Whoever came up with the dots in the graphs needs a raise, and I strongly encourage the team to find other ways to make the graphs more readable without having to pause the video to read. Off the top of my head, including brand logos or color coding the text could be an option, but I also think you could actually highlight whichever bar(s) the speaker is talking about
Having some kind of ltt water mark on I think would be beneficial for sure, like the labs should do. Not that they need to. But, definitely should have a live sort of highlight or different color bar as they talk about that specific part of the graph. That would be a solid improvement.
@jorgedaniel9656 theyve done pulsing yellow brackets in the past, which i think i prefer. Draws your eye to the data thats relevant rather than the name first
Riley is so funny. I don't even remember the ad he was talking about after the punch line. It's like those really old caveman Geico commercials. They were so good the actual thing they were selling got lost in the comedy.
Agreed - Might also be helpful to have little manufacturer logos on the names on the left axis labels or something - there's quite a lot of text there to keep track of which is which at a glance. Removing any redundant text for an icon might make it easier to take in at a glance when the charts are only displayed as briefly as they are.
@@halomika4973 Whilst that's obviously an option it's not a reason not to think about ways to improve the 'scanability' of the graphics so it's not necessary.
About the thumbnail: your sling is on backwards :-) The fabric hugging your body should go over the shoulder opposite your injured arm and the other side (the part closest to the camera) should go over the should on the injured side. This makes it easier to access the arm for any kind of treatment while the sling is on.
Thumbnail update on 8/8/24: it's now the laptop on the left with the workshop in the background and the text "Dumb but good" on the right. Good is on the line below "Dumb but" and is a yellow/golden colour
Has intel released new chips since this video aired or is the data from this still accurate and ryzen CPU’s are killing it in laptop space? Is the ai300 …is it different that ryzen 5/7/9 or ??? Or is this like a generation ahead of its other chips or ?
My old XPS13 had a Killer wifi chip in it, it was the only complaint out the whole machine which was otherwise pretty amazing. I cannot fathom why Dell would not stick with a decent Intel wifi chip.
@@rokmysoxbabykiller got bought out by intel in 2020 and uses intel cards with some branding now, they're pretty common in gaming laptops and have some silly features wireless 'gamers' might appreciate. Apart from that no different to normal intel cards. (eg. killer 1650i=ax201)
It makes me wonder why too. Even cheap phones have good WiFi connection. CHEAP PHONES! In fact nobody expects a bad WiFi chip on a phone. What's making this hard on laptops? They have a larger body to work on!
I purchased the Asus Vivobook S16 AMD back in January purely on recommendations from LTT and its an absolutely brilliant machine glad I took the plunge
Im really looking forward to replacing my slow ass x86 2 hour netbook styyle machine for an ARM based one that is way faster and lasts all day. Gonna give it another couple of generations before pulling the trigger.
You did not purchase this laptop back in January. This chip and new design of the 16 S Vivobook was just released with this processor the AI9 H375h AMD.....haha
@@seththomas3194 You know what your totally correct its impossible for anyone to own this specific 16S Vivobook because Linus is the sole owner of it. However I misscorrected myself of the back of of one of there review of the Asus Vivobook in general I purchased one back in January I stand corrected and apologise
Using dots for visualization for the graphs is genius. It makes it 10x more readable and makes me follow the content that much better. Keep it up! Whoever did that diserves a raise.
thank you guys for making videos but I wanna say its insanely annoying when the video is lower volume and then the ad cuts to a higher volume audio. please normalize the volumes
It’s nice that iGPUs are now starting to be serious considerations when buying a laptop It’s good for creator laptops, since they focus mainly on CPU bound tasks. With these iGPUs, they can also play games at relatively good settings and FPS without needing heavy cooling
This is basically all I care about now - I'm not a content creator so the artificial benchmarks are meaningless. In real-world functions for 99% people, we've not had performance issues for years with CPUs, so efficiency has become more of the king. The iGPU however is of great interest in its ability to render graphics for handhelds and thin-and-light laptops alike.
Upgradable Memory: According to TechPowerUp AMD has said that the Strix Point memory controller supports LPDDR5, LPDDR5x and DDR5. But even if that's true I doubt we'll see any SO-DIMM Strix Point because it's heavily reliant on memory bandwidth for the iGPU and LPDDR5/5x just deliver more of that than DDR5 thanks to higher frequencies being easier - but on the positive side the new LPCAMM2 standard means the manufacturers can provide user replacable LPDDR5x memory modules! Unfortunately it's probably going to be restricted to high-end second-wave devices given how new it is (and how many consumers doesn't know they need to be able to expand memory). It's not like we've seen many CAMM2 modules for desktop boards yet either, it's very early days for both.
I was somewhat surprised Linus didn't mention LPCAMM in the video. Really hoping we start seeing it in consumer devices soon, before everyone starts getting used to laptop memory never being replaceable, ever. (Framework, this is your cue!)
If you were around in the late 80's/early 90's, you'll remember when every fucking thing ever made had "TURBO" as part of the name. That's what "AI" is now.
Yeah but at least "Turbo" didn't forcibly opt me in to private data collection for their "AI learning". You know what, fuck AI in general, man. I have seen FAR more bad things being done with it than good things to warrant its existence. Why is AI even here except to fuck over the consumer and artist more while fucking over search results while providing so little of value except to shareholders? This is almost like that NFT horseshit all over again.
Even if the performance jump wasn't as big as some were expecting, seeing other reviews numbers it's true, the power efficiency improved greatly. It's really remarkable.
For the graphs, I love the little dot you had on the screen to help follow what the speaker is talking about. It might be nice the make the product being reviewed a different color so it's easy to quickly remember what you are comparing everything to.
@@aquiveal I mean if it's color coded then it's not that bad. they can color code the review chip with say green and all the others a different one. then highlight whichever ones linus is talking about at the moment with a different color. or even just highlight a square around all the data that is being talked about with a semi transparent color.
AI only matters to business data centers, for Joe user at home, AI circuitry is just a waste of power and chip area. I don't get why AMD and Intel is making a big deal to put AI in the names. The businesses that care about AI already would know if it's there regardless of the name.
Yeah I think they could do better there. The branding is stupid but the hardware appears to be legitimate. If they want to evaluate hardware performance (it's a hardware review, right?), they should benchmark the actual performance through pytorch, hf transformers, ollama, etc rather than just pass/fail on some consumer applications that either haven't launched yet, or not updated for the hardware. I mean, they criticized the software compatibility with Qualcomm too but still made an effort to find applications that did work to show a like for like comparison. If they want to benchmark tensor/neural performance they need to be comparing it against pure CPU and dedicated GPU performance.
@@spacepxl eh... they did just fine by me. I'm hoping they contiue to just name stuff "AI" while doing as little to actually advance the cancer as possible... I barely like voice assistants. The last thing I need is an app that will just give me misinformation whenever I ask a question.
I do a lot with AI. It's about knowing the use cases and the pros and cons. You don't read an editorial for factual information and you don't read an encyclopedia for opinions. I think stating that a model has an NPU is valid, but it doesn't need to be front and centre as it just isn't popular enough yet. Myself, I have an inference sever that does the heavy lifting and a laptop and phone that can access it. I'm unlikely to buy an NPU equipped laptop without the ability to upgrade the RAM, especially given the future of AI is having multiple small, specialised models all standing by and routing to the most appropriate.
The problem with HuggingFace is that it is mainly a repository, like Github, that runs demos in the cloud. It does not run locally. So AMD including them in their NPU marketing at all is cringe to 11.
7:00 those are terrible news. It doesn't matter how fast and how efficient is a HW, if you can't upgrade/replace RAM its life will be shorter that it could be
Apparently it's not true. If you look at the support page for the CPU it says it supports DDR5 ram. In the framework subreddit they're saying it can support LPCAMM so some clarification is needed
@@walkinmn they don't support SODIMM, which in and of itself is a problem. If they transition to LPCAMM2, that would be awesome. But why would they do that when they can just drop support entirely? -_-
The difference between the price of the memory made by Apple and the one made by Crucial is because Apple's memory is welded on the motherboard. And that increase the speed memory of the laptop... so don't just judge by the price because some sort of technologies can do a way better job on basic tasks or even hard tasks who need huge performance.
@@clemtech3008 everybody is "welding memory on the motherboard" what do you think Apple is the only one lmao gtfo, nothing can justify that price difference other than Apple Sheep giving them their money
@@clemtech3008 But all the different memory configurations have welded memory. Even if there are 2 chips welded onto the motherboard for the 16Gb version instead of 1, the price difference should not be 200$. RAM is super cheap these days, and even if they would hand solder it on the motherboard, the difference cannot possibly more than 50$. They are not hand soldering anything in their production however and the only reason why the 16Gb version is 200$ more expensive is that they can market it as being affordable while still extracting the most amount of money for the customers.
8:32 THIS. I always hate closing my laptop because it always takes so long to reconnect. You won’t believe how jaded 2 seconds will make you until you need to deal with it all the damn time.
@5:03 I see a Papillon! In all seriousness. Upgradability is a much bigger issue than I think most consumers realize. My work laptop is an HP 850 G5. It was getting real slow and glitchy recently. For $20 us, I double the ram to 16gb and now its like a whole new laptop. I have no need to upgrade. I have compared it to our other work laptops, such as the Lenovo T series and newer HP Elitebook's. Its at least on par in real world performance for our usage. I didn't have to get a whole new laptop, and my old one didn't end up in e-waste. It cost me $20 and 10 minutes to "refresh".
@@AntonioNoack No, most programs only save a few percent in memory consumption by running in 32 bit mode, not half like what you said. What is your source on this?
@@AntonioNoack not a lot, just like 20%. And js and other dynamic type languages have all types consume 64 bit regardless if it is running on 32 bit or 64 bit, for consistency
Honestly? Im a bit disappointed by not only the naming, but the removal of replacable ram. They could have just switched to CAMM2 modules to embrace the new standard, but decided no to?
Hey, Linus with blond hair looks not bad at all actually I would suggest considering ombré coloring to make ends pop up better, and maybe correct a tone for a little bit so it would look more in tune with darker roots when they'll show up eventually Hope it's not some meme that I don't getting 😅 Otherwise, nice! Experimenting with the looks is great, hopefully other comments would not be toxic about it, which I doubt a little 😬
re: Wi-Fi issues on AMD, I recently bought a Zepherus G14 with a 6800HS in it, crappy MediaTek chipset. I ended up just swapping it out for an Intel AX210 after a couple of days. AMD please just allow OEMs to use whatever Wi-Fi card they want to
I agree, if I get any laptop with a swappable WiFi card, and it's anything but Intel, it's getting swapped out immediately. MediaTek, RealTek, etc, they all just completely suck.
I have to try this. I have an AMD laptop for a year now with 6800H, and I am struggling with network performance. It is okay for the most part but the constant dropouts, reconnection issues and overall speed changes are driving me crazy.
Funny thing regarding the same cpu diffrent power part, here in sweden for a few years now the retailers have to list the max power for the gpus in laptops since they have the same problem!
Good on you guys for calling out the Wi-Fi chips. Anyone who struggled with the Marvell Wi-Fi in older Surface devices knows what a frustrating struggle it is when you device is packing substandard Wi-Fi support.
6:58 Strix should have ddr5 support and therefore support for SODIMM. It is stated on there website and then was stated again on the slides released on their tech day. On The Slides it is said 2 times that it supports both ddr5 and lpddr5. it even states the frequency of 5600 MT/s. Also ram can be upgradeble via lpcamm2 which is in 1 Lenovo laptop as far as im aware.
@@Juanguar soldered RAM is always LPDDR5, normal sized RAM cannot be soldered, it defeats the point of soldering it in the first place as its mostly done to save space inside the laptop mobos. Plus LPDDR5 or LPDDR5X is almost always faster than SO-DIMM DDR5, although I'll always take the SO-DIMM over the slightly faster MHz.
Well, because they don't want to fall behind.. just a quick reminder, the 400GB/s transfer speed in M3 Max laptops are equal with a 8 channel DDR5 (so 8 RAM moduls..) and a PCIe 8.0 x16 slot for GPU (the GPU communicating with the CPU on that 32GB/s limited PCIe 4.0 x16 port..) And even the 120GB/s in basic M4 and the 136GB/s in X Elite is about 3-4x faster like the regular DDR5 and PCIe 4.0 x16 can offer.. remember the SATA SSD also started with "just" about 4x faster like HDD.. and check where they are now, and how many HDD and SSD are in use now.. (even if it's more expensive, and more expensive to upgrade the SSD, the HDD still lost because it's slower..)
@@TamasKiss-yk4st "And even the 120GB/s in basic M4 and the 136GB/s in X Elite is about 3-4x faster like the regular DDR5". That is wrong. The Memory Bandwith of DDR5 5600 MT/s is 90 GB/s. That is nowhere near the 3-4x that you are stating. To add to that the official JEDEC Spezification of DDR5 goes up to 8800 MT/s. This is faster than any memory currently in an mobile device. The total bandwith of a processor only depends on the bus-width and the memory speed. The M3 uses 6400 MT/s ,the M4 7500 MT/s, the X Elite ~8400 MT/s, Lunar Lake uses 8533 MT/s and Strix up to 7500 MT/s. All of them have the same 128bit bus so the bandwith is only dependent on the memory speed. DDR5 can easily have the same speed as LPDDR5 the only reason it doesnt is because of the trace lenght and therfore resulting worse signal integrity and because intel and amd only rate there memory controllers to 5600 MT/s.
@@TamasKiss-yk4st The biggest speed benefit of SSDs over mechanical hard drives is their latency (on mechanical hard drives, this is seek time). This is multiple orders of magnitude faster on an SSD. Even if it's on a slow SATA SSD that can only read/write at something like 200MB/s, the SSD will just be way more responsive.
did a different editor handle the audio mixxing for the sponsored segment? the volume levels for linus in the beginning are a bit low, but then going to 1:13 is like a jumpscare.
I am loving these video's like always, but can I just say, this video has my most favorite background music I've ever heard on LTT. Honestly, the music has a very original Plants Vs Zombies vibe, and I am freaking loving it. Now, back to your regularly scheduled program, Qualcomm giving Intel and AMD a much needed bi*** slapping.
Basically this eliminates any argument for getting ARM Windows. The battery life difference exists but its not that huge for losing x86 backwards compatibility.
True, and don't forget the Intel answer still not arrived, which will use 3nm technology from TSMC. That also should be way more efficient, so maybe even the battery life advantage won't exists, but the compatibility problem still remain for Qualcomm.
@@TamasKiss-yk4st why are you hating on qualcomm? you're dead wrong here. The qualcomm chips still managed more battery life with a smaller battery. If the batteries were the same the qualcomm would have won with 3 extra hours or about 27% longer battery life, that's HUGE! The compatibility issues will be fixed as time goes on. But these new amd laptops are nothing impressive, unless you take a look at their ridiculous price tags $1,700 bucks for an old crapper that doesn't even perform as well as the macbook except for gaming.....in which case you could get a laptop with a real gpu....these amd laptops are boring trash for most people who don't specifically need these.
M1's had their downsides at launch too, especially for developers, but they figured it out. We're only in the first gen of these new Snapdragon chips, and as long as they sell enough to fund the next couple of gens, I think we'll be pleasantly surprised and be glad to have three fully viable options in the market.
9:40 Everyone, techtubers and consumers, should be asking laptop sellers to be including important specs like power limits on their product pages. Their omission is so obviously anti-consumer, and this has been too common for too long.
Slightly off topic here but, I have taken an issue with AMD iGPU sector. It's not that talked about anymore but they have progressively released better and better iGPU's but only in high end SKU's that are also then sold by manufacturers with discrete graphics. Why can't there be SKU's with Ryzen 5 or 7 CPU performance with best iGPU's? Most of the time a top SKU mobile APU won't even use the iGPU as it will switch to the discrete GPU for any work load. I am aware of sites like Minisforum, however it's never the latest APU (I live in the UK)
I just noticed the little pulsing circles on the charts. I've always had trouble following the charts and had to pause the video, but now I don't have to pause anymore.
Well, if it's still have bandwith limitation, it's not really work.. X Elite have 136GB/s, M3 Max laptops have 400GB/s, M4 already increased the bandwith with 20% (so M4 Max will get 480GB/s and that is still a laptop.. with 8-10 channel DDR5 memory speed, and PCIe 8.0 x16 speed.. where will you get 10 RAM slots in a laptop, and PCIe 8.0 slot? And didn't find any transfer limit for LPCAMM2, but based on Tom's hardware image, it's only about doubled in speed compared to average DDR5.. what is far away from the 480GB/s..) And compare the HDD vs SSD now.. they are only started with this 3-4x faster thing.. and even on a higher base price, on higher upgrade price, the SSD is more popular, and also way faster like the more transfer speed limited HDD..
@@TamasKiss-yk4st We are talking just about upgradability, but even then, for bandwidth LPCAMM is still better than SODIMM. If you try to get 500GB/s on SODIMMs you will need 10 slots (compared to just 4 LPCAMM).
Hey so uh, the Hugging Face support is actually a huge deal. More importantly a couple of icons to the right you can see that Ollama is supported. Don't overlook that - it means that this laptop can essentially self-host a GPT4o comparable experience through Meta's models right there on the machine, completely offline! My only criticism with it would be that they didn't include an installer to get people started with a local LLM in a single click. It's a huge deal for businesses who want the productivity of staff being able to use AI text generators to proof read emails etc without the liability of giving OpenAI access to business secrets.
Being a regular watcher of ltt + wan show I feel like I’m in a multiverse of Linus. One day his hair is dyed, the next it’s not, but currently it is, but previously filmed Linus isn’t. One day all these Linus are gonna make the spider man meme when they all meet.
We need soldered RAM to act like a L4 cache and SODIMM RAM to be like L5. I know it’s not always possible to get SODIMMs running the same speed, but we at least need something for expansion.
It's the same RAM chips whether soldered or separate, I think difference in speed and cost is not significant enough for the added complexity. If there was a scenario where fast RAM is crazy scarce and expensive (a second covid maybe), maybe there eventually pop up a solution where something limited like 4 GB is onboard and the next terabyte is pulled of from something like DDR3 adapters for the lack of anything better. L6 would probably be just SSD at that point. For historical reference, GTX 970 already tried using cheaper and slower VRAM as part of memory configuration, and did not succeed as intended. I had that card, and almost every game never used the last 0.5 GB that was sitting there.
@@talkingonthespectrum can you suggest what they should do instead of just complaining? what you said could range anywhere from valid criticism to some absolutely asinine complaints.
@@ghosthunter0950 a prerecorded 2-3 second piece of audio that's inserted into the video that says "investment disclosure, I have a financial stake in framework. Any mention of the company or their products are for illustrative reasons only unless otherwise implied" could even be 1.5x normal speed
Day 1 of asking LMG to name the screwdriver T-Shirt "Screwdriver 40% Polyester Shirt" (or just having better descriptions of what selling points are for each item, including pockets and such, I can never tell and this is why I haven't bought way more items)
3:35 My friend showed me what urban dictionary has as one definition of getting gapped and I just shook my head and laughed and now I can't unremember it lol. It involves one person missing at least one tooth and some horny dude just going to town using their grill as a special type of toothless glory hole
AMD has always been my goto product since the early 2000s. I have stuck with them no matter what even through their struggles. As someone that mainly codes and for gaming emulation is my thing, their continued focus on improving integrated graphics has saved me loads of money. They are also very power efficient which helps as i run my PC all day everyday