News Intel E-Core-Only N100 Gaming Benchmarks Emerge

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Add ARM variants to those tests. I would love to see the performance per watt battles at the low end to give some perspective of where we are with the two architectures.
I finally found the MediaTek SoC I was trying to remember:

Here's a review of a board using it:

It's pricey, because it's a development kit and the actual SoC is on a daughter card. The first RK3588 board I saw was also like that. I fully expect to see more cost-effective options. Maybe not Orange Pi 5 cheap, but hopefully not far off.

Using just one metric, CoreMark, it's 39.8% faster than the RK3588. Then again, the same metric puts it in the ballpark of a i7-2600. So, that's not terribly promising. But, if MediaTek is right that the absolute maximum power draw is only 8 W, it still could be a strong alternative to something like the N100.
 
Last edited:
  • Like
Reactions: JamesJones44
Yes, but we have to do the best we can with the data available. If you have better data, please share.


This would be a big revelation, if you could substantiate it. For instance, if true, then Rome (Zen 2-based EPYC) should've left no market opportunity for Sierra Forest (Gracemont-based Xeon). All AMD would've had to do is port Zen 2 to TSMC N5 or N4 and call it a day.

I've seen only one detailed performance comparison involving Zen 2 and Gracemont. However, there's a massive disclaimer right at the start of it:
"Please take power measurements in this section with a grain of salt. Research suggests that AMD doesn’t have power measurement hardware, which limits the accuracy of their RAPL counters"​

You can go on to read the article, but that basically makes their Zen 2 comparisons irrelevant, other than perhaps the shape of the curves.

The best we could do is compare package power of a Zen 2 APU against a N300, in order for it to be a fair test. If you're aware of such a comparison, please share.
The reason Zen 2 isn't being done in that way is because scaling up Zen 2 would be very area inefficient, Zen 4 in a similar area (slightly smaller even) even though is on a more advanced node uses less die area and Zen 2 really doesn't scale much lower in power draw either.
Zen 4c however uses a lower power library that lets them scale lower

N305 uses 15W anyway and if you put one next to a 7840U, it's looking very unfortunate for gracemont
 
The reason Zen 2 isn't being done in that way is because scaling up Zen 2 would be very area inefficient, Zen 4 in a similar area (slightly smaller even) even though is on a more advanced node uses less die area and Zen 2 really doesn't scale much lower in power draw either.
Zen 4c however uses a lower power library that lets them scale lower

N305 uses 15W anyway and if you put one next to a 7840U, it's looking very unfortunate for gracemont
My proposed experiment was to compare the performance and package power of a 4800U with N305. That would be a proper test of your claim about Zen 2's efficiency. Let us know if you find such a comparison or have any other data to substantiate your claim.

It should hopefully go without saying that Phoenix is more efficient.
 
My proposed experiment was to compare the performance and package power of a 4800U with N305. That would be a proper test of your claim about Zen 2's efficiency. Let us know if you find such a comparison or have any other data to substantiate your claim.

It should hopefully go without saying that Phoenix is more efficient.
Per work done, 4800U finishes much faster than gracemont cores even at gracemont sweetspots even though 4800U power won't much lower per core (which don't exist on alder lake S)
But bear in mind N305 has severely cut down IO so it's not really that good of a comparison
 
Per work done, 4800U finishes much faster than gracemont cores even at gracemont sweetspots even though 4800U power won't much lower per core (which don't exist on alder lake S)
That's consistent with what you've said, but I have yet to see you post up the results of any independent testing to support these claims. You clearly have some reason you believe those claims are true, so all I'm asking is for you to tell us why. If you don't have it, that's okay, but we're at a point where I think this exchange is entirely unproductive without it.

But bear in mind N305 has severely cut down IO so it's not really that good of a comparison
I thought about that, but doesn't the 4800U just have like x16 PCIe 3.0 lanes? That's not too much more than the N305's x9 PCIe 3.0 lanes, IMO.

The bigger issue might be what the N305 is using for DRAM. I'd hope it were tested with DDR5, to at least be somewhat fair.
 
Anyway, I have to agree that the idle power is a lot higher than I'd like to see. 8 W would be okay. 22-23 W is in the range of what I'd expect for peak active power. The main question is whether the manufacturer customized the PL2 threshold. Given that it's fanless, I'd have to guess they did.
You were pretty accurate for power on your guess. Check out Robtech video where he reviews a fanless N100 and it idles at 9W and draws 22W max or 25W when OC (PL1/2 raised to 30watts)
View: https://www.youtube.com/watch?v=HaLhuwHtALI&start=425
OSRDFJw.png

 
  • Like
Reactions: bit_user
Speaking of buying...
You can't buy only the motherboard + cpu combo right?
You are forced to buy already pre-packaged things?

I ask because its very hard to find what I want for.
 
Speaking of buying...
You can't buy only the motherboard + cpu combo right?
You are forced to buy already pre-packaged things?
There are actually a few motherboards available with these SoCs.

The consumer-oriented models I know about are:

All except the ASRock N100M are mini-ITX. Read the specs carefully. One issue is they all use single-channel DDR4.

Some industrial boards are:

There are a few other brands of industrial boards with these SoCs, but they tend to be somewhat overpriced.
 
  • Like
Reactions: thestryker
Speaking of buying...
You can't buy only the motherboard + cpu combo right?
You are forced to buy already pre-packaged things?

I ask because its very hard to find what I want for.
Curious what your use case is that you can't find what you're looking for, but bit_user's list is about the best you're going to get just board wise. I'm not sure there are any available boards higher than the N100 right now, but there are some SBC N305 though I haven't seen any retail availability.
 
  • Like
Reactions: bit_user
I'm not sure there are any available boards higher than the N100 right now, but there are some SBC N305 though I haven't seen any retail availability.
One problem is that even if you can find an announcement of a board you want, trying to actually find it for sale is another problem. Here's one that claims to support up to the N305, but the AliExpress store of Maxtang doesn't carry any version of it, nor have I found anyone else selling it:

Here's another one I ran across, but they say the minimum order size is 20 units:

If you're flexible on the actual form factor, you could buy this one today:


Up Board works very closely with Intel, and has even made development kits for them. So, it's a good bet they're top quality. On the plus side, it has LPDDR5. The down-sides are that it's soldered down and that I think the form factor is proprietary (or, at least not a common PC form factor).
 
Curious what your use case is that you can't find what you're looking for, but bit_user's list is about the best you're going to get just board wise. I'm not sure there are any available boards higher than the N100 right now, but there are some SBC N305 though I haven't seen any retail availability.
Was looking for something that isnt above 500 USD and that can handle a smaller SFP+ card for 10GB networking.

I guess I will have to compromise for 2.5Gigs for my home lab.
 
Was looking for something that isnt above 500 USD and that can handle a smaller SFP+ card for 10GB networking.

I guess I will have to compromise for 2.5Gigs for my home lab.
I suggest you take another look at the specs of the two ASRock boards, then. One has a PCIe 3.0 x4 slot, with x2 lanes connected. The other has a x16 slot with x2 lanes connected.

Two lanes @ 3.0 speeds is roughly 16 Gbps in each direction, and therefore plenty of bandwidth for running a single-port SFP+ card. You could even run a dual-port NIC, with it being only capable of 80% bandwidth when both ports are getting hammered.

You could get something like this, but I don't think you'll do any better in a small form factor.

Just be aware that they're not using E-core CPUs, though. So, you get more I/O, but at the expense of some efficiency. Also, it does have a fan.
 
Last edited:
Just be aware that they're not using E-core CPUs, though. So, you get more I/O, but at the expense of some efficiency. Also, it does have a fan.
With the potential power consumption of all the IO I'm guessing the case sinks the heat from the secondary networking and the fan/copper plate handling CPU + primary networking.

I wish they'd moved over to a platform based on the pentium gold 8500/8505. There's one of the 6x 2.5gb port ones with the 8505, but I think that's the only place I've seen it. Seems like that's the perfect chip for these devices as long as you power limit it.
 
I suggest you take another look at the specs of the two ASRock boards, then. One has a PCIe 3.0 x4 slot, with x2 lanes connected. The other has a x16 slot with x2 lanes connected.

Two lanes @ 3.0 speeds is roughly 16 Gbps in each direction, and therefore plenty of bandwidth for running a single-port SFP+ card. You could even run a dual-port NIC, with it being only capable of 80% bandwidth when both ports are getting hammered.


Just be aware that they're not using E-core CPUs, though. So, you get more I/O, but at the expense of some efficiency. Also, it does have a fan.
The asrock ones definitively looks interesting.