No but, over the generations they add new technologies with improve raw performance, like forced induction and improved fuel injection. I was really hoping for a core count bump from AMD, considering Intel is beating them in productivity right now. AMD could easily take the outright performance crown with a few extra cores
@@CorneliusCornbread I mean are they really? when you have to suck half the power of a microwave to achieve it and it's causing mass instability with the product to be ran like that is it really a win? it's like finishing a marathon in first place but you took so many performance enhancers you keel over when you finish.
@@verde5738 and if it is proven that this is true, then what say you? It could mean support in many different ways. Like shear force, etc. Its unclear until its released and further explained. So don't assume anything until you have it in your hands and can prove it either way.
@@forog1 When you've been recording hours and hours of talking head videos, it happens. Some of these videos can take upwards of 50 to 100 hours of recording takes because you keep messing up. I can promise you, the most professional sounding channels record dozens of hours worth of takes to get just a 20 minute video finished.
Don’t let all this distract you from the fact that Hector is going to be running three Honda Civics with spoon engines, and on top of that, he just went into Harry’s and bought three T66 turbos with NOS, and a Motec exhaust system.
They already have been, this is a follow up to the MI100/200 which were also pretty good cards and AMD is finally starting to catch up to Nvidia on the software side of things.
@banishedbr shareholders are the ones who decide the direction a company goes. As long as money is made, it doesn't matter if consumers get upset. Individual consumers just don't buy enough to change their minds.
Doubt your PC will be good for 10 years. Sony asked AMD to put a 300 TOPS NPU on its APU for the PS5 Pro for a reason. AI hardware is used for DLSS. Right?
@@tringuyen7519 instead of focusing on upscaling i think they should actually make the cards run that resolution instead of relying on fuckin ai to make their cards perform better
@@hellize4212 Same here. AI stuff just does not interest me at all. My lowly Ryzen 5600 + RTX-3070 Ti + 32 GB of DDR4 SDRAM is WAYYY faster than I am, so making it even faster would be a complete waste of money at this point and for the foreseeable future.
@@banishedbrAMD selling their professional offering cheaper than nvidia is nothing new. They have been doing it like almost 20 years now? And yet nvidia still dominate the sales for professional gpu/compute card. Recently MS starting promoting AMD MI300X. but at the same time they also said they will be among the forefront getting nvidia blackwell solution. It's more like they try to diverting others (espcially the startup) to use alternative to slow them down in this heated AI space right now.
I do hate jumping on my motherboards and watching them break all the time. I knew AMD would be the first to fix my problem. Finally i can get some rest.
It all depends on how you're measuring it. My assumption is they're measuring shearing force. and obviously gpu's act as a lever so forces are being applied a little differently. If they meant 5.8kg that would mean you could rip your gpu out just by pushing on it a little bit.
nah 58 is quite a bit but sounds more than it is. if GPUs keep getting bigger the pull from the GPU will be worse because it extends outwards way more. Try to hold something of 10kg close to your body and then fully stretch your arms away from you the difference is really big. So the GPU doesn't have to be 58kg to put a 58 kg force onto the motherboard.
price to performance means that it will just be cheap enough to be better price to performance, but not better results when you ignore how much it cost
no, it exactly means what it says...better price/performance means better performance for the price you pay compared to the competition...doesnt matter if nvidias accelerator is 10X as fast because for the same money you can get 11 MI300 wich will be faster at the same job ( just for an example )...
A FAR better GPU sag/weight fix would be just to give the card a full metal spine on top. That way all the weight will rest on the I/O screws, giving you a full metal support. I personally push the card upward before tightening the screws on the I/O bracket to counter the sag
Umm... Will the thru holes on the MB PCB for mounting to standoffs also be reinforced for the 120+ LBS, or is the slot going to be fine, but the board is going to shear off at the thru holes?.....
AMD surpassing nVidia (arguable, tho) in their own "game", which now is indeed, not "games", but generative AI, which is quite a bad thing for society (jobs destruction, the artistic field, security issues, fake news, scams, the culture it creates about the easy solution instead of getting skills and learning in general, etc). So, them competing for the current cash cow/hype, and making games a second priority, not great news, IMO. As neither was with NFTs and mining.
@@miyagiryota9238 Nope... As I see it, generative AI is bad always. Heavily disliked the push of nVidia towards generative AI, since the very first announce. Even while I prefer nvidia for my graphic work (IMO, for games AMD GPUs tend to be of much better value, and it's overall from a much more customer friendly company) as I need a lot of nvidia exclusive features (CUDA for video and 3D, OptiX for fast GPU render in Blender, better driver support for 2D apps like the Affinity suite, etc), I yet deeply dislike how much nvidia is pushing AI in every sense. Detested it with all of my soul, and still do.
58 kg load weight will allow the slot to cope with a fumble-fingers dropping their case an inch or two (or three) with a heavy card in it without (hopefully) breaking anything. (e.g. 1 kg under 58 g acceleration force because of a sudden stop will put approx 58 kg stress on whatever is holding it).
No! Motherboard manufactures making super tough PCIe slots are encouraging Nvidia's awful BIG GPU decision! That will drive costs for hardware that shouldn't even exist! Nobody wanted fat un-manageable hardware, nobody's asking to pay for tough hardware they weren't going to use!!!
That 58kg limit might be a spoiler for what will be coming in terms of gpu size in the future. Really tought that this gen or next gen will be the max size and then focus their resources to make the chip cooler
i laugh when i hear about 'npu and gpu without cpu cores' because when i was a kid i had amiga 500 which had a 'maths coprocessor' - what eventually become known as a graphics processing unit. now we're going back to that, its just being assigned to do AI and redesigned to do math for that
At some point, games arent going to run off of the GPU the way we know, itll all be 100% AI generated with increasingly complex prompts designed by increasing suicidal prompt engineers.
I wonder if NPUs are going to be unavoidable in processors going forward. You do NOT want to run Windows 11 on a system with an NPU and have the full recall functionality running on your system.
Yep. Every CPU from now on (except maybe really budget ones, or ones that need extreme power optimization) will have an NPU built in. Both because it's the hot new thing, and because there will undoubtedly be more and more software coming that will take advantage of it to do "AI things".
At this point just combine the GPU with the motherboard, I'm tired of advising gamers to lay their PCs horizontally so they are not creating more electronic waste.
Yea, the slots are blocked anyway and the standard and all already exist. Just so a plastic hatch without the hook, so it shouldn't even make mounting more difficult.
9:16 Nvidia actually beating Nvidia at their own game? You made a mistake here bud, might wanna see how you'll get rid of it or make it clear that you mean that AMD beat NVIDIA at their own game.
fun fact amd and nvidia ceo are the same family members i'm pretty sure nvidia help amd develop there new cpu/gpu bc they have an ai which creates boards
"AMD beats Nvidia" is a bit missleading. YES the MI300 is faster...But its when using an open source OpenCL AI language. Nvidia's super power is CUDA, and CUDA is faster then OpenCL (on Nvidia GPUS). So its not an Apples-to-Apples comparison. But the new ROCm software (AMDs answer to CUDA) used in there Server GPUs is a real threat to Nvidia. Once we see ROCm used as ubiquitously as CUDA, that's when Nvidia will be in trouble. As of now, I think ROCm is not available for consumer GPUS yet.
@@ehenningsen The story I read said that a lot of AI software would need to be rewritten in order to switch from OpenCL/ROCm to CUDA. THats a huge cost for mayby 5-10% gain on THIS generation of GPUs. So, if you were starting from scratch TODAY, yeah, maybe AMD makes sense, but if you're currently deploying AI, then Nvida has their claws DEEP in you already.
Meanwhile perfomance per watt goes to hell. Making the biggest sillicon die for a GPU is simply not sustainable, i could get monolithic designs for CPUs (since they are extremely sensible on latencies and IPC per core).
Price/performance isn't Nvidia's game, that's the also-ran contestants runners-up prize. Nvidia owns the performance crown which I dare say is the most important metric. Hyperscalers won't bat an eyelid at the prices. It's a nice marketing ploy from AMD via MS to promote their gear.
That is why I am more Excited for the RDNA 5 than, RDNA4. Because RDNA 5 architectural change can have way more CUs, than RDNA 3 and 4. and If Redgamingtech is corrent if AMD is insane enough to make an RDNA 5 GPU with a whooping 384cu that is insane, It will be like 3 XTX of combine performance.
AMD is supposed to be better at price to performance, not just performance. Nvidias cards are like $50k and the new ones coming out are supposed to be like $70k. It could be half as powerful and only cost $10k and the price to performance would be through the roof but you'd still be getting half of the performance. I think it's good that AMD will be making some extra cash but we'll have to wait and see how well they stack up against each other.
Nothing was more predictable. Same thing happened to Intel's Itanic. Intel was supposed pwn the entire the server market with Itanium but AMD64 and Linux torpedoed those ambitions
I am not surprised that AMD beats Nvidias price to performance. If you compare with cars you could see Nvidia as a very fast and luxury car with top performance but when it comes to real life, then Fiat purchased Chrysler some years ago. There are much more people buying tiny cars than Cryslers. The question is does the companies need the fast performance of the Nvidia chips or will AMD chips be more suitable at a much lower cost and AMD AI chips have one more ace, they run open source programs, which will mean more competition and lower prices. It was recently publish that Nvidia demand and supply is more even so I guess many companies have come to the same conclusion as Microsoft.
9:08 not sure if that's really beating nvidia. I would argue performance per watt to be more important in datacenter ai workloads than price to performance, and i don't know who wins that.
Nvidea 700w AMD 750w for AI chips. AMD Just announced 100% power to compute reduction around end of next year. Around the time of the release of mi400. Great news for gamers for rdna5
Does intel really having "more cores" make a difference when half of them are slower, have less features, cause performance issues with scheduling and has the same total number of threads anyway? ohhh, and like 20% of them are on RMA
Because mi300x's secret it out, nvidia will have to now reveal blackwell and if it doesnt surpass mi300x then amd will probably not only have the cheapest ai chip but the fastest one, and nvidia will have the slowst and most expensive. If this happens and no one buys from Nvidia, they will most likely try to make up for this in one or two ways, one reduce the msrp of the 5000 series and sell as much as possible or, reduce proces for blackwell and ada ai chips to meet the demand and hope someone buys it. If nvidia also reduces proces of blackwell ai chipsets then we may or may not see reduced pricing of 5000 series wich will onlt further prove that they are trying to nickle and dime the gaming community yet again... I guess it just works?
It's good to have AI competition, then Nvidia might pay attention to the low hanging fruit of the graphics GPU market side of the house. Otherwise it will become an afterthought.
engineering the motherboard to handle 600W would add unnecessary cost and complexity. The psu already has to handle all that power anyway, so tapping directly from it as is done today is the simplest most elegant design.
AMD hasn’t won until they out-perform or sell more units and they aren’t close to doing either. I’m glad they will get the customers who are lower tier and can’t afford Nvidia products though. Good place in the market for AMD.
Ferrari's new engine still has the same number of cylinders as the previous generation.
Hahaha Amd wins?
They have no big or small cores and
No but, over the generations they add new technologies with improve raw performance, like forced induction and improved fuel injection.
I was really hoping for a core count bump from AMD, considering Intel is beating them in productivity right now. AMD could easily take the outright performance crown with a few extra cores
I like it.
@@CorneliusCornbread I mean are they really? when you have to suck half the power of a microwave to achieve it and it's causing mass instability with the product to be ran like that is it really a win? it's like finishing a marathon in first place but you took so many performance enhancers you keel over when you finish.
There's absolutely no way that PCIe can withstand 58kg. That's 100% a typo and it was supposed to say 5.8kg.
prob. with a thicc ass pcb
58kg is the weight of a lean adult human being. No way in hell ANY PCIe/PCB could support that kind of weight.
imagine the shipping cost
@@verde5738 and if it is proven that this is true, then what say you? It could mean support in many different ways. Like shear force, etc. Its unclear until its released and further explained. So don't assume anything until you have it in your hands and can prove it either way.
most likely its not. 5kg card can give over 20kg of force into the pcie connector depending of the dimensions.
Silly that GPUs keep getting bigger and bigger.
This is the price we pay for not investing in photonics instead of silicon chips.
you afraid of 100lbs GPU? ha! puny man!
Not gpus... NVIDIA GPUS!
I mean we are building cases around Nvidia GPUs now which makes them all go up in size.
I like em big, makes me feel like I'm getting my money's worth, even though I'm not.
9:18 I think Nvidia catching up to Nvidia and beating Nvidia at there own game is paradoxical!
I caught that to. I had to rewind and see if I heard right. Not sure how he would mess that up. Seems hard to do by accident lol.
@@forog1 When you've been recording hours and hours of talking head videos, it happens. Some of these videos can take upwards of 50 to 100 hours of recording takes because you keep messing up. I can promise you, the most professional sounding channels record dozens of hours worth of takes to get just a 20 minute video finished.
That means Nvidia has some catching up to do... Now Nvidia is ahead of them
@@AlphaMachina Well, HERE'S A GIANT "NO-SHIT!" TACO!
@@CZcamsSupportServices A giant no shit taco that apparently needed to be explained.
I think they should stop altogether expecting you to plug the GPU. Just instal the GPU on the Case and Plug the Motherboard into the GPU.
At this point that's not a bad idea. That way they could really do 100lb GPU coolers and not matter lol
Or just use a riser like many cases are going anyways?
Back to the future.
Horizontal backplane with the motherboard and gpu vertical, back-to-back.
Card retention slots inside - IBM AT style too.
@@myne00
I love horizontal full size desktop PC
@@myne00
I already do this, the only problem it's that there is no space for the loquid coolers.
@2:00 Sorry, but in the US, it's better you describe the weight as 9547328 ants or 0.00001 fridges.
1/2 a football field apparently.
That makes no sense😂😂
@@DeepDownInTheOceanPrecisely.
@@therealwhite that’s why I laughed
The joke is "Sorry in the US we only understand the weight in terms of 100 hamburgers or 1/2 a fridge."
Don’t let all this distract you from the fact that Hector is going to be running three Honda Civics with spoon engines, and on top of that, he just went into Harry’s and bought three T66 turbos with NOS, and a Motec exhaust system.
9:15 Man Great to hear that Nvidia is catching up to Nvidia
Well there you go, AMD is focusing on HPC/AI and not consumer GPUs.
AMD is a business with shareholders. It should do whatever’s best to maximize revenue & profit. Nvidia’s the same. Right?
@@tringuyen7519 there are more consumers than shareholders right ?
They already have been, this is a follow up to the MI100/200 which were also pretty good cards and AMD is finally starting to catch up to Nvidia on the software side of things.
@banishedbr shareholders are the ones who decide the direction a company goes. As long as money is made, it doesn't matter if consumers get upset. Individual consumers just don't buy enough to change their minds.
@@banishedbr quality > quantity. One 20 million dollar vendor is worth a hell of a lot more than 20 sweaty gamers
Yeaahhhhhh, I'm not doing anything A.I. I've got a beast of a PC build right now and will be using it for the next 10 years or more.
I have the same plan
Doubt your PC will be good for 10 years. Sony asked AMD to put a 300 TOPS NPU on its APU for the PS5 Pro for a reason. AI hardware is used for DLSS. Right?
@@tringuyen7519 instead of focusing on upscaling i think they should actually make the cards run that resolution instead of relying on fuckin ai to make their cards perform better
Can your pc run hellblade 2 on 60 fps max settings?
@@hellize4212 Same here. AI stuff just does not interest me at all. My lowly Ryzen 5600 + RTX-3070 Ti + 32 GB of DDR4 SDRAM is
WAYYY faster than I am, so making it even faster would be a complete waste of money at this point and for the foreseeable future.
Literally all AMD needs is to sell MI300X for 30000$ and still pull in over a 1000% profit
or 10k and make nividea sell nothing.
@@banishedbr😂😂 copium
@@banishedbrAMD selling their professional offering cheaper than nvidia is nothing new. They have been doing it like almost 20 years now? And yet nvidia still dominate the sales for professional gpu/compute card. Recently MS starting promoting AMD MI300X. but at the same time they also said they will be among the forefront getting nvidia blackwell solution. It's more like they try to diverting others (espcially the startup) to use alternative to slow them down in this heated AI space right now.
@@banishedbr Brah, the 7900XT >= 4070ti, costs 700$ vs 800$. The 4070ti is far more popular.
@einstien2409 brand loyalty is strong. Even though nvidia keeps bfing you. I'm going amd next gen for my gpu, and already switched with cpu.
cache, cache and more cache...the amount of cores is currently fine
Hey, how much do you bench press?
3 nvidia's gpu 💪
I do hate jumping on my motherboards and watching them break all the time. I knew AMD would be the first to fix my problem. Finally i can get some rest.
Surely they meant 5.8kg.. not 58kg
That would make it 12 lbs. so very likely.
It all depends on how you're measuring it. My assumption is they're measuring shearing force. and obviously gpu's act as a lever so forces are being applied a little differently. If they meant 5.8kg that would mean you could rip your gpu out just by pushing on it a little bit.
nah 58 is quite a bit but sounds more than it is. if GPUs keep getting bigger the pull from the GPU will be worse because it extends outwards way more.
Try to hold something of 10kg close to your body and then fully stretch your arms away from you the difference is really big. So the GPU doesn't have to be 58kg to put a 58 kg force onto the motherboard.
I doubt it. if it can't hold up my 75 inch TV, then I don't want it
I’d assume it could take in account for transit forces for pre builds. Nothing wrong with overkill though, unless it effects the price drastically 😅
Do you have any videos talking about intel arc battlemage gpus?
Yes please, Idc about 5090s or the new AMD mbs, like less than 20% of gamers will use either. Less than 1% for the 5090.
AMD never tried to be the fastest, AMD always tried to be the most performant at a certain price. And they do it very well.
price to performance means that it will just be cheap enough to be better price to performance, but not better results when you ignore how much it cost
no, it exactly means what it says...better price/performance means better performance for the price you pay compared to the competition...doesnt matter if nvidias accelerator is 10X as fast because for the same money you can get 11 MI300 wich will be faster at the same job ( just for an example )...
Man if it can support 58kgs, gpu cooling solutions are abt to get real interesting
*When your guest goes to get a drink and thinks the GPU in the kitchen is your fridge.* How do you like my 5090?
A FAR better GPU sag/weight fix would be just to give the card a full metal spine on top. That way all the weight will rest on the I/O screws, giving you a full metal support.
I personally push the card upward before tightening the screws on the I/O bracket to counter the sag
Umm... Will the thru holes on the MB PCB for mounting to standoffs also be reinforced for the 120+ LBS, or is the slot going to be fine, but the board is going to shear off at the thru holes?.....
AMD will never beat Nvidia.
Hey, i think the 150+ kg weight limit is not for gpu's that heavy, but to protect pc's with moderately heavy gpu's when transporting :)
AMD surpassing nVidia (arguable, tho) in their own "game", which now is indeed, not "games", but generative AI, which is quite a bad thing for society (jobs destruction, the artistic field, security issues, fake news, scams, the culture it creates about the easy solution instead of getting skills and learning in general, etc). So, them competing for the current cash cow/hype, and making games a second priority, not great news, IMO. As neither was with NFTs and mining.
Yeah when it was nvidia leading its good, when AMD leading then its automtically bad????
@@miyagiryota9238 Nope... As I see it, generative AI is bad always. Heavily disliked the push of nVidia towards generative AI, since the very first announce. Even while I prefer nvidia for my graphic work (IMO, for games AMD GPUs tend to be of much better value, and it's overall from a much more customer friendly company) as I need a lot of nvidia exclusive features (CUDA for video and 3D, OptiX for fast GPU render in Blender, better driver support for 2D apps like the Affinity suite, etc), I yet deeply dislike how much nvidia is pushing AI in every sense. Detested it with all of my soul, and still do.
GUYS THE 5090 TI SUPER DUPER WAS JUST LEAKED USING 1200W OF POWER!!!
58 kg load weight will allow the slot to cope with a fumble-fingers dropping their case an inch or two (or three) with a heavy card in it without (hopefully) breaking anything. (e.g. 1 kg under 58 g acceleration force because of a sudden stop will put approx 58 kg stress on whatever is holding it).
that gpu slot can technically hold me, im only like 125
Nvidia catching up and beating Nvidia at their own game? Error!!!
You beat me to it lmao
How do you put in a GPU which is 58kg , you would need 2 people ?
Naw , I could do it myself , I know how to run a fork lift .
Weak
And multiple hamsters rolling in the wheel to generate enough power
I'll bring my scaffolding
you need GPU loader and buy bigger one to be future proof, 200kg min max load...
The end of the video, you said "what do you think about Nvidia catching up to and beating Nvidia."
get over it dude, it s an honest mistake. everybody does this from time to time
No! Motherboard manufactures making super tough PCIe slots are encouraging Nvidia's awful BIG GPU decision! That will drive costs for hardware that shouldn't even exist! Nobody wanted fat un-manageable hardware, nobody's asking to pay for tough hardware they weren't going to use!!!
Nvidia are moving to the "cinder block" design, probably as cinder blocks are fire proof. 😂
That 58kg limit might be a spoiler for what will be coming in terms of gpu size in the future. Really tought that this gen or next gen will be the max size and then focus their resources to make the chip cooler
i laugh when i hear about 'npu and gpu without cpu cores' because when i was a kid i had amiga 500 which had a 'maths coprocessor' - what eventually become known as a graphics processing unit. now we're going back to that, its just being assigned to do AI and redesigned to do math for that
At some point, games arent going to run off of the GPU the way we know, itll all be 100% AI generated with increasingly complex prompts designed by increasing suicidal prompt engineers.
"Nvidia catching up and beating nvidia at theie own game?"
😱
I wonder if NPUs are going to be unavoidable in processors going forward. You do NOT want to run Windows 11 on a system with an NPU and have the full recall functionality running on your system.
I'm sure and hope there is an option to enable/disable the NPU in the bios just so you can have the control if some folks don't want it
Yep. Every CPU from now on (except maybe really budget ones, or ones that need extreme power optimization) will have an NPU built in. Both because it's the hot new thing, and because there will undoubtedly be more and more software coming that will take advantage of it to do "AI things".
your phone has npu, you need to disabled it
@@zz.d.great4832 no it does not :D
58kg might be a typo... could be 5.8kg which is about 12 pounds
At this point just combine the GPU with the motherboard, I'm tired of advising gamers to lay their PCs horizontally so they are not creating more electronic waste.
58kg is technically impossible to do on that slot, come on, this is can't be serious!
im so glad i went with a horizontal build so i dont have to worry about my GPU snapping
You need to sort your microphone theorem is audio delay… kinda hard to watch
IMHO GPU-s should be mounted on motherboard like CPU and have common cooler.
You know, cards are getting so massive, I am surprised that they don't just make dual PCI-e cards to fit in 2 PCI-e slots.
Yea, the slots are blocked anyway and the standard and all already exist. Just so a plastic hatch without the hook, so it shouldn't even make mounting more difficult.
AMD's changin candle into a massive blowtorch under Nvidia's...
You know what I mean
Nice to have some healthy rivalry 'n competition
Is the 9k series gonna be on am5 still though?
It's starting to sound like we will have to start screwing the GPU to the motherboard to support the back end if they get that heavy.
I don't think Nvidia's game has ever been price/perf. They are about ultimate performance regardless of the price.
9:16 Nvidia actually beating Nvidia at their own game? You made a mistake here bud, might wanna see how you'll get rid of it or make it clear that you mean that AMD beat NVIDIA at their own game.
I know you're trying to get videos out as fast as possible, but cmon
I think that 58kg pci-e slot os not meant to hold gpu´s that massive but to not destroy the motherboard while the gpu will be f***ed
At this point new x090 cards are as desirable as a Cybertruck.
LINEUP FOREVER: 1. nVidia; 2. intel; 3. MediaTek; 4. AMD.
Amd is always making customers believe they're confused!! I was once a fan, not anymore. Nvidia and Qualcomm are the way to go.
fun fact amd and nvidia ceo are the same family members i'm pretty sure nvidia help amd develop there new cpu/gpu bc they have an ai which creates boards
What do you think of nvidia, catching up and beating nvidia at their own game? hmmmm 9:18
Amd scoring another own goal making everything too expensive to keep up with the Joneses (NVIDIA) and INTEL
"AMD beats Nvidia" is a bit missleading. YES the MI300 is faster...But its when using an open source OpenCL AI language. Nvidia's super power is CUDA, and CUDA is faster then OpenCL (on Nvidia GPUS). So its not an Apples-to-Apples comparison. But the new ROCm software (AMDs answer to CUDA) used in there Server GPUs is a real threat to Nvidia. Once we see ROCm used as ubiquitously as CUDA, that's when Nvidia will be in trouble. As of now, I think ROCm is not available for consumer GPUS yet.
Everyone knows people buy Cuda cores to run OpenCL 😂
@@ehenningsen The story I read said that a lot of AI software would need to be rewritten in order to switch from OpenCL/ROCm to CUDA. THats a huge cost for mayby 5-10% gain on THIS generation of GPUs. So, if you were starting from scratch TODAY, yeah, maybe AMD makes sense, but if you're currently deploying AI, then Nvida has their claws DEEP in you already.
What do you think about NVIDIA catching up to NVIDIA and beating them at their own game?
Leading price performance = marketing speak for “this can’t compete on performance but it’s not too expensive”
AMD CPU + Nvidia GPU couldn’t care for other config’s
nvidia's obsession with AI and neglect of their bread/butter GPU market make AMD surpassing them very sweet. They deserve for AMD to spank them.
Ryzen 9000 Desktop CPU's will still be called Ryzen 9000. They wouldn't have announced them as such at Computex if they weren't.
This "AI" trend is remembering me of the "3D" trend we have on the past, everything needed to have 3D in the name
WELP, AM5 till 2025 they said. Nobody keeps up with their promises anymore
can the just not use screw for PCIe?
RIP Akira Toriyama!
Great question at the end.
Meanwhile perfomance per watt goes to hell. Making the biggest sillicon die for a GPU is simply not sustainable, i could get monolithic designs for CPUs (since they are extremely sensible on latencies and IPC per core).
I swear amd would release the 192gb gpu to the average customer and still would be cheaper than a 4090
every time i see the new intel core ultra. only thing i think is ICU.
9:18 can nvidia beat nvidia at there own game?
Price/performance isn't Nvidia's game, that's the also-ran contestants runners-up prize. Nvidia owns the performance crown which I dare say is the most important metric. Hyperscalers won't bat an eyelid at the prices. It's a nice marketing ploy from AMD via MS to promote their gear.
if they do not have neural engine than they should just cancel release
That is why I am more Excited for the RDNA 5 than, RDNA4. Because RDNA 5 architectural change can have way more CUs, than RDNA 3 and 4. and If Redgamingtech is corrent if AMD is insane enough to make an RDNA 5 GPU with a whooping 384cu that is insane, It will be like 3 XTX of combine performance.
day 1 of asking not to put thumbnail topic in the end of the video
Godspeed
AMD is supposed to be better at price to performance, not just performance. Nvidias cards are like $50k and the new ones coming out are supposed to be like $70k. It could be half as powerful and only cost $10k and the price to performance would be through the roof but you'd still be getting half of the performance. I think it's good that AMD will be making some extra cash but we'll have to wait and see how well they stack up against each other.
A 9:18 you flubbed. You said "Nvidia surpassing Nvidia at the own game". I think you meant *AMD* is surpassing Nvidia at their own game.
Nothing was more predictable. Same thing happened to Intel's Itanic. Intel was supposed pwn the entire the server market with Itanium but AMD64 and Linux torpedoed those ambitions
Ugh. AMD's product naming is getting r*tarded. They're the market leader, they need to stop copying Intel.
Dude, AMD is not even in the same League as Nvidia when it comes to AI! 🤣🤣🤣
I am not surprised that AMD beats Nvidias price to performance. If you compare with cars you could see Nvidia as a very fast and luxury car with top performance but when it comes to real life, then Fiat purchased Chrysler some years ago. There are much more people buying tiny cars than Cryslers. The question is does the companies need the fast performance of the Nvidia chips or will AMD chips be more suitable at a much lower cost and AMD AI chips have one more ace, they run open source programs, which will mean more competition and lower prices. It was recently publish that Nvidia demand and supply is more even so I guess many companies have come to the same conclusion as Microsoft.
Why on earth there is Lisa Su photo with the following comment of Satia Nadela 🙈
9:08 not sure if that's really beating nvidia. I would argue performance per watt to be more important in datacenter ai workloads than price to performance, and i don't know who wins that.
Nvidea 700w AMD 750w for AI chips. AMD Just announced 100% power to compute reduction around end of next year. Around the time of the release of mi400. Great news for gamers for rdna5
Does intel really having "more cores" make a difference when half of them are slower, have less features, cause performance issues with scheduling and has the same total number of threads anyway? ohhh, and like 20% of them are on RMA
So does that mean we will get a high end consumer gpu from amd?
58KG GPU, let's goooooo!!!!!
Finally, i can download Cod MW3 on VRAM...
At the end of the video you said Nvidia twice for example you said, do you think Nvidia will be Nvidia at their own game
I frl just bought a new aorus b650 mobo and then i see this..
interesting but NVIDIA stock says otherwise.. 1,140$ today, so a RTX 4080 price lol
What is nvidia's own game? Is it selling x050 cards with the number x070 written on at the price of a x090 card?
Don't know who's going to buy that 5090... with a larger die size and three PCB's it'll be at least 25-50% more expensive than a 4090...
Because mi300x's secret it out, nvidia will have to now reveal blackwell and if it doesnt surpass mi300x then amd will probably not only have the cheapest ai chip but the fastest one, and nvidia will have the slowst and most expensive.
If this happens and no one buys from Nvidia, they will most likely try to make up for this in one or two ways, one reduce the msrp of the 5000 series and sell as much as possible or, reduce proces for blackwell and ada ai chips to meet the demand and hope someone buys it.
If nvidia also reduces proces of blackwell ai chipsets then we may or may not see reduced pricing of 5000 series wich will onlt further prove that they are trying to nickle and dime the gaming community yet again... I guess it just works?
Is the 5090 gonna be called the BOGPU (Big Obese GPU)
It's good to have AI competition, then Nvidia might pay attention to the low hanging fruit of the graphics GPU market side of the house. Otherwise it will become an afterthought.
am I only one who skipped 5090 segment?
It looks like if you want a high-end gaming GPU that's not heavily AI focused, the time to get one is going to be -- now.
At this point , they should just redesign the MB so the gpu plugs in another socket on the MB.
engineering the motherboard to handle 600W would add unnecessary cost and complexity. The psu already has to handle all that power anyway, so tapping directly from it as is done today is the simplest most elegant design.
AMD hasn’t won until they out-perform or sell more units and they aren’t close to doing either. I’m glad they will get the customers who are lower tier and can’t afford Nvidia products though. Good place in the market for AMD.