When the 3000 series came out, I was sure that every review of any of the cards was just sci-fi. the price and availability made me feel like TH-cam was a window to another universe
Thats what i was Thinking... Laptop GPUs are insane in that regards. My 4070 in my Acer Nitro (with bad Vram and only 8gb of it) out performs my 2080Ti sometimes
Electrical engineer here, power mostly gives less returns as you increase it. “More energy- optimized” chips can reduce power by 15% and slow down by like 5% (there are still many other factors that contribute to speedup). But power is an exponentially decaying trend as I said.
The 3080's amazing value proposition at launch being entirely the result of outside factors not quite lining up the way NVIDIA wanted makes a disturbing amount of sense.
well, "value" should be taken cautiously here. because no cards were available for a long time at no price let alone the MSRP ones. the EVGA one (not the Ti) is still being sold on amazon for 1365€ so not really a good value option yet.
I've actually used several of these "Frankenstein GPUs", mostly the 3060m but a few others. I'd say all in, theyre really good if you can get them for the right price. The main issue people have with these is the lack of driver support and the Arutar Frankendrivers (free version), are usually behind on updates. But with NVClean install, you can actually modify legitimate drivers yourself to have support for the Franken GPUs (with anticheat support). That way, theres really no downsides to these cards. Except lack of warranty support that is.
The AMD Frankenstein cards have full driver support, AMD doesn't block them so you can just use the standard AMD mobile drivers. I wouldn't use the Nvidia ones purely for this reason, unless you really know your way around driver code or/and know how to remove all the viruses etc.
@ For the NVIDIA ones. NVClean install doesn’t involve any coding. It tells you the ID of the GPU, and gives you the option to just add it to the driver.
@@Sam-ms8pt I didn't know about this. So is it, in effect, a way to add the blocked card to official drivers? If so then yeah, with the low power consumption of these cards they do become very viable like the AMD equivalents. Thanks for that mate.
I own a msi gp66 laptop with 150w 3080ti no complaints, other than it being a little toasty and some fan noise. Cooling probably a little better on bigger laptops.
A year following a catastrophic GPU shortages in 2022. These Frankenstein GPUs, which range in price from $230 to $320, are available in every online stores. The only difference is that they refurbish some of these components from mobile GPUs and discontinued Bitcoin mining.
that's not true, 150 watts is the maximum a 8pin should safely carry. I have been cryptomining for almost 8 years and it's extremely stupid to push more than that
20 series, 30 series, 40 series, Hardware Raytracing and the drive for High-refresh. I lived in a world where the deep lore was still being written. When real-time Software Ray-tracing was a neat trick that the Demoscene used to wow their fellow Nerds. We can't let creativity die. The Hardware is only as cool as the games and Demos that we can run on it. And people need to be able to afford these Modern Miracles too. Anyways, interesting vid. Thanks 🙂
Imagine a world where we got a version of this chip cut down by some weird amount. Like a 14GB or 7GB card with a 224-bit bus with ~6000 Cuda cores. If the 320-bit rumors are true, then you could also imagine a world where we got a 9GB 3070ti with a 288-bit bus. This would have marked one of the weirdest memory sizes since the 970 3.5GB "4GB."
The fact that changing the thermal interface material made an increase in performance even that big, shows that this core had a lot more performance to give, if it had the power delivery and memory bandwidth it might have made a better beast of a card.
Wow. This is a insane new step for the channel, especially since I didn't think you would cover a modded GPU. This is the type of stuff that makes me want to mod laptop GPUs into full sized cards despite my lack of knowledge on that subject lol. Thank you for making this content man, you make being a budget builder and tech appreciation more satisfying ❤
I've always been interested in how the GA103 fares, and it is great that you've made a video about it! Interestingly though, at the same number of active CUDA cores, the 4080 Laptop offers 30% higher clock speed at the same power (according to my testing) as an efficiency reference. Seems like 8nm is never meant to perform well at lower TDPs. What would be interesting is for there to be a shunt mod on this 3080Tim and see how well it fares as an actual desktop GPU in terms of power consumption. I've heard that it is not too hard, but it does take some soldering skills to get there. But, I am not so sure about how much extra power the mod will bring, as in the video the card is shown to run at sometimes below 100W, indicating that it looks like a voltage limit instead of power limit (good way to check is GPU-Z), and that we can't really solve voltage limits without extreme overclocking equipment that overrides the voltage.
Me too 😭, today at work a guy asked me how much faster a 3060 is than a 2060 super and I said 15-20%. He looked it up, I was right. Off the top of my head.
Sounds like a shunt mod is the best option. BIOS mods are often sketchy and bring drawbacks as driver incompatiblity. ... Ive tested that a lot. You also dont need to be afraid of the 8Pin power connector. If you have a decent PSU this connector is easly able to deliver twice the power. In the era of the 3090 I overclocked and overvolted the "cheaper" models with 2x8Pin to around 1kW powerdraw. The cables got warm, but not hot. We also dont talk about 500W in your case after all. You can find similar info in basically any XOC Forum, where people are not so "afraid". Anyway. Most likely the voltage is also capped. So buying an elmorEVC to hook up to the voltage regulator combined with the shunt mod, will unleash the full potential of this gpu core. The risk of bricking something fairly low. If you need any help or assistance. Let me know. I can guide you thru.
It is my belief that the true value 80 series cards are the one's that share the same die as the top tier cards, whether that be the 80ti or 90. The previous card to have this configuration before the rtx 3080 was the gtx 780, which shared the same gk110/b die.
I cant say youre right though. Ironically, the better measurement is whether the next nvidia x60 class of card can match the last gen x80/70 Ti cards If, the desktop 5060 can match 80-90% of the desktop 4080. The 5060 is worth it to buy, much like the 1060,2060 and 3060 The 4060 struggles to beat the 3060 Ti even without DLSS. What a joke
@@3dcomrade Yeah, in the past at launch a new 60 card would perform just shy of the previous gen 80 card, then through driver optimisation creep just ahead of it after a few years. The 2060, for example, launched with identical rasterised performance as a 1080, the 1060 launched just a little slower than a 980 (though in reality ran games at higher FPS, 1060s doubled the FPS of 960s, for context, with just 30% more on paper throughput as a GTX 770). This was ended with the RTX3060, while it did launch with performance in between the 2070 and 2080 instead of being 5% slower than the 2080 it was just 3% faster than the 2070, and unlike Turin cards that all have underclocked memory (by 1000MHz) you can't OC them like you can RTX 20 cards. When I OC my 2060Super, at 8GB workloads it smokes a 3060. I fully OC'd and ran an effective non-stop stress test for one month, as long as your Turin card has second gen GDDR6 or newer it's an OC beast, if it's on first gen, and especially first gen Micron chips then do not OC it, in fact give it a bit of an undervolt, 1st gen Micron DDR6 cooks itself even when it's doing nothing.
@@potatoes5829 price will almost definitely continue to rise, mainly due to Trump getting into office in the states and the heavy tarrifs he plans to put toward on Chinese imports Nvidia being an American based company will feel this
Little Tim here sipping power while pushing quite respectable FPS. Too bad you cannot really buy it from anywhere. I would have wanted one when I had to last upgrade my setup.
I've seen these models on eBay (UK) for around £330, I was actually really wanting a video on this GPU as it looked like a great deal! It's a shame they've locked the TDP and OCing as it could arguably go quite far!
@EtaCarinaeSC I'm not sure if that's worth it though for most consumers, that's like what. 2080 ti level performance? Obviously, you lose like 5 gigabytes of VRAM, but you save about a £100 easy! The only way I'd be interested in buying this if it was actually only around £350 with performance that was 5-10 FPS below that of an actual 3080 ti.
If you’re taking a look at some of Nvidia’s lost gpus that are also fairly modern, I dare you to take a look at the gt1010. This powerhouse of a card is a further cutdown 1030 that was released sometime in 2021. I’m sure was made for office prebuilds, like the optiplex and whatnot, unfortunately, I don’t think any systems actually released with it. However, there’s listings on eBay of a Lenovo gddr5 version, so perhaps some unlucky souls actually got it with their systems. Anywho, there are actually 2 versions of the card, the ddr4 and gddr5. Gotta be honest, I got it for grins and giggles after Budget-Builds official did a review on it, and let me just say, it’s the first time I’ve seen a gpu struggle to play deep rock galactic.
It's probably due to the card's conservative power profile. It stays 20-30 Watts below the cards normal TDP, and it leaves out 20-30 Watts of potential boost (170W).
@ not that it performs any better at 175w diminishing returns add up fast. I get about a 15% difference in fps going from 115w to 150w and about 3% from 150w to 175w.
RTX 3000 series = cracking PCB-s, imploding VRM-s, thermal holes in the gpu vapor chambers, LAVA hot GDDR6 chips, DOUBLED the actual CUDA core numbering (compared to how the FP32 throughput was quazi 2 times faster) (one of the sources is Hardware Unboxed RTX 3080 Launch Review), already nuclear amount of PWR consumptions from the RTX 3080 10GB and above. No one cares about the RTX cards in general because they are almost all either crap because of poor performance, crap because of nuclear PWR consumption or crap because of 12-pin PWR melting. RTX 2000, RTX 3000 and RTX 4000 all have quite huge hardware issues and hardware flaws.
It's in my CPU benchmark roster, due to how much of a bottleneck they can be once a lot of tyranids appear on screen. If I feature it in GPU reviews, it will have to be with the caveat that you NEED a decent CPU to go with it!
@IcebergTech understandable. I was messing with graphics settings in it last night and didn't notice much difference between them. there does appear to be a pretty large cpu/optimization bottleneck
I think now there are 2080 Ti 22GB 3080 20GB and 4090(D and non-D) 48GB for "AI Shenanigans", also heard of existence of 3070 Ti 16GB and 4080 Super 32GB, but haven't seen it in the fringes of the market yet
Saw a few listings on Xiaoyu (more eBay-esque side with more 2-hand/less-formal stuffs than like AliExpress/Taobao), but they only accepting lot orders starting from 50+ units, at the price point that is not really competitive against RTX A6000 (Ampere) For "AI Shenanigans", VRAM bandwidth is the only thing needed to pay attention to
Ptm7950 isn’t just a pad, you can buy a tube of it. I got one on my desk. When I sent my 6900xt to northwestgpu for repair he also didn’t know it was available as a paste. It’s the best sh*t ever no cap. Only thing better is liquid metal and for obvious reasons I don’t care for it, rip my hot pink rx550 4gb super oc dreams, limited by a mole. Power connector when u needed a sata!
The Laptop 3080 is close to a 3060 desktop while the Laptop 3080 Ti is around a desktop 3060 Ti in terms of performance. I have seen a few Gaming Laptops that had 3080 Tis in them. So that makes me think that Nvidia had quite a few full GA103 dies but not enough to warrant making desktop GPUs out of them. Well look at it this way your advice could help owners of Gaming Laptops that have 3080 Tis in them.
my 3070tiM notebook is 12% away from desktop 3070, i dunno what you are talking about. 3060 notebook can actually be faster than 3060 desktop as it has 9% more cuda cores.
then the 4090 laptop has half the performance of a desktop and is getting smoked by 7 series gpu..... there is a reason people dont buy anything higher then 6/7 series is lapops and it has to do with the premium you are paying and the insane loss of power. So 60 series has alsways been a nice spot in ot works and is same as desktop and cost around 900-1300€ new. Thats what most people are willing to spend on a system like that. Anything higher you just build a full setup cause price to performance does not match at all.
@@nsreturn1365 NO. The speetspot in laptops in 3000 and 4000 series is upper-mid range. 3060M only has 6GB of vram, whereas 3070m ti has desktop 3070 specs and therefore has close performance in most laptops. 4070m is a scam and has low end GPU specs, whereas 4080m has specs between desktop 4070 and 4070 ti, so it performs like a desktop 4070, which is 50% faster than 4070m. Even at the same 100w power limit, 4080m is about 20-30% faster than 4070m simply due to having much more cores which makes it more energy efficient.
@@nsreturn1365 "anything higher you just build a full setup" ah yes, I love carrying my desktops with me. Since laptops are bought for portability, this is what I assume you're suggesting to do with a desktop one gets instead of a laptop. P.S. also you're suggesting buying laptops with least energy-efficient GPUs. Great advice, man. 4080m needs just 65 watts to perform like a full power 4070m. Yeah why would anyone want efficiency in a device that is limited on cooling am I right?
this is actually in no way an actual representation of the performance of a RTX 3080 Ti mobile, i have a Strix Scar 17 with a 12950hx and 3080 Ti laptop i get much higher FPS in a lot of the games you tested maybe the low power usage on yours is seriously gimping it but mine usually runs between 160-175w depending on how demanding the game is on the CPU (this laptop prioritizes CPU power in certain titles which lowers the TDP of the 3080 Ti slightly)
@@Rotwold hmm yes it does have a vapor chamber but overall i think the cooling is quiet comparable because the fans on the graphics card you have are much bigger fans in a less restrictive environment (a pc case vs a limited intake laptop chassis) you also get much better temps than i do even though mine uses a vapor chamber and liquid metal but I'm pretty sure the massive performance issue is because of the power limit as the official 3080 ti laptops hover in the 150w and 175w range i just didnt like that your graphics card was referenced as an actual 3080 ti laptop/mobile variant when its actually even more gimped than a laptop 3080 ti
Oef you are putting it up against the 102 3080... those come with a Core Count of 10240 TMUs 320 ROPs 112, the 2 extra MB it has, means it has a big effect in more demanding games. You will see sh*tty 1% and 0.1% lows. It lacks the frequency curve to prevent that from dropping. If you had the overclock capability to push this around 1800 to 1900 you could probably do some real damage to the 3070 line up.
I was wondering whether you have considered doing reviews on the 6750 gre’s? They seem like a good deal but I haven’t seen any real coverage from most of the tech reviewers I follow😊
I've got one! Unfortunately it's not going to get a standalone review until sometime in the new year, but it will be in a "console equivalent PC" video in a couple of weeks' time.
This is China refurbished cards with mobile GPUs consume more than what the sensors show. Their consuming is fake :) Check through the outlet ampere meter. First time i saw it on rtx4000m cards and then checked rtx3000m.
GA103 is 256bits wide on the memory bus, it's not meant to be a 320bits at all, how are you supposed to fit those nearly 100 more connection to the core while both GA104 and GA103 share nearly the same amount of balls. You shouldn't base your informations on some "leak"/rumors, those are false, I bet the person that shared the info about GA103 being a native 320bits extrapolated this from 3080 rumors and wrongly said it was for GA103.
Shunt mod it so it can use full power and then retest it. It's annoying (and bad journalism) that people don't test GPUs with the power limits removed.
When the 3000 series came out, I was sure that every review of any of the cards was just sci-fi. the price and availability made me feel like TH-cam was a window to another universe
everyone I know with a 3000 bought it after the 4000 series came out
Months after the 3000 series came out I still had to pay 1100 dollars for a 3070 💀I think that’s my biggest regret
@wenimechoinasama2 I got mine and the prices dropped two months later.
@wenimechoinasama2 I'll buy it for 350 lol
@@isaacibbotson4632 I would but I rely on my pc too much lol
the efficiency of this modded gpu is actually insane though. The regular 3080 ti uses more than 3x the power for 50% more performance
Thats what i was Thinking... Laptop GPUs are insane in that regards. My 4070 in my Acer Nitro (with bad Vram and only 8gb of it) out performs my 2080Ti sometimes
To be fair, undervolting 3000 series cards is pretty viable and effective and can drop the wattage significantly.
Electrical engineer here, power mostly gives less returns as you increase it. “More energy- optimized” chips can reduce power by 15% and slow down by like 5% (there are still many other factors that contribute to speedup). But power is an exponentially decaying trend as I said.
@@JJ0203 I don't measure FPS, I only check how many Watts I'm pulling from the wall. If the game feels laggy I turn up the voltage FOR MOAR POWER!!!
The 3080's amazing value proposition at launch being entirely the result of outside factors not quite lining up the way NVIDIA wanted makes a disturbing amount of sense.
Careful... Wouldn't want to alert the horde :P
Yeah. Unfortunatelly we will never see again a 102 die in a non 90 class card.
well, "value" should be taken cautiously here. because no cards were available for a long time at no price let alone the MSRP ones. the EVGA one (not the Ti) is still being sold on amazon for 1365€ so not really a good value option yet.
@@inkredebilchina9699 that's mostly due to mining.
@@inkredebilchina9699 Yep, the one promising post-RTX generation for value never delivered on it thanks to crypto.
Just be glad we never got an RTX 3030!
Um 3050 6gb?
I mean if it was less than $100 🤔
Yeah a 6gb ddr 64 bit gpu 😆
So far......... 😂
RTX3030 4GB, GTX1060 3GB performance. Wouldn't be terrible idea...for the right price.
I've actually used several of these "Frankenstein GPUs", mostly the 3060m but a few others. I'd say all in, theyre really good if you can get them for the right price. The main issue people have with these is the lack of driver support and the Arutar Frankendrivers (free version), are usually behind on updates. But with NVClean install, you can actually modify legitimate drivers yourself to have support for the Franken GPUs (with anticheat support). That way, theres really no downsides to these cards. Except lack of warranty support that is.
The AMD Frankenstein cards have full driver support, AMD doesn't block them so you can just use the standard AMD mobile drivers. I wouldn't use the Nvidia ones purely for this reason, unless you really know your way around driver code or/and know how to remove all the viruses etc.
@ For the NVIDIA ones. NVClean install doesn’t involve any coding.
It tells you the ID of the GPU, and gives you the option to just add it to the driver.
@@Sam-ms8pt
I didn't know about this. So is it, in effect, a way to add the blocked card to official drivers? If so then yeah, with the low power consumption of these cards they do become very viable like the AMD equivalents. Thanks for that mate.
"Nvidia's Lost GPU Line" more like with the lack of availability
Back then.....miners and serial scalpers rejoiced!
The 3080ti laptop is more like always nvidia´s 3070 desktop. So you get a rare 3070 with 16 gigs of ram whats effectifly a Quadro Card.
no, 3070 desktop is same chip as laptop 3070ti, 3070ti desktop is laptop 3080 , all of them are capable of upgrading to 16gb
I own a msi gp66 laptop with 150w 3080ti no complaints, other than it being a little toasty and some fan noise. Cooling probably a little better on bigger laptops.
@@alanliang9538 I think they were talking about specs, not performance. 3080m ti has about the same performance as desktop 3070 ti.
A year following a catastrophic GPU shortages in 2022. These Frankenstein GPUs, which range in price from $230 to $320, are available in every online stores. The only difference is that they refurbish some of these components from mobile GPUs and discontinued Bitcoin mining.
The theoretical limit of a good 8pin is 310W. You should shunt mod this card to see what it would have been capable of with a higher power limit.
Not good idea on cheap VRM ;/ Also tis heat-sink probably can only handle 150W
that's not true, 150 watts is the maximum a 8pin should safely carry. I have been cryptomining for almost 8 years and it's extremely stupid to push more than that
Where the actual fuck did you get 310W
@@excess.subiefl0w it can take way more than 150W. 150 is limit for crappy casual cables.
💀
12:54 It was pretty good French actually.
as a french native yes
20 series, 30 series, 40 series, Hardware Raytracing and the drive for High-refresh. I lived in a world where the deep lore was still being written. When real-time Software Ray-tracing was a neat trick that the Demoscene used to wow their fellow Nerds. We can't let creativity die. The Hardware is only as cool as the games and Demos that we can run on it. And people need to be able to afford these Modern Miracles too. Anyways, interesting vid. Thanks 🙂
Imagine a world where we got a version of this chip cut down by some weird amount. Like a 14GB or 7GB card with a 224-bit bus with ~6000 Cuda cores. If the 320-bit rumors are true, then you could also imagine a world where we got a 9GB 3070ti with a 288-bit bus. This would have marked one of the weirdest memory sizes since the 970 3.5GB "4GB."
The fact that changing the thermal interface material made an increase in performance even that big, shows that this core had a lot more performance to give, if it had the power delivery and memory bandwidth it might have made a better beast of a card.
love these mobile GPU turned reg GPU videos
and also interesting story behind AD103!
Wow. This is a insane new step for the channel, especially since I didn't think you would cover a modded GPU. This is the type of stuff that makes me want to mod laptop GPUs into full sized cards despite my lack of knowledge on that subject lol. Thank you for making this content man, you make being a budget builder and tech appreciation more satisfying ❤
You can buy such cards from China
Thanks for sharing a piece of history with us Iceberg, this was a pretty interesting thing to learn on a Friday morning lol
I've always been interested in how the GA103 fares, and it is great that you've made a video about it!
Interestingly though, at the same number of active CUDA cores, the 4080 Laptop offers 30% higher clock speed at the same power (according to my testing) as an efficiency reference. Seems like 8nm is never meant to perform well at lower TDPs.
What would be interesting is for there to be a shunt mod on this 3080Tim and see how well it fares as an actual desktop GPU in terms of power consumption. I've heard that it is not too hard, but it does take some soldering skills to get there.
But, I am not so sure about how much extra power the mod will bring, as in the video the card is shown to run at sometimes below 100W, indicating that it looks like a voltage limit instead of power limit (good way to check is GPU-Z), and that we can't really solve voltage limits without extreme overclocking equipment that overrides the voltage.
I feel attacked, i am one of those with the encyclopedic knowledge in that galaxy brain of mine
Me too 😭, today at work a guy asked me how much faster a 3060 is than a 2060 super and I said 15-20%. He looked it up, I was right. Off the top of my head.
@@phantoslayer9332 I was asked which die 980ti used and i blurted out GM200 without thinking...
Yeah.
@@DLTX1007 yep, that exactly.
New video. Let's goo boyss
"The usual TH-cam things"... I felt so inclined.
Sounds like a shunt mod is the best option. BIOS mods are often sketchy and bring drawbacks as driver incompatiblity. ... Ive tested that a lot.
You also dont need to be afraid of the 8Pin power connector. If you have a decent PSU this connector is easly able to deliver twice the power.
In the era of the 3090 I overclocked and overvolted the "cheaper" models with 2x8Pin to around 1kW powerdraw.
The cables got warm, but not hot. We also dont talk about 500W in your case after all. You can find similar info in basically any XOC Forum, where
people are not so "afraid".
Anyway. Most likely the voltage is also capped. So buying an elmorEVC to hook up to the voltage regulator combined with the shunt mod,
will unleash the full potential of this gpu core.
The risk of bricking something fairly low. If you need any help or assistance. Let me know. I can guide you thru.
The intro to Iceberg videos with that music just hits so hard
People finds loads of cards in warehouses and i still have to use a 980 ti :( But really cool video and congrats for getting such a nice card.
It is my belief that the true value 80 series cards are the one's that share the same die as the top tier cards, whether that be the 80ti or 90. The previous card to have this configuration before the rtx 3080 was the gtx 780, which shared the same gk110/b die.
I cant say youre right though. Ironically, the better measurement is whether the next nvidia x60 class of card can match the last gen x80/70 Ti cards
If, the desktop 5060 can match 80-90% of the desktop 4080. The 5060 is worth it to buy, much like the 1060,2060 and 3060
The 4060 struggles to beat the 3060 Ti even without DLSS. What a joke
@@3dcomrade the 5060 would also needs to be not anymore expensive
@@3dcomrade
Yeah, in the past at launch a new 60 card would perform just shy of the previous gen 80 card, then through driver optimisation creep just ahead of it after a few years. The 2060, for example, launched with identical rasterised performance as a 1080, the 1060 launched just a little slower than a 980 (though in reality ran games at higher FPS, 1060s doubled the FPS of 960s, for context, with just 30% more on paper throughput as a GTX 770). This was ended with the RTX3060, while it did launch with performance in between the 2070 and 2080 instead of being 5% slower than the 2080 it was just 3% faster than the 2070, and unlike Turin cards that all have underclocked memory (by 1000MHz) you can't OC them like you can RTX 20 cards. When I OC my 2060Super, at 8GB workloads it smokes a 3060. I fully OC'd and ran an effective non-stop stress test for one month, as long as your Turin card has second gen GDDR6 or newer it's an OC beast, if it's on first gen, and especially first gen Micron chips then do not OC it, in fact give it a bit of an undervolt, 1st gen Micron DDR6 cooks itself even when it's doing nothing.
@@potatoes5829 price will almost definitely continue to rise, mainly due to Trump getting into office in the states and the heavy tarrifs he plans to put toward on Chinese imports
Nvidia being an American based company will feel this
@@3dcomradethat joke was created to keep AMD alive.
5:46 AD103?
Title got me jumpscared thinking Bergman somehow got a hold of a 4010
bad bot
@@Zoey_KL whatever u say bro
I want to see a 4050M on a desktop PCB. It would be a worthy upgrade to the 3050 6GB as a new slot-powered card.
I love to see this kind of videos, it reminds me that not always I know everything there is about GPU chips
Little Tim here sipping power while pushing quite respectable FPS. Too bad you cannot really buy it from anywhere. I would have wanted one when I had to last upgrade my setup.
The 528.xx driver fron Arutar allows the power slider to be adjusted. Might be worth trying and a followup Short.
Timmy is more like a 3060
i've never seen 3060 doing 66% of 3080ti performance at 135w... I want to get my hands on such 3060s...
I've seen these models on eBay (UK) for around £330, I was actually really wanting a video on this GPU as it looked like a great deal!
It's a shame they've locked the TDP and OCing as it could arguably go quite far!
shunt modding is easy tho. I did it and it runs at 200w, its insane, gets around 83% of 3080ti performance without the 12gb vram issues
@EtaCarinaeSC I'm not sure if that's worth it though for most consumers, that's like what. 2080 ti level performance? Obviously, you lose like 5 gigabytes of VRAM, but you save about a £100 easy!
The only way I'd be interested in buying this if it was actually only around £350 with performance that was 5-10 FPS below that of an actual 3080 ti.
@@benchmarkerlovesgaming140 in some games it performs better, where the vram maters. Or when you want to use RTX.
Your french is perfect ! Keep it up !
Would be curious to see the cards power limited to see how the dies scale
I knew not sleeping yet had a use today
Cool video! This is indeed the wetdream of anyone remotely interested in joining any card manufacturer!
Great video and great find!
There should be an older frakendriver with the power limit controls unlocked
3080 *Tie* never gets old
Somehow still more concise than Ryzen MAX AI HX 395
If you’re taking a look at some of Nvidia’s lost gpus that are also fairly modern, I dare you to take a look at the gt1010. This powerhouse of a card is a further cutdown 1030 that was released sometime in 2021. I’m sure was made for office prebuilds, like the optiplex and whatnot, unfortunately, I don’t think any systems actually released with it. However, there’s listings on eBay of a Lenovo gddr5 version, so perhaps some unlucky souls actually got it with their systems. Anywho, there are actually 2 versions of the card, the ddr4 and gddr5. Gotta be honest, I got it for grins and giggles after Budget-Builds official did a review on it, and let me just say, it’s the first time I’ve seen a gpu struggle to play deep rock galactic.
As a French you have done it quite well with your accent well done , Bravo a toi 😉
Yep, not that far from a native speaker actually.
God bless this guy mentioning the holy grail
Didnt realize that is what is in my laptop and performs so great!
probably performs even better, because in this video the power limit was lower than on most laptop 3080 ti s
It would be interesting to see it moded.😊
See if Der Bauer could do some voltage mods for you.
When are you gonna make a video about CRT Monitors!!!! your channel is so unique and classy, its gotta happen some day :O
RTX A5500 7424 / 232 / 96 config I'm guessing.
Send little timmy to buildzoid for some actually hardcore overclocking treatment
Is the song from "Dance With The Dead"?
Jensen Huang would like to: *KNOW YOUR LOCATION*
What about compare 3080s by massively undervolt cards in order to match the power limit of Tim?
Just so you know, you pronounced '69' and '66' really well!
Bonus points for the Monty Python reference. 😂
You don't prononce the "t" in malheureusement
Very interesting piece of kit.
Wonder how well that PTM7950 would work on a Radeon HD 7950. :D
being in the red zone means you're at 80% full on VRAM. You could have maxed it out.
great video! thank you :)
Have the same GPU in my x17 R2. Runs at 135-150w and up to 170w~ with max fans. That GPU can do better not sure why its being strangled.
It's probably due to the card's conservative power profile. It stays 20-30 Watts below the cards normal TDP, and it leaves out 20-30 Watts of potential boost (170W).
@ not that it performs any better at 175w diminishing returns add up fast. I get about a 15% difference in fps going from 115w to 150w and about 3% from 150w to 175w.
I hope they find it
RTX 3000 series = cracking PCB-s, imploding VRM-s, thermal holes in the gpu vapor chambers, LAVA hot GDDR6 chips, DOUBLED the actual CUDA core numbering (compared to how the FP32 throughput was quazi 2 times faster) (one of the sources is Hardware Unboxed RTX 3080 Launch Review), already nuclear amount of PWR consumptions from the RTX 3080 10GB and above.
No one cares about the RTX cards in general because they are almost all either crap because of poor performance, crap because of nuclear PWR consumption or crap because of 12-pin PWR melting.
RTX 2000, RTX 3000 and RTX 4000 all have quite huge hardware issues and hardware flaws.
Why didn't you try overclocking it?
would you consider using space marine 2 in one of your benchmarks?
It's in my CPU benchmark roster, due to how much of a bottleneck they can be once a lot of tyranids appear on screen. If I feature it in GPU reviews, it will have to be with the caveat that you NEED a decent CPU to go with it!
@IcebergTech understandable. I was messing with graphics settings in it last night and didn't notice much difference between them. there does appear to be a pretty large cpu/optimization bottleneck
What's the song you're using in your intro?
Unicorn Heads "808 Doorbell Chime"
th-cam.com/video/x7gNHCsD-XY/w-d-xo.htmlsi=rq9A2-PfabX7zByg
I remember I had a 1060 with a 1080 die (with gp104 instead of gp106) I wonder if there is any difference...
so that would be like a 3075 Ti?
Thanks! It was interesting. 🙃
I think now there are 2080 Ti 22GB 3080 20GB and 4090(D and non-D) 48GB for "AI Shenanigans", also heard of existence of 3070 Ti 16GB and 4080 Super 32GB, but haven't seen it in the fringes of the market yet
I heard about those too, it's crazy
Saw a few listings on Xiaoyu (more eBay-esque side with more 2-hand/less-formal stuffs than like AliExpress/Taobao), but they only accepting lot orders starting from 50+ units, at the price point that is not really competitive against RTX A6000 (Ampere)
For "AI Shenanigans", VRAM bandwidth is the only thing needed to pay attention to
just try shunt mod with more power u can still deliver over 200W vis that 8pin
Ptm7950 isn’t just a pad, you can buy a tube of it. I got one on my desk. When I sent my 6900xt to northwestgpu for repair he also didn’t know it was available as a paste. It’s the best sh*t ever no cap. Only thing better is liquid metal and for obvious reasons I don’t care for it, rip my hot pink rx550 4gb super oc dreams, limited by a mole. Power connector when u needed a sata!
Not bad for a lower TDP card 😅
such a unique video ,tho the t is silent in malheureusement 😂
The Laptop 3080 is close to a 3060 desktop while the Laptop 3080 Ti is around a desktop 3060 Ti in terms of performance. I have seen a few Gaming Laptops that had 3080 Tis in them. So that makes me think that Nvidia had quite a few full GA103 dies but not enough to warrant making desktop GPUs out of them. Well look at it this way your advice could help owners of Gaming Laptops that have 3080 Tis in them.
my 3070tiM notebook is 12% away from desktop 3070, i dunno what you are talking about. 3060 notebook can actually be faster than 3060 desktop as it has 9% more cuda cores.
then the 4090 laptop has half the performance of a desktop and is getting smoked by 7 series gpu..... there is a reason people dont buy anything higher then 6/7 series is lapops and it has to do with the premium you are paying and the insane loss of power. So 60 series has alsways been a nice spot in ot works and is same as desktop and cost around 900-1300€ new. Thats what most people are willing to spend on a system like that. Anything higher you just build a full setup cause price to performance does not match at all.
3080 laptop has pretty much desktop 3070 performance, it's not even close to 3060, not sure where you got your numbers from.
@@nsreturn1365 NO. The speetspot in laptops in 3000 and 4000 series is upper-mid range. 3060M only has 6GB of vram, whereas 3070m ti has desktop 3070 specs and therefore has close performance in most laptops.
4070m is a scam and has low end GPU specs, whereas 4080m has specs between desktop 4070 and 4070 ti, so it performs like a desktop 4070, which is 50% faster than 4070m. Even at the same 100w power limit, 4080m is about 20-30% faster than 4070m simply due to having much more cores which makes it more energy efficient.
@@nsreturn1365 "anything higher you just build a full setup" ah yes, I love carrying my desktops with me. Since laptops are bought for portability, this is what I assume you're suggesting to do with a desktop one gets instead of a laptop.
P.S. also you're suggesting buying laptops with least energy-efficient GPUs. Great advice, man. 4080m needs just 65 watts to perform like a full power 4070m. Yeah why would anyone want efficiency in a device that is limited on cooling am I right?
so this card is the dream´s owners of 3070/3070ti because of 256bit/16gb of vram
You should compare it with the 3070/3070ti to see the gains
I wonder how feasible it is to swap out the Vram to gddr6 20GB. ? I mean is anyone skilled enough to remove the current vram and replace it with 20gb
this is actually in no way an actual representation of the performance of a RTX 3080 Ti mobile, i have a Strix Scar 17 with a 12950hx and 3080 Ti laptop i get much higher FPS in a lot of the games you tested maybe the low power usage on yours is seriously gimping it but mine usually runs between 160-175w depending on how demanding the game is on the CPU (this laptop prioritizes CPU power in certain titles which lowers the TDP of the 3080 Ti slightly)
Your laptop probably has a vapor chamber and power delivery far better optimized to utilize the performance available in the chip
@@Rotwold hmm yes it does have a vapor chamber but overall i think the cooling is quiet comparable because the fans on the graphics card you have are much bigger fans in a less restrictive environment (a pc case vs a limited intake laptop chassis) you also get much better temps than i do even though mine uses a vapor chamber and liquid metal but I'm pretty sure the massive performance issue is because of the power limit as the official 3080 ti laptops hover in the 150w and 175w range i just didnt like that your graphics card was referenced as an actual 3080 ti laptop/mobile variant when its actually even more gimped than a laptop 3080 ti
Oef you are putting it up against the 102 3080... those come with a Core Count of 10240 TMUs 320 ROPs 112, the 2 extra MB it has, means it has a big effect in more demanding games. You will see sh*tty 1% and 0.1% lows. It lacks the frequency curve to prevent that from dropping. If you had the overclock capability to push this around 1800 to 1900 you could probably do some real damage to the 3070 line up.
Oh wow
What a great video i am about to just watch like rat now
now i understand why my 3080 is a beast !
Looks like I am a galaxy brain
So it is like a 3065 then
I was wondering whether you have considered doing reviews on the 6750 gre’s? They seem like a good deal but I haven’t seen any real coverage from most of the tech reviewers I follow😊
I've got one! Unfortunately it's not going to get a standalone review until sometime in the new year, but it will be in a "console equivalent PC" video in a couple of weeks' time.
@ cool, which one? The 10 or 12 gb version?
@@sjargo11 10GB
daaaweed... i will never get over the pronounciation :D
i know, i know, this is my second comment about this. its not my fault!
This is China refurbished cards with mobile GPUs consume more than what the sensors show. Their consuming is fake :)
Check through the outlet ampere meter.
First time i saw it on rtx4000m cards and then checked rtx3000m.
he did check. It used 135w
How does it compare to 4060 ?
Nice
GA103 is 256bits wide on the memory bus, it's not meant to be a 320bits at all, how are you supposed to fit those nearly 100 more connection to the core while both GA104 and GA103 share nearly the same amount of balls.
You shouldn't base your informations on some "leak"/rumors, those are false, I bet the person that shared the info about GA103 being a native 320bits extrapolated this from 3080 rumors and wrongly said it was for GA103.
me watching this video on Nvidia 970
The RMA 30 series is such a joke
imagine if they let samsung build 4090. we will get a whole bunch of 4090D for very cheap😂
This card has regular 3060 12gb level of performance
Replace die with 3080 ti board 👌
If it's lost, then can I get it for free?
u need to find it first.
Dang
there are plenty of these m graphics cards taken from laptop. now there are desktop 4090m 4080m cards in china 😂
Shunt mod it so it can use full power and then retest it. It's annoying (and bad journalism) that people don't test GPUs with the power limits removed.
4:43 name of the movie?
Monty Python and the Holy Grail
A classic
8:36 thighs??
Good French
Yes
:)
shunt mod it, noob