I Tested Two $699 Laptops for Coding… It Got Weird

Ah, they did it. Dell's new XPS13 cost the same amount as Apple's MacBook Neo. I wanted to answer how it is for developers, so I ran some dev tests on them. So, this is the base XPS13, Intel's new Core i5 Wildcat, 8 GB of RAM, 512 gigs SSD, and this here is the famous MacBook Neo. It's an A18 Pro chip in there. So, this and this have the same chip in it. 8 gigs of memory, and this one has 256 gig SSD. Also, there's no fan in here. These two are about the same size. The lid goes back just as far. The XPS weighs 1,56 g and the Neo weighs 1,241 g. Both are extremely light laptops. One finger open on both of them. XPS's have come a long way. Last time I tested one of these, it wouldn't even sit flat on a desk. Now, it's good. But, did they fix the rubber feet? I still don't like the strip. I don't like the strip because it expands and contracts. My wife's still kept coming apart, so I had to recut it and reglue it back in there all the time several times. That's why I prefer the four rubber feet of the MacBook. I don't know, is that patented or something? Why can't more companies adopt that? It's easy. One cool thing that the Dell has is a backlit keyboard and I can actually cycle between different levels of it, but Neo doesn't have that. So, in the dark, it's uh hard to see the keys. With Apple, you have to get the MacBook Air to get the backlight. As far as the trackpad, MacBook Neos clicks anywhere you press. Whereas the Dell's, they're

using the old diving board design, so you have to press towards the bottom. The top doesn't register. The screen on the Dell goes up to 120 hertz. And it's also a touch screen. I can't believe we get some of these features in the Dell for $699. That's crazy to me. The Neo doesn't have touchscreen, obviously, and it only has 60 Hz. Scrolling code on the XPS is just smoother. and under this lamp. Not that you necessarily have a lamp like this, but just as a demonstration. You see the XPS screen is matte. This is glossy. Neither one is going to look good outside. And this is at maximum brightness with a high contrast theme in VS Code. Now, ports. The Neo has a headphone jack and the XPS doesn't. Come on, Dell. On a laptop for students, no headphones. Courage. Both of them have two USBC ports. I like the way that Dell places them. one on each side for convenience. The MacBook Neo has both USBC ports on one side and nothing on the other. And both of Dells are USB 3.2, so they have the full 10 gig speed, whereas the Neo, only one of them is 10G and the other one is the old 480 megabit. And I always forget which one is which cuz they're not labeled. So yeah, that would be nice if they were labeled. Storage XPS has double. Yeah, the speed the speed up is huge. Uh, also smaller SSD chips are not as fast as the larger ones. And this definitely shows. Dell is

using really nice SSD speeds here. This one, what's going on? That's very slow. So sequential, that's when you're copying a large file. For example, Dell is at 6,860 for read, 4,873. The MacBook Neo, 1,735, and 1500 for write. That's really kind of bad. The other number that matters a lot is that random, especially if you're doing development and you're compiling code. Lots of little files here and there being read and written. And here on the Dell, we have slightly worse numbers for read. 237. MacBook Neo has 549 there. And the right on the Dell is much better. 635 versus MacBook Neo's 52. And the XPS drive is actually removable. You can swap it out for a larger one. on the Neo 256. That's what you get and you're stuck with it forever. Now, I will give the Neo this. When the lid is closed and you pop it open, the Neo is on and available just like that. The Dell a little bit slower, but still not too bad. I spend a ridiculous amount of time jumping between different computers and networks. And not all those networks are ones that I completely trust. Now, considering how much of my work happens online, I'd rather not have my traffic exposed to whoever happens to be on the same network. And that's where Surf Shark comes in. I just connect and I'm good to go. One tap and all my internet traffic, including my git pushes and package downloads, travel through an AES 256 encrypted tunnel. There's also clean web, which helps block ads and trackers in the background. I don't even have to think about it. They've also added scam

text protection, which checks incoming text messages for signs of scams before you accidentally click on something dangerous. Surf Shark doesn't keep activity logs, and their RAM only servers don't retain data after a reboot. Their no logs policy has even been independently audited. Now, my favorite part is that I don't have to worry about how many devices I'm using. One subscription covers my laptop, my phone, and whatever devices I happen to be using. It just gives me a little peace of mind when I'm accessing servers or doing development work away from home. Head over to surfshark.com/alexiskin and grab four extra months free on your plan. And if it's not for you, there's a 30-day money back guarantee. Check out the link in the description. Stay private, keep doing your thing, and now back to the video. All right, enough poking at it. Let's see some performance in code. Now, historically, Intelbased machines had to be plugged in to get the best performance. And I do have it set on best performance mode. The Apple Silicon Max have the same performance characteristics whether they're plugged in or not. And I like this little Sabrant device. It can actually charge multiple machines. Very useful for me because I have a bunch of these that I need to charge and test all the time.

and it actually puts out all that power from this little box. It also shows me how much wattage each port is drawing. These were not plugged in for probably the last hour as I was testing them. And right now, as soon as I plug it in, the Dell is drawing 55 watts while the MacBook Neo is drawing 31 watts. I'm going to test the Python performance of these. And I first want to do it plugged in to see what the difference will be plugged and unplugged. So, first up is the Python Matter test and go. All right, here we go. This test is crazy because it pegs all the cores to the maximum. It's the Mandelroad algorithm. You know, Ben Juan Mandelroad, he created the fractals patterns discovered, I should say, not created. I'm here in the XPS. Of course, the MacBook Neo is fanless. This is an intense test. So, right now, just about 45 seconds into it, the uh palm rests still feel cool on the MacBook Neo. They feel warm on the Dell, but not like uncomfortable. Just a little warmer, that's all. Ooh, wow. That fan really helps out. The Dell fan chip, one of those MacBook Neo still going. All right. Wow. 57.6 seconds on the Dell. 90 seconds for that test on the Neo. Just wanted to show you what it looks like if you're looking at the performance of each core while the test is running.

Each one of these has six cores and they're all maximized. Dell is showing utilization almost 100% there. And we can see that from the chart. With the last XPS13 chip, it was Intel versus Snapdragon. That's Qualcomm's ARMbased chip. Intel took 80 seconds and Snapdragon 122. This time, we're doing even better. I ran it multiple times. Got an average of 65. Neo did a little bit better, too, on average. So, Intel did it again by 21 seconds. Apple, you got a problem. Going to have to put the A20 Pro in the new ones. But, of course, the XPS was plugged in for that. And I expected the Intel laptop to need full wall power in order to get full performance power. So I pulled both and I ran it again. After all, these are laptops. They're supposed to perform when they're away from the plug on your lap. Now on battery, the Neo took about 92 seconds, which is kind of uh within the margin of error of what we saw so far. So pretty much the same. But here was the surprise. The XPS took 67 seconds, also within the margin of error. So, it doesn't need the wall. A lot of you are going to be looking to do some AI stuff because that's what's hot right now. I don't know. I never heard of it. Have you? Llama CPP, a very popular AI inference engine. Runs on most things out there. There's their new logo. Pretty nice. What do you think?

It's open source, so you can grab it, clone it, build it, and that's what I did. The XPS took 3 minutes and 28 seconds, and the Mio M and the Neo took 4 minutes. Now, to be completely fair, the Mac build also compiles Apple's GPU backend, but still, I'm going to give this one to Intel. Now, I also wanted to really push these machines to see how far they can get. If you're working on these really hard, this is to simulate that. Basically, I ran the multi-core test five times back to back. And the MacBook Neo, remember, does not have a fan to cool it down between those runs or during the run. And throughout the test, the MacBook Neo went from 34.1 seconds, that's per run, down to 36.6 seconds by run five. So about 7% slower. With no fan, the chip probably slows itself down as it heats up as it should to protect it. It's called throttling. But the XPS did not slow down. It just kind of bounced around a bunch. 73 seconds, 71, 63, 67, 63. That's quite a bit slower than the Neo. About twice as long. Why the XPS loses this one, I don't entirely know.

All right, right now they're processing. All right, I got them working pretty hard in the background right now. Of course, this has no fan at all, so we don't hear anything, but tell the CPU load pretty high. All those cores are pegged. Now, here's Dell. That's the fan blowing. It's back here. You can definitely hear that fan. And this CPU is also pegged. Isn't that cute? MacBook Neo. There's the hot spot. What do we got? About 35, 36. Ooh, wait. I saw 38 in there. Oh, yeah. 38.5. What's over here on the Dell? Ooh, we got lots of them. 42, 36, 38, 40. Yeah, 42 is the highest I seen. The XPS did show that it was a little bit warmer than the MacBook Neo. While running that test on my actual workday, I have VS Code open, 15 tabs in Chrome, a terminal, Slack, couple of other apps like Notion and To-Doist. This is something I showed a couple years ago when I did the MacBook Air comparison on realworld tests on an 8 gigabyte machine and the Mac OS basically just handles it fine. The memory pressure is there but it's low.

It offloads everything that it doesn't need immediately to compressed. Now, right now I do have a couple things open and we can see that it's using just over 2 gigs of swap. Now the XPS's memory looks like we got 9.3 GB out of 26 committed. That's the page files here. I opened up a couple of tabs and memory popped up from 5 to 6 1/2 gigs. Popped open. Same two tabs on the Neo. Went up to about 6 gigs of used as well. They're about the same. Now, for local AI, I was limited by the amount of RAM here. You always want as much RAM as possible to do your models cuz the models have to fit inside the the RAM of the GPU. You can run it on the CPU, but that's not advisable. I talk a lot more about this in my other videos, so check out the rest of the videos on the channel. I go into more depth than this, but suffice to say for now to test these machines, you're not going to be running big models anyway. So I ran quen 34B at 4bit quantization. Quantization is basically when you take the full model at 16 bits and you compress it down to four bits to fit into the memory. Both use Llama CPP, which by the way has uh really improved their metal performance recently. So that's for Apple Silicon. You can also run Apple's MLX on Apple Silicon as well. But for Quen 3 4B, smaller model, not going to matter much. Reading the prompt, the XPS is actually faster. 156 tokens per second there versus 142 on the MacBook Neo. Reading the prompt is called prompt processing, by the way.

Sometimes it's referred to as PP. After prompt processing, the next stage is token generation, which is like writing the answer, the text, you get it printed out. The NEO wins this one. 17.9 tokens per second versus 15.7. You can think of a token as like 3/4 of a word. Essentially, both of these can write faster than you can read. Yeah, prove me wrong. Well, I guess I'll speak for myself. I can't read that fast. It's actually closer than I expected. Intel's graphics aren't a joke anymore here. They're integrated graphics and they're pretty good. Now, that was 4 billion parameters and 8 billion parameter model. On 8 gigs, ah, on the GPU, neither one can actually run that. At 4bit, the model is already five gigs and it needs working room on top of that. These GPUs just don't have their own memory. They have to borrow from the system. So that 8 gigs that we tal about, that 8 gigs of memory total, GPU can only use a part of that. And Llama CPP sees only 4.3 GB available for the GPU on the XPS and 5.3 on the Neo.

And neither is enough. So when you run an 8 billion parameter model, things spill over to the CPU and not just the GPU. they spill over to system RAM as well and CPUs are not great at doing this kind of parallel processing that machine learning models need to do. GPUs are good at that. So on the CPU, the Neo does 7.1 tokens per second. It's slow but still usable. The XPS 0.23 tokens per second. That's one token every 4 seconds. I can read that fast. My guess is it's just simply out of RAM and swapping to disk the whole time. battery. I unplugged both of them, fully charged, screens at half brightness, and then I ran the Python Compile Multicore and the 4B model twice. For that test, the Neo finished in 32 minutes, and the XPS 33. But while the Neo took only about 13% of its battery to do that task, the XPS took 21% of the battery.

That's about 60% more battery use for the same work. So, it used to be that the XPS was 100 bucks more than the Neo. Now, they're the same. Plus, you can get certain deals on these. But the XPS gives you a backlit keyboard, better screen storage, you can upgrade, faster Python compiles plugged in or not plugged in, which is a huge shift for Intel. I like it. Also, the XPS can be configured with 16 gigs. The Neo can't. It's kind of a tough decision because if you're in the Apple ecosystem, then the MacBook Neo is kind of a no-brainer. It just works seamlessly with the ecosystem. It's also really wellb built and the trackpad and the keyboard are both good. And Dell coming out with this device is a really welcome change as well. I'm happy that the XPS line is back online. Did I just say that? That's weird. Either way, I'm really happy that this is happening. This kind of competition at the lower end. Prices on all electronics are pretty much going up. And here are two entry-level laptops, and neither one skimps on quality. If you want to see my XPS video, the Intel XPS 13 versus the Qualcomm one, watch that video over here. And here's my MacBook Neo initial impressions and test video here. Thanks for watching and I'll see you next time.