AMD’s STEPPING UP!
Вставка
- Опубліковано 10 лип 2024
- AMD's Ryzen 9000 is getting a new feature, Intel's in a lot of trouble, you can drop in more VRAM to a GPU and RTX 4090 "Super"! Stay tuned...
***Items featured in this video available here***
►Newegg (Affiliate): geni.us/newegg1
►Amazon US (Affiliate): amzn.to/3b9UjKB
Join The Discord: / discord
Twitter: / gamermeld
Facebook: Facebook: / gamermeld
Gamer Meld Merch: teespring.com/stores/gamermeld
Gamer Meld Sponsors: www.gamermeld.com/sponsors
Support Me On Patreon: / gamermeld
SOURCES:
www.techspot.com/news/96261-n...
wccftech.com/teclab-geforce-r...
x.com/1usmus/status/180808181...
www.tomshardware.com/pc-compo...
www.tomshardware.com/pc-compo...
• AMD Zen 6 Medusa Early...
TIMESTAMPS
0:00 RTX 4090 "Super"
2:25 AMD Adds Huge Upgrade
4:30 Extend GPU VRAM
6:25 Intel's In Trouble - Наука та технологія
I moved from Tampa to Santa Clara a few months ago and I’m thinking of purchasing a single family home there, but with real estate prices currently through the roof, is it still a good idea to buy a home or should I invest in stocks for now and just wait for a housing market correction? Looks like NVDA, TSM and AMD and AVGO are strong buys this week.
it’s a personal decision, but according to Forbes, housing activities will remain stagnant for the most part of the year, so maybe hold off a little.
well you could put a downpayment on a home and as well diversify as much as you can into Ai, energy and big pharm. stocks like Pfizer and JnJ, ASML, MLM and S&P 500 ETFs. I’d suggest you go with a managed portfolio, but even those don’t perform so well, so it’s best you reach out to a proper fiduciary to guide you, that’s what works for my spouse and I making a whooping $738k in Q4 last year
I’m an art collector, this is not very new to me but has a nuance to it. Can you assist me?
There are many advisors to choose from. But I work with 'Leah Foster Alderman' and we've been working together for years and she's fantastic. You could check her if she meets your requirements. Just research the name. You’d find necessary details to work with
Thanks for the recommendation, I just googled her and I'm really impressed with her credentials, I reached out to her since I need all the assistance I can get.
I'm here for the news. Either presentation style is fine. Do what makes sense when it makes sense.
Same. I don't follow this channel *nearly* enough to care about how each video is presented. All that matters to me is staying in the know when it comes to major advancements in the PC hardware realm. So far the YT algorithm has been doing a good job at reminding me that this channel exists when such advancements happen, which means that Mr. Gamer Meld is doing a good job. Keep it up.
yeah, I actually prefer the talking head videos. someone to look at while hearing about the news.
I don't min the "talking head" videos; I treat them as a podcast and listen while I'm doing other stuff.
same. surprised it makes me more productive for my adhd ass
Id love some "monster" APUs for desktop, the 780m is nice but desktops have more room for additional power and cooling.
I think APUs also have potential in the budget section, as having both in one can be a big cost saver.
I would love to make a steamos based mini pc of those that is below 700$ that would coooooook@@Posh_Quack
Instead, but a iGPU less CPU and throw in a GPU. APUs on desktop are a total waste.
Learn how to overclock.. My 8700G at 5.4Ghz allcore with 7400 DDR5/ 2467 Fabric, and 780m at 3200mhz I’m getting over 100 fps in Warzone 1080p without upscaling..
@@einstien2409it's not that easy
stop appologizing for "talking head" video's. It does not bother me. Getting the info from you is what I want. But every video appologizing for something I do not care about is anoying.
Bootlick
*annoying
He has been doing it for years. It is more infuriating him apologizing all the time! If people had a problem with they would of been gone a long time ago!?
Just do what's easiest for you to get the videos out.. and instead of apologies you should just say "thank you"
@@anubis1800 That, what you just said. That would make the most sense!
A mix is good :) Need to get that info ASAP sometimes !
That expandable VRAM for local LLMs is going to be incredible.
I'm think the memory is too slow for that
Currently, when using the CPU, it is not the CPU that is limiting you in the slightest, but the RAM is much too slow
@@Psycho-Duckyup, you can see this by tightening the DDR5 timings and than see 10+% performance gains in framerates on vanilla Zen4
@@Psycho-Duckhopefully it is, or it becomes another Nivida gpu shortage. Unless they make it so it can only apply to gaming, so big corps don’t buy vram expansions and 60 or 70 series cards for LLM causing the gpu prices to possibly soar.
@@Psycho-Duck this is my question, just because it's available doesn't mean it's fast enought to keep up. But I'm guessing here.
Honestly I love your talking head videos man, I like being able to put a voice to a face. Makes it feel waaaaay less robotic! So mate, I'm proud of you doing whatever you like but I prefer your talking head videos and if you personally like the other format... throw that in from time to time too. = ) Much love from Australia!
PS: Wanted to add when you do the mix of the two with content and your video in the corner is a great way to show things I personally thought. So like a mashup of the two
A 6-8 core with decent apu will win a lot of more budget limited fans.
There's about 6bn people who're never going to afford a new gpu
6-8 core with decent apu? hat should it mean? 😲 Did you mean, an AOU, which has 6-8 CPU Cores ang a good IGP with 16-24 CUs (Compute Units)?
@@Theworthsearcher strix halo is rumored to have 40!!! CU's that's about ~4060 level of performance from an APU.
@@TheworthsearcherAOU? I've heard of APU, CPU, GPU etc. Never heard an AOU. Can you explain it please. Thanks.
@@mohd5rose APU typically means cpu with igpu (integrated graphics) but will soon include NPU like tensor cores or other types aux processing units. In two intel generations, when they're integrating battlemage level APU/GPU cores with CPU cores they'll be drawing down on half of todays Nvidia's market and 3/4 of AMDs. The only chance those two have to retain their market share against intel is to rapidly scale performance per watt improvements and robustly implement new code support for "ai" type processing. Intel is going to need 600 watts just for the "APU/NPU/CPU/iGPU" processor.
@@kazadori164Strix Halo’s 40 RDNA 3.5 CUs will more like a 4070 than a 4060. RDNA 3.5 has vastly improved RT than RDNA 3!
IMO the presentation style should go to efficiency and clarity; whatever is best at the moment, so a mix of different styles, but showing the proof of everything is always a must.
Pumped for Ryzen 11000 series, a true redesign, and probably the last full redesign, as Silicon is getting to it's Physical limit; which reminds me other Elements combos were being tested, and also Photonics
there isn't a physical limit, just the limit of the time it's being made. I'm sure people said something similar 20+ years ago too
@@lassebrustad Really? How do you know there are no Physical Limits?
@@saultube44 just look at the history, no limit within tech has kept up, and CPUs are still improving a lot, which is a sign that there's still nowhere close to any real limits, it's all just the limits for these days, and in 10+ years, todays limits will look like crap
anyway, why pumped for Ryzen 11000 series, if you're so sure about CPUs being close to the limits, when you're also claiming a true redesign, which typically includes pushing the current limits even further?
@@lassebrustad No, you have to proof, now, that there's no limit; I'm not asking for a re-comment on it, which is only your opinion and no Science about it.
About Ryzen 11000: I can't believe you made such a dumb question; but this proves how you deeply ignorant you're about the subject, and just trying to turn my words against me, kid; so why don't you Google your ignorance, instead of displaying it, in front of The Internet?
@@lassebrustad transistors can only get so small before they reach the size of an atom so there is a physical limit however i think it will still take a while untill we reach it (1 silicon atom = about 0.1 nm iirc)
RTX 4090 supers would probably get so hot it would visibly turn red, the only way to cool the beast is by sticking 3 industrial grade fans and put it on the highest setting for cold intake air and fast exhaust lol
Can't wait for those power connectors 😂
Blackwell AI GPU will dissipate 1kW & be liquid cooled exclusively! RTX 5090 will probably be the same!
I've said it for a long time now, they need to add M.2 slots on every GPU with DMA and Directstorage. I would love to add a PCIe5.0 4TB drive to my GPU, like a L3 cache.
Thanks for sharing!
Best wishes
Being able to add vram to your GPU could literally extend the life of a GPU for quite some years.
Imagine a 4060 Ti, it will be enough for 1080p for quite a while if you don't care about max graphics on every game, but at this point some games will need more than 8GB even in 1080p.
With that you'd be able to spend way less than a new GPU would cost you and perhaps even upgrade to 1440p and still being able to get playable frame rates.
That's also why I don't think it will be adopted, imagine helping your customers needing to buy your product less times...
I am TOTALLY FINE with whatever style you choose (talking head or voice over) ... I want you to make the video that WORKS BEST FOR YOU! :) If one way is easier for you to make then by all means go for it!! It really doesn't matter to me TBH....
There IS ONE THING tho... there were like 2 or 3 times where you did a talking head video and must have done something different with the audio setup (I think you were standing and it was just during the ad-read... but this was a while ago so I don't remember ALL the details :/ ) Well there was a pretty bad reverb/echo on your voice and it was super noticeable and pretty bad :( - BUT it was only during the ad read, so it was pretty short and was back to normal in less than a minute :) So I think because you were standing up you were projecting your voice much more than usual and so the sound was just bouncing off of everything lol
Like I said, it was only a couple times this happened and it was a while back... SO I'm pretty sure you know about it.. but if not then now you know to look out for it :)
these vids are fine. you get to the point, no fluff. love it. I do think you could rasp up your voice a bit though it's almost disturbingly smooth
3:23 this makes it to where your ryzens CPU's dont ramp up to max frequency/power when doing semi CPU intensive tasks. Have you played a game and noticed your CPU shoot to 90c almost immediately? this sort of in theory fixes that. There was always some weirdness going on with certain old games and ryzen CPU's where the games did not like ryzens architecture (if im not mistaken)
We love your normal self.
Gamer Meld Bro, please promise me, you will take a vacation anywhere soon. You pump out so much content on a yearly basis, you deserve some rest.
Nvidia gatekeeping performance on the 4090 is a very Nvidia thing to do. Is anyone really surprised ?
"new technology" by Pamnesia - AMD offered expanded memory via something similar to an SSD as early as the datacenter counterpart of the Fury.
I wonder if they thought about stacking cache on gpu or to vram kinda like they did with x3d chips. Maybe get some speed bumps. Maybe intigrate it all into one chip.
These vids are great!
Great content and happy to with how ever you do your content. Love the DBZ hat 👍🏼
I prefer a mix. I feel you've never needed to apologize for the usual style not being present.
How would they keep the power connectors on a "4090 Super" from melting, given it's going to eat even MORE power than the 4090 which already has "melting power connectors" issues?
AMD needs to do this Curve Shaper to their GPUs as well. When I had the 7900xtx it could take 1.070v undervolt at max clocks (100% load) 24 7. But at medium loads 2ghz-2.3ghz I crash in games, so I have to give it 1.130v to be stable all around and it's just a loss in efficiency doing so
I hope Strix halo has an 8 core full gpu variant. I’m worried they are gonna only make the full gpu for the 16 core and I couldn’t afford that
Wow, I need to learn more about this-Superposition benchmark. I know Heaven and Valley but this has gotta support DX12 and Vulkan. If possible, I wanna test it on my A12 9800.
I was surprised when I first saw you on camera but this way nobody can say you are a robot! Keep up the good work. Feel free to do either or a mix of both.
CXL will be a transformative tech, cache coherence bus. This would be a great use for it.
Don't apologise my man I like your videos despite getting these news usually earlier from other sources, but your delivery is good.
Talking Heads ..... We're on a road to nowhere. I thoroughly enjoy the talking heads, so I am quite happy with Gamer Meld becoming one as well, as I do the same that @JamieStuff below does. Have a GREAT Day!!
I say pick which ever video style you like doing the best. I can only speak for myself but the video delivery style comes second to the actual info you provide us with!
Adding 8/16 GB of super fast pcie 4.0 / 5.0 ssd/ram to be used as vram is totally viable
dawg, you as the daily news is lit
Your variety of video styles is fine. Thanks for everything you’re doing! 😎🤙
Lol curve shaper, They have just done what my modded version of Ryzen SMU tool does. Since I can have really high offsets when the CPU is hot but it would crash at low temps from trying to boost to high so It switches between 3 CO profiles based on below 30C, 30-60C, 60C+
to be fair, i enjoy your overview videos such as this at 1.75x speed... content and voice quality are more important to me than any particular visual video style.
while i'm in the industry myself and keep a pretty tight finger on hw, i enjoy your perspective as it is substantially more mainstream/vox pop than the stuff i dig into personally, thus helping me round out my perspective and not lose touch with with the people who are the majority of the sector by being way too far into the details.
the presentation style that makes the most sense, is the best! nobody complains, so do stuff like you want to. your videos are the perfect videos to literally be 100% up-to-date anyway, so well that there's no suprises like "shit, I just bought a CPU last month and now there's a new and better generation out" 😄
the new youtube layout swaps video's and comments locations from the old layout, going to result in less engagement, hides the whole comment section in theater mode.
Dropping in VRAM into your GPU is a consumer AI Game changer. I have a 4090 and 24GB is NOT ENOUGH. Forget about ChatGPT, you can run UN-BIAS LLM's on consumer cards...
RTX 4090 super, now with 25Gb of VRAM
wow this 5x3 power curve setting is going to be very interesting.
I am hoping the extended VRAM capacity using Panmnesia CXL memory expanders technology is developed - that would be fantastic for folks like me who have an RTX 3080 10GB or basically anyone with a GPU under 16GB >> keep up the good work 👌💯👀🎯😎🌟
Just one suggestion: sometimes if talking about numbers, such as comparing two speeds, its useful to have a graphic showing those numbers together so the information stands out clearly.
I actually think you could make some money doing voice-work, also consider looking up dubbing courses etc. - as they can teach you a ton of tricks, old school acting could also, but as you already have a very clear and articulate voice, then it perhaps might not teach you as much (thou it's great fun)
Why should GPU use NVME for capacity increase when it can access RAM and have 10x more speed for swapping VRAM?
Volume, I suppose. An optane would work quite well there, I suppose, if speed is a concern. Also, in some use cases such as llms volume is perhaps way more important than speed.
I think you have a good mix of "talking head" and showing what you're talking about, it works 👍
Vram addition ssd would be a blessing for many gamers!!
4:08 so that means when idle you can downclock and undervolt and while under load you can overclock
Most modern CPUs pull less than 20w at idle.. If you’re worried about saving money at 20w you don’t need to be playing on a PC.
@@LegionGamingTV i dont have a super modern cpu.
They should make it so that 2 vid cards can use the ssd PCIE slots and a link cable such that the vid cards can share their VRAM. sLI like functionalities are problematic for various reasons related to syncing frames or scanlines and framebuffers, but using the secondary vid cards vram as a sort of buffered memory wouldnt have the same issues; it should be faster than transfering down to the RAM, at any rate, and certainly much faster than dumping from and to disk. Past that, the transfer could take place outside the PCIE bus/lanes, which is helpful for high IO tasks (blockchain mining and the like), though normally video games dont need anywhere near the full capability of PCIE 16x within their generation. Even if they end up being that intensive, offering a secondary transfer path would be helpful (think dumping data half and half to each vid card, and then keeping it between the vid cards after that). Also perhaps for various calculations, such as post 3d render effects, the required data could be copied to the second vid card as soon as it is available, which then does the calculation, and sends back the result to the primary vid card. If they primary vid card finishes the 3d state before said results are in, it could just do the calculations itself as the prequisite data would still be there, and post the frame if it finishes, but if the 2nd vid card finishes up before the 1st is done, the first could just abort and present the frame. At the very least, this would reduce frame latency, if not be a major boost to fps.
I just want to get use out of my second vid card damnit... it just sits there hoping somehow there is some physx to do, but not many games use that
I don t much care about the clip format. I m here for the aggregated news and comments on it. thanks and keep it up
4:40 - Why can't GPUs use system memory? You know RAM? Your latency will be 50ms.
I was used to no talking head. The talking head feels like lip syncing
I like all of your videos and actually didn’t even realized what changed… so everything’s no is fine with me, your videos are still great
Mix of video styles is always good 👍🏼
Budget APUs should replace the mid to low end GPUs IMO and if you want top notch graphics you pay for hte top end dGPU
Mix it up! Keeps things interesting and likely nothing will become monotonous that way. Enjoy your viewpoints and the news you bring is definitely interesting!
I love your "face for radio" man... Keep the talking head videos. And I agree with some of the comments... STOP APOLOGIZING! Just don't have an aneurysm in the middle of the video again and it should be fine!
Addons to gpus we did already back in the days hehe.. So this is coming back ;).. late 90s that is. Some of you were not alive back then but yes.
Imagine 4090 with 128 GBs of VRAM 😍
That’ll be many years from now. 😂
Why is none of the pc sellers offering AMD ? I just checked the major companies. It’s just weird that intel is the only thing offered right now
( I really only want AMD )
for the average consumer, they know intel and rely on that
Build your own computer. It isn’t _that_ hard. Just *don’t* follow The Verge’s shitty PC Build Guide and you should be fine.
That 4090 looks interesting. But 3 GHz isnt that much for OC. Or am I wrong? I have a TUF 4090 OC which clocks just slightly above 3000 MHz when overclocked on its default vBIOS. With the Galax 666w vBIOS (And therefore with 1.1v), it does 3100 to 3150 MHz. Thats with a simple waterblock on it. Without water, I lose around 50 MHz.
Ironically I found your channel only recently and I enjoy the talking head videos and now I’m curious what you did previous that you think is better??
he had a script and the videos were planned out. i personally dont really notice a big difference tho so i dont really care what style he does
@@bal7ha2ar yeah I scrolled back a fair ways and watched the older videos and honestly I enjoy the talking head style more and more casual script. I find it informative while not having a audiobook textbook vibe but more like a casual conversation with someone I might know who does all the talking (wow as I wrote that it was so specific and I can’t tell if it made sense or if my lack of sleep is affecting my cognition really bad).
@@masterinico it made sense dont worry :)
Strix halo on desktop? I guess we need 2 128bit camm RAM slots
I like the talking-head videos. I like to see how enthusiastic you are when talking.
Mix is good :)
So with this performance boost we the customer should not buy a 5090 unless there's a 30% increase or more in performance . That's what this video is telling me.
Either video is fine for me, theyre informative and well presented
Intel is dead before launch
Dumb
@@thetheoryguy5544 cope intel meat rider. cope...
Been dead since 14th launch.
Hoping qualcomm keeps amd sane
Intel going lose billions pretty soon their making a chip plant I think what it is. In a very hot area in a country.
intel makes more money than crappy AMD so how is that explain? Oh forgot you can't.
I don't know about other people who have a 4090, but I don't come close to utilizing any of of the power even playing AAA games 🤷
Info is info whether you are a talking head or not. I liked the 'podcast' comment below. Keep up the good work!
Please do "Normal Video" back in days
NVidea won't like the expanded gpu ram kekw
nice video, keep it up
5:32 Nvidia is not going to like this. Jensen said , " The more you buy, the more you save ". He wants you to buy more Nvidia GPU . More specifically, buy more expensive ones from them😄
For me , more video is better. Short is not a problem at all 😊
I like any video from you, so talking heads are good enough for me and many others.
Amd radeon pro ssg?
wait for ryzen 12000 series
Presentation style does NOT MATTER
iGPU is the ultimate secret weapon for home lab AI. The only problem right now is that they aren't really powerful enough. That's why The M-chip is selling like crazy, because it is kind of powerful enough to train large models that need 100+GB of memory. Which dGPU can't really offer, unless you have like 4+ of the highest end card...... Intel and AMD need to realize that. So True AI for desktop can really happen. NPC or whatever simply can't make the cut.
Waiting for the 9000 to drop si that the 7800x3d is at cheaper price
9000x3d could be worth it if the performence is better than the 7800x3d
i vote for a mix style, not all talking head.. not all screen record
I really don't care about the presentation, i personally grateful into having a condense news source of tech on the daily basis.
Intel isn't worried in terms of fastest peformace CPU's for consumer platforms. Panther Lake and Beast Lake are just around the corner. Intel Xeon cannot yet match EPYC and Threadripper, Intel needs to seriously take Xeon to the next level, which Sierra Forrest aims to do with a 288 E-core workstation processor. Not that Zen 5 isn't exceptionally impressive as it most certainly is. Intel needs to take notes from AMD extensively on a faster CPU without needing a higher clock speed or power draw. AMD and Apple are crushing Intel on power efficiency. Arrow Lake and Lunar Lake are Intel's attempts to prioritize power efficiency.
Intel knows this, Pat Gelsinger knows this, and they have demonstrated in 2024 a dedication to change. Meteor Lake accomplished the same goal. AMD and Intel IMO are now truly neck and neck. NVIDIA is still in a league of their own obviously, Huang Jensen is untouchable.
Intel is nearly a gen change behind zen 5, hoping they get better. Power efficency and multithreading are being killed by intel rn, don't want just 1 competitor
I haven't noticed any difference in the video style.
I'm not going to be much help in deciding which content to make, because I've always enjoyed the content from this channel that I've managed to catch. So I'll simply say please continue to keep making videos when possible, they are enjoyed & appreciated.
Thank you for making something for AI. I don't want to go bankrupt for more vram.
NVidia's "boost clock" might be the single most meaningless thing ever.
I'm interested in what you're working on - so talking head videos are fine. :-)
we shouldn't be forced to find a way to cram more vram into our system. it's the developers job to optimize their games.
Intel here , exe file lives forever I promise
Bro I love the so called talking head videos but I will support either way.
I'm all in on ryzen 99999999x
zen6 might be in the ps6?
Based off of its heavily cutdown L3 cache and lower clock speed it will be equivalent to a zen 4 Ryzen 7 7700 if it is a zen 6 CPU.
I love everything you do
There's a lot of heavy lifting in the AMD Medusa details calling out Intel. Essentially, "AMD will continue to improve their architecture, and TSMC will move to a smaller node at a larger scale." Not a very sexy, eye grabbing title, considering Intel will also be improving their architecture and moving to a smaller node.
Arrow Lake will be Intel's deciding factor for the next 1-3 years of Intel desktop trajectory, as Lunar Lake will be for their mobile side, and whatever comes after that will be essentially their final shot in the desktop and mobile markets. If the 20A (2nm) process, and follow-up 18A (1.8nm) and 14A (1.4nm) process nodes yield even moderately well, TSMC will actually have a serious fab competitor, and one that isn't under constant Chinese threat of non-existence. I want them all to succeed, simply because I want all of these mega-corporations to make the best products they can at the cheapest prices they can so I can get more frames for less dollars.
Finally that adding Ram is real