pi 5 can often be under 5 watts for light workloads.

 

[TerryLaze]

I mean EVERYONE has benched both for months. A quick google search would get you some verifiable results. Its just lazyness.

Doesn't mean that the CPU will perform the same in whatever case (or lack thereof) the lattice will have. It might but it also might not, so it is unverified until tested.

 

[NeoMorpheus]

I'm curious about the power draw. I've seen several posts were the n100 locked to 6w TDP was still drawing 14+ watts of power for light workloads. The Pi 5 can often be under 5 watts for light workloads.

Considering that lying is a rewarded trait in anything related to intel, specially when it comes from their marketing dept, this simply follows the trend.

 

[Notton]

Comparing a raspberry pi5 to basically a mini pc? You can sometimes get minipc's between $100-$150 on amazon when on sale.

Having recently looked at Raspi5, I was shocked to see how expensive it was.
With an N100 mini PC, you get all the necessary stuff, including a 500GB NVMe SSD and 8~16GB DDR4 3200 SODIMM. This all costs around US$150~$220, depending on discount.

Where as with a Raspi5, the board itself "only" costs US$90~$100. However, when you buy a starter kit, or tack on all the extras you may want (Like a case, heatsink, fan, microSD card, power adapter, etc.), it easily hits US$150. If you toss in a NVMe HAT with 500GB SSD, it goes over US$200.

I figure people buy the raspi when they want to learn something new, or like to tinker.
You buy an N100 miniPC when you don't want to tinker with it. When you just need something that is easy to setup and works out of the box.

I'm curious about the power draw. I've seen several posts were the n100 locked to 6w TDP was still drawing 14+ watts of power for light workloads. The Pi 5 can often be under 5 watts for light workloads.

N100 is 6w average, but can hit a peak 15w, if the heatsink is adequate enough.
Something to also consider is power loss at the AC-DC power converter. DRAM, SSD, LAN, and Wifi card can also suck down extra power. N100 miniPCs are often paired with cheap chips, so they may not have the best power efficiency.

 

[NedSmelly]

Where as with a Raspi5, the board itself "only" costs US$90~$100. However, when you buy a starter kit, or tack on all the extras you may want (Like a case, heatsink, fan, microSD card, power adapter, etc.), it easily hits US$150. If you toss in a NVMe HAT with 500GB SSD, it goes over US$200.

IMHO application creep was beginning with Pi3 and more so with Pi4, with people using them for tasks traditionally done by x86 because "you could", rather than "you should" (e.g. NAS and Linux desktop). Somewhat unfortunate that Pi Foundation has leaned into this with Pi5, ending up with an SBC that falls into value/feature purgatory.

I wonder whether Pi Foundation should've just committed with two branches: embedded SBC such as Pi Zero/Pico, and a full-blown Arm 'NUC' form factor (instead of the current 'ports everywhere' layout).

 

[JamesJones44]

N100 is 6w average, but can hit a peak 15w, if the heatsink is adequate enough.
Something to also consider is power loss at the AC-DC power converter. DRAM, SSD, LAN, and Wifi card can also suck down extra power. N100 miniPCs are often paired with cheap chips, so they may not have the best power efficiency.

If you have a link to a review that measures package power I would appreciate if you can link to it. I've not seen a review where the power consumption was less than 10 watts in minimal work loads when capped to 6w Thermal Dissipation. In several of the reviews it was above 6 watts idle.

The issue is many are not measuring the CPU package itself, but given that a Raspberry Pi 5 rarely breaks 10 watts under normal workloads unless you get crazy with add-ons. It's hard to believe the N100 will be competitive in the low power consumption arena (at least on the cheap anyway).

 

[bit_user]

Having recently looked at Raspi5, I was shocked to see how expensive it was.

Where did you look?

With an N100 mini PC, you get all the necessary stuff, including a 500GB NVMe SSD and 8~16GB DDR4 3200 SODIMM. This all costs around US$150~$220, depending on discount.

$150 shouldn't count, since it's a promotional discount that is available to only a very small number of customers. We should compare standard pricing. The $190 price tag gets you just the compute module, lite carrier board, 64 GB eMMC storage, and active cooler. Not including case or power supply.

I'm not sure where you see a 500 GB NVMe SSD option. The RAM is already soldered on the compute module.

Compare that to the PiShop.US (an official US distributor) price for Raspberry Pi 5 (8 GB):

• $80.00 - Base price
• $ 6.95 - Active cooler
• $12.95 - 32 GB microSD with OS (I don't see a 64 GB option)

Total: $99.40

If you want to upgrade each computer to a cheap NVMe drive (e.g. the $34 TEAMGROUP MS30 512GB), then the PiShop.US price for the NVMe base is $18.95, bringing the total to $118.35. However, at that point, you could drop the 32 GB SD card, bringing the price to $105.90 (before adding in the NVMe drive, itself). With the NVMe drive, the total comes to $139.90.

Where as with a Raspi5, the board itself "only" costs US$90~$100. However, when you buy a starter kit, or tack on all the extras you may want (Like a case, heatsink, fan, microSD card, power adapter, etc.), it easily hits US$150. If you toss in a NVMe HAT with 500GB SSD, it goes over US$200.

Not sure where you're getting those prices, but you should say. Furthermore, you're not comparing apples-to-apples, because the LattePanda Mu starter kit doesn't include a case or power adapter, by default. LattePanda doesn't appear to give us the option of dropping the eMMC from their starter kit, but if we add in the same $34 drive to their kit's $190 base price, that brings it up to $224. For some people or institutions, the difference between $140 and $224 is very significant, particularly if we're talking about a cash-strapped school trying to buy machines for as many students as possible.

If we're talking about things like value for money or versatility, then there's no question the Intel board is the better option. However, a Raspberry Pi 5 is quite capable and good enough for a lot of purposes (including low-power desktop).

I figure people buy the raspi when they want to learn something new, or like to tinker.
You buy an N100 miniPC when you don't want to tinker with it. When you just need something that is easy to setup and works out of the box.

Raspberry Pi works out of the box. Always has, for me. It's actually the other SBCs that require varying degrees of tinkering to get working properly. IMO, one of the selling points of the genuine Raspberry Pi boards is how well they generally work and how good the support tends to be.

N100 is 6w average, but can hit a peak 15w, if the heatsink is adequate enough.

Source? LattePanda themselves say it can consume up to 35W!

 

[TerryLaze]

Source? LattePanda themselves say it can consume up to 35W!

Your source?! Because it's not this article since they say:
LattePanda Mu has an adjustable TDP between 6W and 35W.

So the whole system, while notton was talking about the n100 alone.
"This means that there is plenty of headroom for powering external devices from your carrier board."

 

[bit_user]

Your source?! Because it's not this article since they say:
LattePanda Mu has an adjustable TDP between 6W and 35W.

Here's what the fine print says:

"LattePanda Mu's innovative design allows tailoring the thermal design power (TDP) from 6 watts to 35 watts, making it versatile for various applications and performance requirements."
​

So the whole system, while notton was talking about the n100 alone.

The carrier board specs make it pretty clear they're not talking about system power, since even the Lite board's power specs are:

Power Input

• USB Type-C: 15V (Max 3A)
• DC 5.5x2.5mm: 12~20V (Max 10A)

"This means that there is plenty of headroom for powering external devices from your carrier board.

That's Les Pounder (article author) talking, and all he means is that if you subtract the compute module's power from the carrier board's input, the amount left over is your power budget for peripherals. However, he is not speaking as an authority on the power consumption of the SoC - he's just citing its default TDP that he saw somewhere.

The authoritative source is none other than Intel, themselves:

Intel® Processor and Intel® Core™ i3 N-series	N305	N300	N200	N100
TDP (PL1)/PL2	15W/35W (cTDP 9W)	7W/25W	6W/25W	6W/25W

Source: https://edc.intel.com/content/www/u...core-i3-n-series-datasheet-volume-1-of-2/001/
​

So, we know the N100 can use at least 25 W, since that's the default PL2. Furthermore, we see that some models default to a 35 W PL2, showing the SoC is capable of that much.

Finally, if we consult the datasheet:

"No Specifications for Min/Max PL1/PL2 values." (page 143).
Source: https://cdrdv2.intel.com/v1/dl/getcontent/759603
​

Not sure exactly what they're saying... either there are no upper or lower limits, or they're just not telling us what they are.

BTW, I just ordered a N97 board (different make/model), a couple days ago. I should be able to confirm the max value it lets me set PL1, when I get it.

 

[das_stig]

Strip out all the stuff you don't need, you could make a powerful pfSense/router/NAS type systems.

 

[evdjj3j]

This "article" is writen by an idiot. I mean come on... The "We've yet to confirm these details but in Geekbench 6 tests, LattePanda claims that the LattePanda Mu is twice as powerful as the Raspberry Pi 5 for multi-core operations, and just under twice as powerful for single core." is fun. I mean EVERYONE has benched both for months. A quick google search would get you some verifiable results. Its just lazyness.

At this point most Toms articles are just regurgitating the work of others. I'm not sure why AI isn't being used to generate these articles.

 

[evdjj3j]

Doesn't mean that the CPU will perform the same in whatever case (or lack thereof) the lattice will have. It might but it also might not, so it is unverified until tested.

Which the new Toms will never do.

 

[bit_user]

Which the new Toms will never do.

To be fair, Toms never tested mini PCs. A couple of ARM-based boads/boxes have gotten a somewhat cursory set of tests (Raspberry Pi included), but anything in the realm of Apollo Lake, Gemini Lake, Jasper Lake, or Alder Lake-N only gets their attention when they're scratching around for news to meet their daily article quota.

 

[Notton]

At the power consumption, it doesn't say if it's measured at the wall, the DC power jack, or through software, but I assume it's from the wall.
Lower down the power consumption measured from software is 6W

MeLE Quieter 4C Review (a pocket-sized fanless PC with Intel N100) - Liliputing

MeLE Quieter 4C Review (a pocket-sized fanless PC with Intel N100)

liliputing.com

I have a couple N100 mini-PCs. One from Intel NUC, the other is some NUC-like. They perform closer to this linked one below, and do 8W, but they also have active cooling.

https://www.notebookcheck.net/Testi...00-GB-SSD-for-less-than-200-USD.768945.0.html

I assume laptops/chromebooks with N100 would have a power configuration that is closer to 6W, but I can't find any reviews of them.
What I do know is that my N5100 equipped chromebook only consumes 10W with the 300nit IPS screen at max brightness.

 

[bit_user]

Lower down the power consumption measured from software is 6W

MeLE Quieter 4C Review (a pocket-sized fanless PC with Intel N100) - Liliputing

MeLE Quieter 4C Review (a pocket-sized fanless PC with Intel N100)

liliputing.com

Given that it has no airflow, nor room for a heatsink of any substance, of course that thing is going to run on the lean side!

Did you miss this part of the review?

"All the actively cooled Intel Processor N100 mini PCs I’ve seen have much higher power limits than the Quieter 4C. And whilst there are many factors that affect thermal throttling, fans are usually more effective at dissipating heat than passive cooling.

Some N100 mini PCs I’ve seen have Power Limit 1 set as high as 20 watts and Power Limit 2 set at 25 watts ..."
​

Below that, the reviewer compares the Q. 4C against GEEKOM Mini Air12, with the Q. 4C at PL1=20W and the actively-cooled Mini Air12 at PL1=15W (both have PL2=25W). And guess what? The Mini Air12 smoked it on virtually all the CPU benchmarks (except Geek Bench/Windows, where they performed roughly equal)! That proves what I said, which is that you picked a comparatively low-end, thermally-limited model to use as a data point.

The review's frequency plots also clearly show that even a power limit of 20 W keeps the N100 at well below its peak clock speed.

​

In that plot, once Tau is exceeded, the PL1=8W configuration drops the peak core clock speed to between 1.1 GHz and 1.3 GHz. At PL1=20W, the peak clock speed varies between 1.8 GHz and 2.0 GHz. So, the multi-core performance of that SoC is very much power-limited, even at 20 W!

 

[TerryLaze]

Given that it has no airflow, nor room for a heatsink of any substance, of course that thing is going to run on the lean side!

Did you miss this part of the review?

​

"All the actively cooled Intel Processor N100 mini PCs I’ve seen have much higher power limits than the Quieter 4C. And whilst there are many factors that affect thermal throttling, fans are usually more effective at dissipating heat than passive cooling.​

​

Some N100 mini PCs I’ve seen have Power Limit 1 set as high as 20 watts and Power Limit 2 set at 25 watts ..."​

​

Below that, the reviewer compares the Q. 4C against GEEKOM Mini Air12, with the Q. 4C at PL1=20W and the actively-cooled Mini Air12 at PL1=15W (both have PL2=25W). And guess what? The Mini Air12 smoked it on virtually all the CPU benchmarks (except Geek Bench/Windows, where they performed roughly equal)! That proves what I said, which is that you picked a comparatively low-end, thermally-limited model to use as a data point.

The review's frequency plots also clearly show that even a power limit of 20 W keeps the N100 at well below its peak clock speed.

​

​

In that plot, once Tau is exceeded, the PL1=8W configuration drops the peak core clock speed to between 1.1 GHz and 1.3 GHz. At PL1=20W, the peak clock speed varies between 1.8 GHz and 2.0 GHz. So, the multi-core performance of that SoC is very much power-limited, even at 20 W!

That's awesome!
The article' result match up with this at 15-25W but they also show 8-10W

So at 8-10W it's still well faster than the Pi5 at same-ish watts but you can also decide to run it at 25W to actually get twice the performance of the Pi.
And if it has even more headroom above 25W then 25W won't even bother the CPU any, you are not hurting/degrading it with that much safety headroom.

Geekbench 6.2.1
Single-Core Score	1191	1226
Multi-Core Score	2305	3249
OpenCL Score	3389	3315

 

[bit_user]

The article' result match up with this at 15-25W but they also show 8-10W

Yes, they show how much performance you lose/gain depending on how much power you're willing to budget. However, it's not a perfect analysis considering how the better-cooled GEEKOM Mini Air12 @ 15W/25W frequently outperforms the Quieter 4C at 20W/25W on multi-core tasks. That indicates either cooling or maybe its RAM is holding back the latter.

I was surprised that the iGPU seems to run unconstrained, within even the 8W/10W power limits. Then again, given that the Alder Lake-P laptop CPUs have iGPUs 4x as large, I guess I probably shouldn't be surprised at all.

 

[bit_user]

I just learned that Hardkernel recently announced their ODROID-H4 generation of x86 boards.

• https://www.hardkernel.com/shop/odroid-h4/
• https://www.hardkernel.com/shop/odroid-h4-plus/
• https://www.hardkernel.com/shop/odroid-h4-ultra/

They range in price from $99 to $220, but that also spans the 4-core N97 to the 8-core N305. They all support a DDR5 SODIMM, which is nice to see, since many of these boards only support DDR4. Compared to the LattePanda's soldered-down LPDDR5, regular DDR5 uses more power but has much lower latency.

For an extra $15, you can get a mini-ITX adapter kit:

H4 Mini-ITX Kit – ODROID

www.hardkernel.com

Another aspect I'm glad to see is that its M.2 slot supports full PCIe 3.0 x4 connectivity, if you use 2 SATA ports or less. I can't be sure, but I think the full eval carrier board for LattePanda might also support x4, if you're not using SATA. The specs seem rather vague. In just about every other Alder Lake N-series board I've seen, the 2280 M.2 slot connectivity is limited to x2 lanes or even just x1!

Some more detailed information is available on their Wiki:

https://wiki.odroid.com/odroid-h4/start
​

FWIW, Hardkernel is based in South Korea. I'm not sure where the boards are manufactured, however. I have previously ordered products direct from their website and found both shipping charges & shipping times (to the USA) to be pretty reasonable.