This is normal. It's called "rolling through a take." He just missed it when it came to editing it out. This happens quite a bit on channels that upload 1-2 times a day.
I think AMD's biggest "problem" is their previous gens are so good I know a lot of folks that are just sticking with Zen 3 or 4 because they already have more CPU power than they got work for it to do. I know my 5700x is barely using half the threads most times, and my friends say the same of their 5800x3d and 7500f. These chips are running so well with so little heat none of us on our Discord feel a need to upgrade.
Maybe upgrade my gpu and switch out my psu and aio to a new one since ive been running them for so long but im pretty much sticking with the 5900x until if the ryzen 10xxxX comes out and thats only if it has more cores and/or x3d on both ccds. I just want to make my money worth it at this point and probably just going to build a whole new system leaving this one whatever way it is from that point and use it for a server pc or something.
I agree with what you said about previous gens being so good. Many people are just satisfied with getting a Zen 3 CPUs. The extra bonuses from Zen 4 might be nice but many just don't want to spend the extra money when the prices for Zen 3 have been going down a lot. I recently (back in late feb/early march) upgraded from an i7 4790K to a Ryzen 5950x. It was a complete upgrade (minus GPU) as I got a new NVMe SSD (previous used SATA SSD), motherboard, CPU, new cooler (bought an AIO for the first time), case, and a PSU (upgrade from a 750 Watt to a 1,000 Watts). I would have gone with Ryzen 7950x but the cost would have easily been like $400 to $600 more (with the parts I spec'd with it). It's such a major upgrade from what I had, I'll be fine for many years). By the time I upgrade again, AM6 will probably be out (way down the line) and I'd probably just get the latest AM5 at a major discount. It was nice going from a CPU with 4 cores and 8 threads to having 16 cores and 32 threads. They will come in handy when I eventually do some video editing or even capturing gameplay.
I have a 5700x. I gotta upgrade to AM5 soon to keep going with my high end nvidia GPU buying addiction because soon I'll want an RTX 6090 and I don't think the 5700x will be all that great with a GPU that strong.
It's smart, transistors are getting very difficult to decrease in size. We have entered the architecture era of super focus to redesign and optimize hardware/software systems to continue the evolution for awhile. I do believe that we will see some massive gains this way, but that will eventually lose steam too.
Not if we find a new transistor, femto-scale engineering kind. Or a very fast switching one, some have been demo'd at 20Ghz+. Or optical computing, very low power use, extreme frequencies, like 100Ghz.
Optical is not faster. Its bigger and slower and use more power than today most advanced transistors. And they can run at 10 GHz max but same problém with electrical@@jamegumb7298
I love your videos brother, and like some of my other favorite Tech CZcamsrs, you speak freely, like a normal person would and not like a person who is trying to sound professional like a lot of news and weather presenters lol. The thing I love about your channel, besides the content you bring, is the desire for everyone to have a great day at the end of your videos. I wish a lot more people did this. On that note, I say to you, Have a great day, and also be blessed, as you are a blessing to us. Keep it up brother ☺☺☺
yeah since right now we have reached a point where we are limited by the architecture. Its like building a castle on wood beams. A strong foundation will allow u to not only build that castle, but also expand it
3:48 yea... i really havent had a problem with "download huge game packages" since I was finally able to get off of dial up and get a cable modem.... 20 year ago
For most people the problem is running out of storage rather than downloading it I mean 1tb 5 years ago was enough to have a bunch of games now you can only have a few since everything tends to be 100gb or more which is ass
They are also generally games poorly optimized by the current regards of what these companies keep endlessly talking about as innovative nowadays. So we see games take 400Gb and use 24GB v ram. When those same games optimized would use only 6 to 8 GB v ram, and roughly 100 to 200GB of space too.
@@Qmohd1 it's ONE video, each day - Monday through Friday... you must be the busiest person on Earth not being able to find 20 mins per day... maybe you should give your wrist a break and stop trying to compete for the largest right-handed forearm.... you know you can PAY to lose your virginity, right? you don't have top spend your ENTIRE day fantasizing about your Mom.... smh
If they compress data files and textures, file size are lower and that means the entire game file size is smaller so you can download it faster since most new game files are close or over 100GB these days. If it can be implemented in all games then valve and other game launchers can use it to compress most if not all game files and that their entire servers have more space to work with. That also means if games also use less vram since the textures are compressed and can load extremely fast with better quality.
AMD, Intel and Nvidia need to squash their beefs and work with Apple, Google, Sony, Microsoft and Meta. The sooner they make apps into physical and permanent products, like in the earlier stages of technology, the sooner we can make a WEB 3.0 and a multi-national citizenship system.
First, DX9 already support texture compression. What AMD & NVidia showed are just new compression techniques based on AI. The decoding of these AI compressed textures don't necessarily need AI to decompress so it can be easily implemented as a new compression mode OR with NO changes to runtime at all. During the initial stages of compression, AI can be used to adjust the weightage of the compression parameters to maximise luminance & colour details to get the best image quality after standard JPEG decompression. (Search JPEG compression process if you have comsci background). I don't want to explain what's discrete cosine transform. Below is my guess, I haven't read the papers: For NVidia, it looks like they will do this automatically if your GPU don't have enough RAM. This reduces the need for texture streaming. It STILL means that lossless compression is used during downloading of game files so it's still 150-200GB for Microsoft Flight Sim 2020. For AMD, they may be proposing to deliver lossy compressed textures during download but at a quality that is "Visually lossless". This will really help their console customers. For Consoles, the RAM is unified so texture streaming is NOT needed. Same for APU with Radeon 780/880/890M iGPUs. Next, Intel Fab is currently working with ASML on the 18A (and better) High-NA EUV wafer machine, which TSMC is not currently involved in. TSMC's most advanced EUV machine from ASML can make wafers down to 3nm or 30A at similar yield. Check Asianometry website or YT channel for this topic. What this means is, Intel CAN leapfrog TSMC for the 2nm/20A node and can downsize to even lower with the High-NA EUV process in 2-3 years. If high-NA EUV works... Lastly, Intel Meteor/Raptor-lake don't support AVX512 so it loses badly to AMD Zen 4 & 5. This is very obvious when running the QUBIC mining client which tests different AI workloads and AVX512 powers the fastest algorithm. Zen 5 having a 2x front-end to instruction/data load really helps due to a wider data bus to/from cache. Similarly, Zen 4 Threadripper supporting qual/octo-channel RAM can feed the FP units much faster than in a dual-channel Zen 5. Zen 5 based EYPC/Threadripper are going to be monsters!
Imagine amd releases a universal open source compression technique that makes 8gb gpus become viable again and everyone runs to buy 4060tis because of this release. Seems like a classic amd move.
5:30 it’s the contrary, AMD talk about the plan of intel to go ahead of Tsmc in term of process, amd is saying they’ll continue that path and taking the risk to be behind in term of process if intel makes it. The backside power delivery and all this crazy bet of intel.
I'm not exactly excited about Intel falling so far behind AMD in the CPU space. It's better for all of us when they're neck-and-neck. More competition means more options and better prices. It's also becoming more dangerous for Taiwan to be doing so well in the compute space for so long. China keeps edging closer and closer to invading and having control over cutting-edge technology (especially in AI) is the biggest motivating factor at this point.
Didn't TSMC state they had solved the issue with Quantum Gate Tunneling several years ago? I know they have made breakthroughs to even smaller transistors than 3nm.
So a better Frame Generation with AI. Better compression is also good, 'coz every single data type is getting more resolutions and more of them; maybe it could help us generate better compression for lossless images or improve the current image compression formats
So GPUs needs to be more powerful now to decompress/upscale textures instead of just having more memory space.... I wonder if memory is cheaper or the processing unit that does this upscale/decompress thing.
Might be memory bandwidth capacity. Only so many attachment points for wires etc…. Rather than update how circuit boards are printed you can send compressed textures across pins and the core can quickly decompress them.
Nvidia started to convert parts of gpu hardware to software when they introduced dlss and frame generation. By addind texture compression and real time model and environment generators they are planning to reduce the need for hardware even further. Soon NVidia gpus will become pure software to download from their site.
But see, game graphics are so good at this point that we don't need to improve them. Especially after UE5.3 or whatever. A 4090 already covers Cyberpunk with 4+ rays 4+ bounces modded path tracing and 8k textures, at 8k native render resolution. I KNOW this will be used to push the envelope further and games will still be 100-200gb, they'll just be able to brag about the now even better graphics or even bigger worlds (that one may be valid, but only if the worlds are detailed and don't feel empty). Just keep games how they look now (or will/would in UE5.3 with 4R/4B path tracing as the practical max), with hardware as powerful as it is now, and focus on making the existing stuff more efficient in terms of power consumption, heat, size, and data storage, then add features (like AMD improving at ray/path tracing). And also, we don't need this compression to be 67 to 1. 10 to 1 or 20 to 1 is fine, and will likely improve the appearance more, too because that example compared to the reference is garbage. It's like TAA and "quality" upscale rendering vs 4x MSAA rendering at native resolution.
All these people watch these videos everytime he uploads a new one. I have for awhile, but I have noticed most you comment then just cry and complain. You must not understand what your watching. I don't respect people who's entire life is just constant complaints.
they are essentially the SAME chip... they just selected all the best binned 13900K's and upped the power and speed by a small amount... lol ... - nah it's actually a new chip but nearly identical to 13th Gen... they kept everything the same except for I think the 14700K by adding 2 more 'big' cores.... you MIGHT get a 14th gen chip that's always 1% or 2% faster than the 13th gen equivalent, but that's still within the margin of error, so pretty much all the big names on youtube and the big websites ALL just consider them the same chips when it comes to RAW performance, so it honestly wouldn't change anything if they did use a 14th gen :)
I'm STOKED for this new compression tech by team green AMND red!!! ( and you KNOW Intel will have their own version soon too lol) I hope to learn more about the AMD tech, since you weren't able to cover much detail on it, but Nvidia's tech looks to me outstanding! As much as I cannot STAND Nvidia's management and the way they treat their customers AND partners, I will give them credit where it's due and this stuff looks to me AMAZING! But like ALWAYTS you'll probably need to have one of their newer cards to use it, so as ALWAYS I'll be using team Red because I support what little bit of open-source/open-ness that they have and share... So I will support the the company... "team" that aligns more with where and how I want to see the industry to head in.... and even tho they have been making MASSIVELY DUMB DECISIONS as of late, they're still the same team at the core and I think they're a GREAT balance to Nvidia's nauseating dribble about themselves..... I just wish AMD would think more about doing their OWN THING and not copying EVERYTHING nvidia does.... it's actually kinda pathetic.... who making that decision anyway??? :( - - Oh but I am SUPER EXCITED for Zen 5!!! I'm still using a 5900X and feel a bit left out being on DDR4 and pci Gen4 ... but I'm sure I can hold out a few more generations before my baby starts to feel slow lol.... (obvious I'm being sarcastic... I don't even need anything this powerful, let alone an upgrade... I most just watch youtube, some movies or shows, and a bit of gaming occasionally....so upgrading would just be about vanity at this point lol :D) But I'm actually VERY EXCITED for it! Progress is awesome and I LOVE seeing it up and close everyday and hearing all the cool things about before its released... and that is EXACTLY why I'm subbed here @ Gamer Meld!!! :) Thanks for everything you do bro :)
That is good news, it is encouraging, but a) that's early silicon, and b) you don't know what clock speeds that engineering sample was running, nor what the other systems being compared were running. In short ~ not going to say that's meaningless, but it's only a pre-release leak to tease us. It doesn't actually mean anything unless we have a whole lot more detail. For one thing, we know that AMD have made changes to the AVX512 pipeline, such that they don't have to break those ops up into 2 parts. It now runs them through in one piece. That has (I understand) more or less doubled AVX512 performance. Let's say a 75 ~ 80% increase. Ok, that's good, so if you have a benchmark that exclusively tests AVX512 performance, you'd expect a Zen 5 chip to have nearly double the performance of an earlier AMD chip. And that's what we see. Ok ~ so does that mean it will deliver an 80% jump in frame rates? No ~ that's not what it means. Is it going to help frame rates? Yes. Sure it will. By how much? Remains to be seen. And it depends on how the game (or other job / task) was written. If it makes big use of big floating point numbers and the AVX512 pipelines, then expect to see impressive gains. If it never goes near those instructions ~ Then there's questions about speed and timing and latency and test-data and ~ let's wait for independent 3rd party reviews to land. I am excited, but let's not get crazy about this. Linux Mint 22 is due to land (the Beta / test) some time soon. I'm excited about that too.
Both images are not similar. Yes, the compressed file looks good, but there is obviously a significant loss in image quality. Impressive compression? Yes. Comparable quality? Hell no.
Some things we just don't want compressed though. The compressed and uncompressed images are clearly different to the human eye and may lower the overall visual experience if one cares about that. It may be useful for cases where a person does not care, meticulously, about the visual experience, but knowing nvidia, they would probably try to force it on people just to see higher FPS across all their GPUs across all games 😐
If you mean the DLSS, I would say, you are right. And nvidia fans just save nvidia. But rtx as well... Really? You have to turn on dlss for it? 😲 That is just... Maybe, 6090 will give us 120 FPS with rtx/path tracing at 1080p without upscaling. 😞
@quintrapnell3605 and even then, AMD still uses AI just at the hardware level. I've been enjoying driver level frame generation and FSR for unsupported games on my 7900 XTX
@@nesseihtgnay9419IA is already on most 2005 PCs. Even more on GPU Nvidia or AMD use IA. Every game uses IA. I know hearing gaming everywhere and IA everywhere can be annoying but it's better for us.
Finally someone that understands that die size matters. Also we already have the rtx 4060 being a 30/40 class GPU since it's 147mm^2. Keep in mind the reticle limit is higher than ever so 147mm^2 is like 80-90mm^2 on 28nm.
@@badass6300 I thought the 4060 was 159mm² and 4060TI was 190mm² The 1650 is 200mm² with the same 128 bit bus as the 4060TI 16GB (FYI you can hold alt and press 0178 on the number pad for ² 0179 fpr ³ but make sure numlock is on, otherwise you'll reload or change the web page)
@@denvera1g1 Yes, it's 159mm² my bad, still that's a 40 class chip. The rtx 3050 6GB is 200mm², the RTX 3050 8GB is a cut down chip from 276mm^2 down to 224mm^2, the gtx 1650 is 200mm^2. Only the gtx 1000 series which were just as scummy, if not more than the rtx 4000 series have small die sizes. For example the gtx 1080 is 314mm² which is straight up a 60 class card, and should have costed 350$, maybe 400$ due to lack of competition from AMD, instead of 600-(700)$ (FE)MSRP... The gtx 950 is also a cut down 960, but still comes out at 176mm² out of 226mm² Of course the gtx 600 series the original bs generation, where a 3-4 weeks before launch the gtx 680 was called a gtx 660...
yea it worried me too for second... but then I realized that he wouldn't have uploaded it if something bad happened... I think he just slightly choked on his own spiddle, but was able to overcome his cough reflex like a beast!!!... it happens to the best of us :)
If AI decompressed textures going to look like AI p*rn, we should not be pumped for it😅. I am waiting both Ryzen 9000 and Intel's Arrow lake, what gaming performance and pricing. But if Ryzen comes out at expected performance and prices, it won't beat 7000 series current price performance ratio.
AI dont exsist yet, it is just smart programing/algorytem, the co called AI is a program that have full acces to google and can find info super fast. But real AI a self thinking computer dont exsist as of 2024
5th time. Do not speak directly to your microphone. Speak to us. We hate our popsonants. Try 45 degree angel to your microphone. There are good quides on CZcams.
When you started swallowing your tongue (7:53), I slapped the side of my monitor and you straightened right out.
He usually takes so clearly I found myself wondering if he uses AI voice cloning or something, this was a surprise!
thats actually kinda funny
This is normal. It's called "rolling through a take." He just missed it when it came to editing it out. This happens quite a bit on channels that upload 1-2 times a day.
🤣🤣🤣🤣🤣
😨
I think AMD's biggest "problem" is their previous gens are so good I know a lot of folks that are just sticking with Zen 3 or 4 because they already have more CPU power than they got work for it to do. I know my 5700x is barely using half the threads most times, and my friends say the same of their 5800x3d and 7500f. These chips are running so well with so little heat none of us on our Discord feel a need to upgrade.
Sticking with my 5900x
Maybe upgrade my gpu and switch out my psu and aio to a new one since ive been running them for so long but im pretty much sticking with the 5900x until if the ryzen 10xxxX comes out and thats only if it has more cores and/or x3d on both ccds. I just want to make my money worth it at this point and probably just going to build a whole new system leaving this one whatever way it is from that point and use it for a server pc or something.
I agree with what you said about previous gens being so good. Many people are just satisfied with getting a Zen 3 CPUs. The extra bonuses from Zen 4 might be nice but many just don't want to spend the extra money when the prices for Zen 3 have been going down a lot.
I recently (back in late feb/early march) upgraded from an i7 4790K to a Ryzen 5950x.
It was a complete upgrade (minus GPU) as I got a new NVMe SSD (previous used SATA SSD), motherboard, CPU, new cooler (bought an AIO for the first time), case, and a PSU (upgrade from a 750 Watt to a 1,000 Watts).
I would have gone with Ryzen 7950x but the cost would have easily been like $400 to $600 more (with the parts I spec'd with it). It's such a major upgrade from what I had, I'll be fine for many years). By the time I upgrade again, AM6 will probably be out (way down the line) and I'd probably just get the latest AM5 at a major discount.
It was nice going from a CPU with 4 cores and 8 threads to having 16 cores and 32 threads. They will come in handy when I eventually do some video editing or even capturing gameplay.
I have a 5700x. I gotta upgrade to AM5 soon to keep going with my high end nvidia GPU buying addiction because soon I'll want an RTX 6090 and I don't think the 5700x will be all that great with a GPU that strong.
Whether R5K or R7K = 10-year life. mb
The main point of Threadripper isn't the speed per se - it's the PCIe lanes and applications that are optimized for massive core counts.
That, and the additional memory capacities and bandwidth.
7:50 Are you okay?
i think he had a stroke
That was funny. no worries gamer meld
Why did I think the video was buffering 😂😂
never eat the extra spicy jalapeno burrito before recording your stream...
high ping
Just what I wanted, AI hallucinations in my video game textures.
hahaha mb
Render resolution upscaling/downscaling, fxaa, and frame generation have left the chat
People start lookin up, why is this one wall talking to me. Or- why do my wheel nuts look like dogs noses in CP2077? GPU found digital dmt.
Nvidia will sell you 4gb cards as 16gb cards With BCT (Bendover compression technology)
😂 💯
BCT with DLSS on ULTRA PERFORMANCE
BCT at ultra super ti performance (144p) upscaled to 16k with the gpu costing 1 gigashit 😂
Respect for almost never shuddering btw
7:50 that's a stroke my dude get yourself checked out
Don't be an ass ! Just some stomach gasses !
It's smart, transistors are getting very difficult to decrease in size.
We have entered the architecture era of super focus to redesign and optimize hardware/software systems to continue the evolution for awhile.
I do believe that we will see some massive gains this way, but that will eventually lose steam too.
Not if we find a new transistor, femto-scale engineering kind. Or a very fast switching one, some have been demo'd at 20Ghz+.
Or optical computing, very low power use, extreme frequencies, like 100Ghz.
@@jamegumb7298 Yea I think I heard of the possibility of using molybdenum or something instead of silicon
@@jamegumb7298 optical computing is the future we can already put electrons into a phase state on silicon.
Optical is not faster. Its bigger and slower and use more power than today most advanced transistors. And they can run at 10 GHz max but same problém with electrical@@jamegumb7298
Until now, realism in games was a physical boundary but now it's an essential boundary. no more pixel perfect games.
loving the energy and content... cant wait to get my hands on the duel 3d v-cached CCD CPUs
I love your videos brother, and like some of my other favorite Tech CZcamsrs, you speak freely, like a normal person would and not like a person who is trying to sound professional like a lot of news and weather presenters lol. The thing I love about your channel, besides the content you bring, is the desire for everyone to have a great day at the end of your videos. I wish a lot more people did this. On that note, I say to you, Have a great day, and also be blessed, as you are a blessing to us. Keep it up brother ☺☺☺
So you're saying making architecture changes works out better than just dumping more power into a CPU.
yeah since right now we have reached a point where we are limited by the architecture. Its like building a castle on wood beams. A strong foundation will allow u to not only build that castle, but also expand it
Someone needs to inform intel about this lol.
Cough cough intel products
7:52 Cut! Do it again. This time do it with more feeling.
🤣
I'm more interested in AI assisted video compression. Get those video file sizes even smaller as long as it stays true to the original images.
3:48 yea... i really havent had a problem with "download huge game packages" since I was finally able to get off of dial up and get a cable modem.... 20 year ago
For most people the problem is running out of storage rather than downloading it I mean 1tb 5 years ago was enough to have a bunch of games now you can only have a few since everything tends to be 100gb or more which is ass
They are also generally games poorly optimized by the current regards of what these companies keep endlessly talking about as innovative nowadays.
So we see games take 400Gb and use 24GB v ram. When those same games optimized would use only 6 to 8 GB v ram, and roughly 100 to 200GB of space too.
Great informations! Thanks for your effort and work on this video! Greetings from Austria
but x3d when
september, clearly you have not watching the channel
Yes i saw it but if it'll truly happening then that is good.
@@AlizHalo damn wish I had all day to watch every single video
Have not been watching*
@@Qmohd1 it's ONE video, each day - Monday through Friday... you must be the busiest person on Earth not being able to find 20 mins per day... maybe you should give your wrist a break and stop trying to compete for the largest right-handed forearm.... you know you can PAY to lose your virginity, right? you don't have top spend your ENTIRE day fantasizing about your Mom.... smh
7:50 is me having any conversation
If they compress data files and textures, file size are lower and that means the entire game file size is smaller so you can download it faster since most new game files are close or over 100GB these days. If it can be implemented in all games then valve and other game launchers can use it to compress most if not all game files and that their entire servers have more space to work with. That also means if games also use less vram since the textures are compressed and can load extremely fast with better quality.
*Your time stamps are NOT working, as I said on previous video, add first time stamp as **00:00**, and not **0:00*
2:40 How come 4096x4096 uncompresed => 256 MB? Such resolution at 8 bit amounts to 50.33 MB and twice as much at 16 bit.
Sounds like to me the AM5 will live on until Zen 7 which is the 1100 series.
AMD, Intel and Nvidia need to squash their beefs and work with Apple, Google, Sony, Microsoft and Meta.
The sooner they make apps into physical and permanent products, like in the earlier stages of technology, the sooner we can make a WEB 3.0 and a multi-national citizenship system.
First, DX9 already support texture compression. What AMD & NVidia showed are just new compression techniques based on AI. The decoding of these AI compressed textures don't necessarily need AI to decompress so it can be easily implemented as a new compression mode OR with NO changes to runtime at all.
During the initial stages of compression, AI can be used to adjust the weightage of the compression parameters to maximise luminance & colour details to get the best image quality after standard JPEG decompression. (Search JPEG compression process if you have comsci background). I don't want to explain what's discrete cosine transform.
Below is my guess, I haven't read the papers:
For NVidia, it looks like they will do this automatically if your GPU don't have enough RAM. This reduces the need for texture streaming. It STILL means that lossless compression is used during downloading of game files so it's still 150-200GB for Microsoft Flight Sim 2020.
For AMD, they may be proposing to deliver lossy compressed textures during download but at a quality that is "Visually lossless". This will really help their console customers. For Consoles, the RAM is unified so texture streaming is NOT needed. Same for APU with Radeon 780/880/890M iGPUs.
Next, Intel Fab is currently working with ASML on the 18A (and better) High-NA EUV wafer machine, which TSMC is not currently involved in. TSMC's most advanced EUV machine from ASML can make wafers down to 3nm or 30A at similar yield. Check Asianometry website or YT channel for this topic.
What this means is, Intel CAN leapfrog TSMC for the 2nm/20A node and can downsize to even lower with the High-NA EUV process in 2-3 years. If high-NA EUV works...
Lastly, Intel Meteor/Raptor-lake don't support AVX512 so it loses badly to AMD Zen 4 & 5. This is very obvious when running the QUBIC mining client which tests different AI workloads and AVX512 powers the fastest algorithm. Zen 5 having a 2x front-end to instruction/data load really helps due to a wider data bus to/from cache.
Similarly, Zen 4 Threadripper supporting qual/octo-channel RAM can feed the FP units much faster than in a dual-channel Zen 5. Zen 5 based EYPC/Threadripper are going to be monsters!
Seems like you had to reboot at the end but forgot to remove from video
Sir Ryzen 9000 and 8000
Relesed date difference is 6 months only.
Any defects drawback flaws manufacturing error in this two types of processors .
does any body know how to apply these mitigations at the layman level especially inept game devs lol
I do 3d work and was tempted by threadripper but Ryzen often seems to catch up for a fraction of the price
I mean we overall need better APUs because we need them for consoles
Imagine amd releases a universal open source compression technique that makes 8gb gpus become viable again and everyone runs to buy 4060tis because of this release. Seems like a classic amd move.
5:30 it’s the contrary, AMD talk about the plan of intel to go ahead of Tsmc in term of process, amd is saying they’ll continue that path and taking the risk to be behind in term of process if intel makes it. The backside power delivery and all this crazy bet of intel.
TSMC will use backside power too.
I'm not exactly excited about Intel falling so far behind AMD in the CPU space. It's better for all of us when they're neck-and-neck. More competition means more options and better prices. It's also becoming more dangerous for Taiwan to be doing so well in the compute space for so long. China keeps edging closer and closer to invading and having control over cutting-edge technology (especially in AI) is the biggest motivating factor at this point.
So, you are saying we finally got 'middle-out' compression?
I found ryzen 9 3900 for 180euro and ryzen 9 5900 for 230eu but i will go with 3900 because i dont want to upgrade bios or to buy a new motherboard.
7:52 Dude was about to throw out, he controlled it & swallowed everything, what a profressionnal 👍🤣🤣🤣
hahaha bro you recorded the video you could have cut that out thats funny AF! 7:50
Good thing that CPU size could change. If a transistor does 3nm you can have a quantum effect leading to signal going throughout the other transistor
Didn't TSMC state they had solved the issue with Quantum Gate Tunneling several years ago? I know they have made breakthroughs to even smaller transistors than 3nm.
There Is roadmap to 2040 with 0.2nm transistors @@MadManUSMC
I'm more excited about the compression tech. :)
Great roundup once again, thanks.
So a better Frame Generation with AI. Better compression is also good, 'coz every single data type is getting more resolutions and more of them; maybe it could help us generate better compression for lossless images or improve the current image compression formats
Bro literally burped his train of thought out 😅
Who is noticing a 2mm scratch detail on a lid... There's very few reasons for 4k textures on any object...
Nature photographers crop the hell out of their shots - every pixel counts. There are more things to life than gaming.
Now nvidia and amd will add a dedicated texture decoder in their gpu and increase the prices by 200$ 😂
Super excited for AMD!
Both Technics from AMD and Nvidia IF THE RESULTS LOOKS WORSE THAN THE ORIGINAL THAN DONT BOTHER
So GPUs needs to be more powerful now to decompress/upscale textures instead of just having more memory space.... I wonder if memory is cheaper or the processing unit that does this upscale/decompress thing.
Might be memory bandwidth capacity. Only so many attachment points for wires etc…. Rather than update how circuit boards are printed you can send compressed textures across pins and the core can quickly decompress them.
@@TryHardNewsletter might be, but the final point is, will it be more expensive and just an excuse to raise the price in the future
Moving from 1024x1024 to 4096x4096 within the same file size is LITERALLY "sixteen times the detail"!
i think this was my favorite episode ... and just like that...i was gone....
Nvidia started to convert parts of gpu hardware to software when they introduced dlss and frame generation. By addind texture compression and real time model and environment generators they are planning to reduce the need for hardware even further. Soon NVidia gpus will become pure software to download from their site.
Definitely allows for programmable GPUs. Been a long time since I said anything like that.
If the 16 core Granite Ridge is that far ahead of Zen 4, then a32-core Zen 5 Threadripper is going to rock.
ive noticed you having the stuttering a few other times, i seriously hope your ok, please get medical help if you need it
But see, game graphics are so good at this point that we don't need to improve them. Especially after UE5.3 or whatever. A 4090 already covers Cyberpunk with 4+ rays 4+ bounces modded path tracing and 8k textures, at 8k native render resolution. I KNOW this will be used to push the envelope further and games will still be 100-200gb, they'll just be able to brag about the now even better graphics or even bigger worlds (that one may be valid, but only if the worlds are detailed and don't feel empty).
Just keep games how they look now (or will/would in UE5.3 with 4R/4B path tracing as the practical max), with hardware as powerful as it is now, and focus on making the existing stuff more efficient in terms of power consumption, heat, size, and data storage, then add features (like AMD improving at ray/path tracing). And also, we don't need this compression to be 67 to 1. 10 to 1 or 20 to 1 is fine, and will likely improve the appearance more, too because that example compared to the reference is garbage.
It's like TAA and "quality" upscale rendering vs 4x MSAA rendering at native resolution.
AVX-512 benchmarks with 2X wider execution of AVX-512 instructions show much greater performance. Yes. It gives impressive results.
What is this new company azeus? Did Asus try a rebrand to get a better reputation?
Considering that my 3700X is barely able to keep up in new games, I think I'm liking my odds when Ryzen 9000 comes out when upgrading :DDDDDD
let's not do that intro move please. Great video and as always great tech info. Till the next one, Enjoy.
Both
The texture thing won't help low memory gpus, games will use as much they can like now, it will be the same but with better textures.
AM5 DDR5 ECC support is dead?
Stole that intro movement straight from Graphically Challenged lmao
Strix halo is something big
We going in reverse with technology...we suppose create mini chips but instead we do bigger.
All these people watch these videos everytime he uploads a new one. I have for awhile, but I have noticed most you comment then just cry and complain. You must not understand what your watching. I don't respect people who's entire life is just constant complaints.
7:50 editing fail
Why is it compared to the 13900k instead of the 14900k?
they are essentially the SAME chip... they just selected all the best binned 13900K's and upped the power and speed by a small amount... lol ...
- nah it's actually a new chip but nearly identical to 13th Gen... they kept everything the same except for I think the 14700K by adding 2 more 'big' cores.... you MIGHT get a 14th gen chip that's always 1% or 2% faster than the 13th gen equivalent, but that's still within the margin of error, so pretty much all the big names on youtube and the big websites ALL just consider them the same chips when it comes to RAW performance, so it honestly wouldn't change anything if they did use a 14th gen :)
@@Nobe_Oddy That makes sense. Thx!
I'm STOKED for this new compression tech by team green AMND red!!! ( and you KNOW Intel will have their own version soon too lol) I hope to learn more about the AMD tech, since you weren't able to cover much detail on it, but Nvidia's tech looks to me outstanding! As much as I cannot STAND Nvidia's management and the way they treat their customers AND partners, I will give them credit where it's due and this stuff looks to me AMAZING! But like ALWAYTS you'll probably need to have one of their newer cards to use it, so as ALWAYS I'll be using team Red because I support what little bit of open-source/open-ness that they have and share... So I will support the the company... "team" that aligns more with where and how I want to see the industry to head in.... and even tho they have been making MASSIVELY DUMB DECISIONS as of late, they're still the same team at the core and I think they're a GREAT balance to Nvidia's nauseating dribble about themselves..... I just wish AMD would think more about doing their OWN THING and not copying EVERYTHING nvidia does.... it's actually kinda pathetic.... who making that decision anyway??? :(
- - Oh but I am SUPER EXCITED for Zen 5!!! I'm still using a 5900X and feel a bit left out being on DDR4 and pci Gen4 ... but I'm sure I can hold out a few more generations before my baby starts to feel slow lol.... (obvious I'm being sarcastic... I don't even need anything this powerful, let alone an upgrade... I most just watch youtube, some movies or shows, and a bit of gaming occasionally....so upgrading would just be about vanity at this point lol :D) But I'm actually VERY EXCITED for it! Progress is awesome and I LOVE seeing it up and close everyday and hearing all the cool things about before its released... and that is EXACTLY why I'm subbed here @ Gamer Meld!!! :) Thanks for everything you do bro :)
cpu!!!
This AI crap is awesome, but it also means you'll own nothing.
That is good news, it is encouraging, but a) that's early silicon, and b) you don't know what clock speeds that engineering sample was running, nor what the other systems being compared were running. In short ~ not going to say that's meaningless, but it's only a pre-release leak to tease us. It doesn't actually mean anything unless we have a whole lot more detail.
For one thing, we know that AMD have made changes to the AVX512 pipeline, such that they don't have to break those ops up into 2 parts. It now runs them through in one piece. That has (I understand) more or less doubled AVX512 performance. Let's say a 75 ~ 80% increase. Ok, that's good, so if you have a benchmark that exclusively tests AVX512 performance, you'd expect a Zen 5 chip to have nearly double the performance of an earlier AMD chip. And that's what we see. Ok ~ so does that mean it will deliver an 80% jump in frame rates?
No ~ that's not what it means.
Is it going to help frame rates?
Yes. Sure it will.
By how much?
Remains to be seen. And it depends on how the game (or other job / task) was written. If it makes big use of big floating point numbers and the AVX512 pipelines, then expect to see impressive gains. If it never goes near those instructions ~
Then there's questions about speed and timing and latency and test-data and ~ let's wait for independent 3rd party reviews to land.
I am excited, but let's not get crazy about this.
Linux Mint 22 is due to land (the Beta / test) some time soon. I'm excited about that too.
he think: what does it meant to be slower??? ehhh....?
Both images are not similar. Yes, the compressed file looks good, but there is obviously a significant loss in image quality. Impressive compression? Yes. Comparable quality? Hell no.
Some things we just don't want compressed though. The compressed and uncompressed images are clearly different to the human eye and may lower the overall visual experience if one cares about that. It may be useful for cases where a person does not care, meticulously, about the visual experience, but knowing nvidia, they would probably try to force it on people just to see higher FPS across all their GPUs across all games 😐
Depends on, if it is lossless, or not.
@Theworthsearcher that's fair to be honest.
If you mean the DLSS, I would say, you are right. And nvidia fans just save nvidia. But rtx as well... Really? You have to turn on dlss for it? 😲 That is just... Maybe, 6090 will give us 120 FPS with rtx/path tracing at 1080p without upscaling. 😞
...thanx dude!
Skynet is taking forever to spawn!!
"A.I." in my GPU, no thanks...
AI is already in your GPU, if you have RTX
Yeah I was coming here to comment that you probably have tensor cores unless you’re on AMD
@quintrapnell3605 and even then, AMD still uses AI just at the hardware level. I've been enjoying driver level frame generation and FSR for unsupported games on my 7900 XTX
Nothing wrong with Ai Don't be edgy
@@nesseihtgnay9419IA is already on most 2005 PCs. Even more on GPU Nvidia or AMD use IA. Every game uses IA. I know hearing gaming everywhere and IA everywhere can be annoying but it's better for us.
Thanks for your informative videos as always.
2:00 Oh good, Nvidia can now sell us a 100mm² 64 bit 30 class card and call it a 5070TI Ultra thanks to this memory compression.
Nah they'll just cut back the VRAM again.
Finally someone that understands that die size matters.
Also we already have the rtx 4060 being a 30/40 class GPU since it's 147mm^2.
Keep in mind the reticle limit is higher than ever so 147mm^2 is like 80-90mm^2 on 28nm.
@@badass6300 I thought the 4060 was 159mm² and 4060TI was 190mm²
The 1650 is 200mm² with the same 128 bit bus as the 4060TI 16GB
(FYI you can hold alt and press 0178 on the number pad for ² 0179 fpr ³ but make sure numlock is on, otherwise you'll reload or change the web page)
@@denvera1g1 Yes, it's 159mm² my bad, still that's a 40 class chip. The rtx 3050 6GB is 200mm², the RTX 3050 8GB is a cut down chip from 276mm^2 down to 224mm^2, the gtx 1650 is 200mm^2.
Only the gtx 1000 series which were just as scummy, if not more than the rtx 4000 series have small die sizes. For example the gtx 1080 is 314mm² which is straight up a 60 class card, and should have costed 350$, maybe 400$ due to lack of competition from AMD, instead of 600-(700)$ (FE)MSRP...
The gtx 950 is also a cut down 960, but still comes out at 176mm² out of 226mm²
Of course the gtx 600 series the original bs generation, where a 3-4 weeks before launch the gtx 680 was called a gtx 660...
If you look at those compressed pictures, they come out looking like; 6 6 6.
nvidia just got a reason for giving 8gb vram 5060 too
And I’ve got a reason not to buy it.
No on at 60 class GPU we need 12GB VRAM.
7:51 nice belching.
it was acid reflux after he saw a pic of your mom ;)
burn LOL
@@Nobe_Oddy oh how funny im dying :)))))))
@@Adreno23421 huh?
@@gamingcomputers7485 Nobe Oddy`s comment was a sick burn!
Don't rely on benchmarks run by nvidia or amd itself. Wait for 3rd party to verify.
AMD Zen 3 cpus still to Good to look to upgrade just yet...lets watch/wait until early 2026
could u not have edited 7:50 out lol
I like true content
yea it worried me too for second... but then I realized that he wouldn't have uploaded it if something bad happened... I think he just slightly choked on his own spiddle, but was able to overcome his cough reflex like a beast!!!... it happens to the best of us :)
he's keeping it real.
Stagnation of cpus..nows its softwares turn to improve greatly
New Tech only makes my budget smaller(
7:50 Dude, if this is happening frequently, I think you need to visit a Doctor. We're concerned for you.
If AI decompressed textures going to look like AI p*rn, we should not be pumped for it😅.
I am waiting both Ryzen 9000 and Intel's Arrow lake, what gaming performance and pricing.
But if Ryzen comes out at expected performance and prices, it won't beat 7000 series current price performance ratio.
Your Favorite Number: Nannniiiinaannnnn....
I am that AMD fan boy everyone makes memes about :D
Is nVidia coming up with this textures trick just to keep selling their GPUs with low amout of VRAM 😂 ?
AI dont exsist yet, it is just smart programing/algorytem, the co called AI is a program that have full acces to google and can find info super fast. But real AI a self thinking computer dont exsist as of 2024
It looks like he almost collapsed Huehuehue
AMD's got a Moore's Law revival lol
A-Zeus? :P
BUUUHUUUUUUUT, this is your favorite word lmao drinking game for it
Oh great, we're going to get another upscaling technology . Here come the 5060 4gb models 😂
Remember folks do not buy the 5060 if it doesn’t have 12GB VRAM on a 192 bit memory bus.
5th time. Do not speak directly to your microphone. Speak to us. We hate our popsonants. Try 45 degree angel to your microphone. There are good quides on CZcams.
Gamermeld leaks be like: RYZEN 20000 IS LEAKED
Dude are you okay?
Asus: eyzuz