Just a quick FYI: The IR/Infineon TDA21472 isn't a "MOSFET"; it's a (70A) power stage, which, while it does contain MOSFETs, also has a driver stage and current/temperature sensors. As a rule of thumb, if you see two or more individual (3 to 8 pin) packages on each output phase, those'll typically be (high and low side) MOSFETs. If you see just one (14 to 32 pin) package, it's usually an integrated power stage. Not trying to nitpick semantics here; just trying to provide helpful info. Hope you take this in the spirit in which it was intended. Carry on the good work; love the channel.
Thermaltake hasn't figured this one out yet, from Steve's video on their holeless airflow technology. (Video link, skip to 11:50 to get your mind blown ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-vhkYcO1VxOk.html)
I actually did that once and it worked, but then when I actually needed rma I got screwed for 3x more expensive product... Even with 10 year warranty claim (which ironically just expired few days ago), I got really screwed and since then I was soo disappointed that I never bought anything from that brand again... Geeez, thanx G.skill :S
There were those that left behind some schmoo that said "VOID" on it back in the day. Those you kinda had to scrape/lift off with a PTFE(teflon) tool. Anyway, never understood why you couldn't attempt to fix something on your own, that just seems ridiculous.
I love hove you handle those electronics. Definitely helped me to be more comfortable around pc parts. So far nothing got damaged but then again I didn't toss them directly. :D
Electronics can take a surprising amount of abuse. Had an electronics professor who would smack motherboards against the table to fling the solder out after heating it up. I'm not saying you should do that, but it definitely made me less nervous around them :D
Yes, knowing that those parts are not that easy to break is great. In the end, the only parts that need attention are pins (on cpus or mobos) and sata connectors on hd/ssd.
Maybe I'm the only one here who gently pulled a GPU out of his case after thinking he unlocked the PCIE slot only to see after removing the GPU that the guts of my MOBO were hanging out and the PCIE slot itself remained attached to the GPU pins. Now to your point all I did was cut the guts to be flush with where they were hanging out and applied liquid electrical tape to the affected area and the PC works to this day, but alternatively I let my brother swap the PSU on my server rig and he fried the entire thing somehow.
he doesn't care he gets sent them. Plus he doesn't like amd he only ever handles amd cards rough. It's because he says AMD fans moan about it. OF COURSE THEY DO, you only ever do it to amd cards.
also I hadn't even gotten to 2:50 yet I was on 1:07 he is a asshole when it comes to taking care of the $1000's worth of amd products he gets given. I guess youtube fame went to his head. Actually come to think of it. Due to how he handles the card AMD should've just sent the card to a retailer so it could've been sold to someone who clearly wanted it. Cause he surely could care less about it. AMD should've left him to buy one himself since he surely would pay for it to "review" it. XD
@@Earthless_Rock a lot of places is still actually not many. A few games which are better optimised for things like Vulkan, maybe. It's competitive to the RTX3080 in rasterization, but is competitive with the 2080ti in Ray Tracing and has nothing like DLSS. However in pure rasterization performance then sure, it can in a few titles get up there with a RTX3090 and in a few work station apps but still most work station apps will prefer CUDA over OpenCL soooooo... yea.
"Apparently now that Nvidia talked about it people are aware that airflow exists, they invented it in fact" I like when he takes digs at either company's fanboys
I enjoy the irony of AMD moving away from a blower reference design, so Nvidia has now used a hybrid one. But I didn't even understand why someone could think there was a problem until Steve's sarcastic remark
@@hughjazz44 silicon is strong, the soldering is strong, the thermal paste can handle relatively rough bumps, the screws aren't cheap easily breakable alu and the fans will not dislocate unless you actively pull on it. that card is fine, it's like you flipping out because he threw a brown glass bottle against the wall. Yes it's glass, but it's those brown ones you find in medicine bottles and they can take quite the beating
Well this is super neat. Tear downs are one of the more interesting things for tech stuff. I don't have any practical use for this information but it's always cool to see how engineers make things.
Love these detailed teardown vids Steve. Really helps with the little things like how you detached the cable with tweezers. Installing my EK waterblock on my 6800 XT today I will definitely have this one playing in the background.
4 года назад
Didn't even watch the review yet. But for this i have always time. Keep up the good work tearing down things.
@ Whole screw drivers are generally better quality because you don't have bits that slip, but mostly, they're way faster to work with. We prefer high-quality, dedicated tools.
4 года назад
@@GamersNexus thank you for your answer, appreciate it.
GN really renewed trust in independent reviews. With no holding back due to fear of getting blanked by companies. Just nice to have a source to trustfully rely on for quality parts. And stuff to stay away from. Thanks. 💎
Honestly, if they had just kept their mouths shut and not had PR campaigns that were doing nothing saying how much Nvidia sucks (hint: this is what Intel is doing with their CPU's, AMD should realize this because they are on the other side of this with their CPU division), then maybe this wouldn't have seemed so underwhelming. It is a decent improvement over their 5000 series GPU's.
@@StormsparkPegasus They delivered what they promised in terms of performance nothing disappointing there. And calling a nearly x2 improvement over the last gen just from a year ago "decent" is a severe understatement. One year and getting nearly twice the performance is more than "decent". Compare to Nvidia that took 2 years to improve at most 45% on an improved node vs their previous gen, meanwhile AMD achieving their massive gain while using the same manufacturing node as their last gen.
@@kittyyuki1537 considering the cooler design, temps and driver instability for overclocking, the card may cook itself over time. Cooling VRMS via piece of metal connected to main cooler via 2 little thermal pads? With no direct airflow? I hope the person, who came up with this particular idea, will burn in hell, just like the card will. For me, reference design is a NO-NO. It´s only viable if you want to put an aftermarket air cooler or water cooling on it.
@@Morpheus-pt3wq To "cook" a card you have to run the components out of spec... So far the temps that were reported were all in spec so pls clarify what you mean
Great work! Your videos have taught me a lot about the technical side of hardware as I dive back into PC gaming. What I appreciate the most is your consistent methodology in each of your videos. I've watched some that are a few years old, and I can compare them with brand new reviews and teardowns of the 3000 and 6800 series.
i have removed countless of those stickers ( even the playstation ones that leave a pattern behind ) just by using some alcohol , soaking them, and then you can just pick them up with no damage to them, or ruining the pattern on the ps ones, and the adhesive will work like new after drying :D .. those stickers are useless anyways
@@CaveyMoth Been there and done that. Quite a boring affair when you do it yourself to be honest. First one I did was way back... Geforce 4 Ti 4200, before RU-vid even existed. The tiny fan on that quit working so I converted it to become a passively cooled one with the Zalman passive cooling solution, to which I also bought the extra fan just in case.
@@CaveyMoth I did tear down my KFA2 GTX 1070 Mini to remove all the dust and repaste it and that was a quite enjoyable process - but it only had 14 total screws (4 heatsink, 4 shroud, 3 per fan) and was super simple. The size also helped by making it easy to handle!
THANK YOU! I did it for my 6900 XT! I applied MX-4 and now my temps are 71 with Hot Spot 100, before I was 75 with 111! Took me about one hour to do it. 👍😁
I disassembled this card a few time. Pads on the die is 0.3mm, vrm 1.5 mm, memory chips 1.12mm. After several cards modded for paste, I ve found that 1.5mm on vrm, 1mm on memory and 0.5mm nylon washers are the best. There is no 0.5mm washers available anywhere, so i use my finger tip to sand down a M3x1mm washer on a 220 grids sanding paper. Perfect die contact everytime
I have to underclock, undervolt, and crank up the fans on mine to keep the temps under control. I’m too scared to try to replace the thermal pads or mod it to use thermal paste.
I absolutely LOVE the format for these videos. Steve is learning along with us on little treks of discovery everytime a new gpu releases. Wholesome af.
As someone that uses 'universal' waterblocks; these types of teardowns are essential, So thank you. (as an aside, this cooler is about the absolute best case scenario for this use-case)
6:40 Those stickers are not meant for you, or for any countries with great consumer protection laws. They're meant for us, third world countries with an out of touch and tech illiterate government where it's completely legal to void warranty under these cases as well as scalping straight from the retailer.
Doesn't mean that they don't work on people who don't know better in countries with consumer protection laws. It's always important to point out when they don't apply.
Thanks to you i modified my 5700xt a year ago. Huge improvement. I very like your reviews, because you dont say what pc parts to buy. Neutral, facts, reliable reviews.
it's called efficiency and optimization, something you may not understand or don't care. In doing so they only had to produce more screws, more plates and use more thermopads is equivalent to spend more time and money and make an old system on the fact of dissipation. the question is wasn't it better to do one piece? in fact it will be you to use a crap piece of product. because i goin to use a waterblock on that. I wonder when things can be done better why not?
@Aaron Moody But that is also going to make a practical difference... Though from a price/performance ratio it is probably not exactly a value proposition. My point was: These are reference design cards. The solution meets the required spec at MSRP. These are not targeting the OC crowd, so if the VRM cooling has the specified headroom (which is usually 30C from max temp when running at 20C ambient) it's perfectly fine. Doing more would cost more, but benefit very few of the total intended audience (granted, GN viewers are disproportionally swayed towards OC'ing the snot out of everything - but we're not "normal" in that regard).... If a reference card isn't noisy, flimsy, loud or hot it is a good card.. unless it's performance doesn't match it's price point.
I really wish the follow up piece on this had actually come out. Of all the GN content I've watched the one piece that doesn't materialize is the one for the card I actually got!
Yay, they did it! I thought they had failed, but in the end they did manage to fuck up their reference card's cooler. Congratulations AMD, I almost thought you had lost your essence in the reconstruction.
I think this explains how AMD's cards pull less wattage but are comparable or potentially even hotter than the Nvidia RTX 3000 series in most reviews I have seen... Nvidia clearly invested more thoughts into adequate cooling where AMD is still kinda like, "let the AIB partners sort it out". Then again saying this, perhaps it's a bit hypocritical when I have a watercooled CPU and GPU anyways.
@@WhenDoesTheVideoActuallyStart not normalized but I think we can say the GeForce RTX will be louder than the Radeons at stock clocks, at least as far as founders editions go.
I apologize if this has already been answered, but I'm very interested in what thermal pads you're using here Steve to add between the backplate and the card. I'd get me some of those, maybe even whip us up an affiliate link?? :) P.S. thanks for these awesome tear-downs, and thanks to Andrew and Keegan for some rad 4K camera work and editing. I'm a new subscriber and GN is one of my faves now.
What, AMD does not use the secret glue! I love these tear down videos. It means the consumer really gets to see what they getting for their hard earned money. Thank you GN, really appreciate it. Will you be doing a 6900XT tear down?
Jay2cent called you out and said he falls a sleep watching your videos. I think this channel is better, and i think you guys are better. Then again, i fell a sleep multiple times watching nice hair dude reading numbers from graphs.
A teardown with no dremel is always a good one. Thanks for making these! Do you really need to remove the entire cooler assembly to gain access to the fans? That's a weird and cumbersome solution!
For those of you who want to know the thermal pad thickness, I have looked everywhere to try and find an answer but I can’t. So I ordered a digital caliper and will tear my down and measure all the pads & replace them with 12w/Mk or higher and update the results be a couple days time I get everything done
It's not just unenforcable, it's actually illegal to put those void stickers on products, the FTC put out a letter scolding manufacturers for it not long ago.
In regards to your criticism of the baseplate design, is there any benefit from dumping all the mosfet heat into the vapour chamber if you can get away with keeping them within operating temps by using the baseplate as a giant heat spreader? That way the vapour chamber has more cooling capacity for the GPU and VRAM, which are connected directly to the vapour chamber. I assume the two thermal pads that connect the baseplate to the vapour chamber are there to assist the cooling of the mosfets on that side of the board because there is less metal in the baseplate on that side of the board. Or am I missing something in regards to the importance of mosfet cooling?
Steve: Forcibly jams fingers into fan deforming it Also Steve: "That will probably cause RMA issues later.." I have this card and I didn't feel comfortable putting enough pressure on the fan to make it make contact. I HIGHLY doubt that will ever be an issue unless its defective right out of the box.
@@SonicDestiny True I just got 1337 RTX 3090s and had no problem. Ambushing people with a baseball bat and yoinking their 3090s was cheap and easy. /sarcasm
When it comes to thermal compound, I definitely prefer longer lasting ones, especially for a videocard cooler, which I honestly never want to have to open up. I've done it before, but it can be a real pain, and it's not that hard to accidentally damage them. A lot of people end up damaging their cards by using the wrong thickness of thermal pads too.
I thought it was a little strange because you say you don't have a thermalpad on the blackplate but here I have a RX 6800 Sapphire reference model and the blackplate is so hot when the plate reaches 70º that you can fry an egg ... Can you tell me why????? I suspect my card has thermalpads
@@GamersNexus True that. People seem to think you have to handle them with the tips of your fingers, with gloves and an ESD strap.. in reality, in my 20 years building computers I've only lost a handful of parts to damage from handling, and they were all platter drives. I appreciate the more down-to-earth approach. Thanks for the review!
@@GamersNexus people get angry because these things are very expensive and (in this release cycle atleast) rare. You don't have to handle it carelessly, just because it can take it.
When the duel bios cards get released by aib's i wonder if you can flash the 6800xt bios on the 6800 boards ? Might unlock clockspeeds like the 5700 / 5700xt maybe even cores depending on if they are lasered off or are defective/ bios locked off
I suspect the shape of the die has to do with the Infinity Cache. AMD probably wanted to keep the latency down as much as possible which meant minimizing the physical distance from the CUs that need access to it to the cache itself. 128 MB of cache takes up a finite amount of area. If you put the cache around a square design, the "border" formed by the cache would be thicker. By making the aspect ratio of the GPU more rectangular it creates a larger perimeter, thus spreading out that cache more thinly around the border. A thinner border means quicker traversal to the furthest parts of the cache and therefore lower latency. AMD also stated that a larger cache would have incurred too much of a latency penalty. While I believe this, the Infinity Cache is already 20% of the die and even if 256 MB was effective from a speed/latency standpoint would it have been cost effective from a die area standpoint? Probably not. But just thinking about this design, that 20% of is just cache is really just incredible. AMD has a very lean architecture that can really rip through instructions when well fed. Ampere's die is not sacrificing 20% of it transistor count for cache, that's all additional space for computational logic. The fact that they trade blows in rasterization is mind boggling. On the downside, 20% of your die being cache means you can't spend it on better ray tracing cores, ML cores or more CUs in general. Can you imagine a "Biggest Navi" where all the space is CUs and it's using the fastest HBM it can get? If CDNA could game it would be a monster.
Hex nex , witch thicness of thermalpads did you useed for your Padmod? I saw you were allready stocking them to get contact with the "Heatsink". Would be happy to use K5 Pro from CS :)
Got lucky today and actually bought a 6800xt. Wanted to wait for a nitro or red devil. Hopefully can buy one soon then sell this reference model, my reference is sapphire branded.
That's a lot of SMD pots! On both sides! Well done AMD, you didn't cheap out at all. Too bad it's expensive AF still can't deside if I should pull the trigger on 6800XT or wait for something cheaper (not insterested into non XT at all)
