Ryzen 9000G Will CANCEL GPUs!

Next Gen AMD Ryzen APUs are going to make GPUs obsolete, I told you these were coming to the US, Intel CPUs have a serious issue and melting 4090 connectors are even worse than we thought! 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:
videocardz.com/newz/repair-sh...
www.techpowerup.com/321461/nv...
www.tomshardware.com/pc-compo...
www.nvidia.com/en-us/geforce/...
videocardz.com/newz/amd-offic...
wccftech.com/amd-strix-point-...
TIMESTAMPS
0:00 Melting 4090s Are Worse Than We Thought
3:20 Intel CPUs Have A Serious Issue
6:08 They're Coming To The US
8:06 Forget GPUs

[@chauto5987]

"NO MORE GPUs Needed" is the term I hear every 1 or 2 year when new AP come out but we know it bullsh@t

[@FatherLamb]

Really depends if the person is playing 1080p (most still do) lol Current APU's can do some magic around 30-60fps 1080

[@D_Wito]

Well its true "no more gpu's needed" you are quite fine with apu nowadays, its better to have GPU but there's no need for them to get decent performance

[@ehenningsen]

We played without GPUs in the 1980s and 1990s. Yeah, i wouldnt recommend taking someone's word for it, that you only need a CPU for high fidelity.

[@novantha1]

I mean, you really don't need a discrete GPU if you don't convince yourself you need one. If you're not playing the latest AAA titles at 100+ FPS at 4k it's actually not nearly as bad as people make it out to be. Particularly if someone was to do primarily emulation, or retro gaming, or even just playing titles from before the 2020s (which includes a huge selection of masterpieces) this is definitely more than enough.

[@Rewind-xq6xs]

In fact, Apple already proved it’s possible. You can also play overwatch 1080p low settings on new AMD apu with about 100 fps? It’s real bro.

[@dystopia-usa]

I would gladly forgo a GPU (& 1440p resolutions) for an integrated graphics solution that could run high-quality 1080p at a native 90-120 FPS.

[@Teilzeit02]

And for a reasonable price lmao, there are no bad products just bad prices

[@selohcin]

Right, but that's never going to happen because games becoming ~30% more demanding every 3 years or so.

[@user-xn9gd5dn9s]

There definitely is a niche for it in the consumer market. Personally, I'd prefer qhd + high refresh rates, some people like 4k 60fps, each to their own. Would definitely wait to see how it performs on launch though and let the tech tubers review it.

[@malik-mahdi]

actually 780m can almost do it (in lowest settings) it gets 60 to 100 fps

[@ShreddedWheat305]

@@malik-mahdi The 780m is slightly slower than my past card the RX 470...if it came out around 2017 the 780m at 1080p would be a heck of a deal but slow nowadays.

[@Militaryman64]

When APU'S performance becomes par with GPU'S makes me salivate to thing what small form factors i can build 🤯

[@Sniper9773]

It's just a matter of time. Maybe in about 6 years we'll have apus that rival low/mid end gpus.

[@arenzricodexd4409]

​@@Sniper9773they also said that back in 2011 when first APU comes out.

[@kaseyboles30]

Problem is apu's have the same die size limits as gpu's and cpu's and they're trying to be both. Now with chiplets and the like you might be able to do a singe package "chip" that puts both wafers onto one substrate like how modern desktop Ryzen chips work, but now we're talking 'chip' sizes on par with threadripper and server cpu's with some serious power draw and heat dissipation at stock settings. And you cannot upgrade just the cpu or just the gpu. Though there are some pro's as well including simplification and possibly some interesting things that might doable where those to items share the same memory sub-system to counter what traditionally been a bottle neck.

[@PineyJustice]

@@kaseyboles30 You forget that the ps5 is an APU and has the equivalent of a 6700 in it along with an 8 core zen 2 processor. It's certainly possible, the limitation is ram bandwidth, unless a special package is created that has socketable GDDR6-7 or HBM is added to the cpu package to service the GPU these will forever struggle, this is evidenced by the 780m not really being all that much faster than the 760m. Zen cpu cores are so tiny it's basically a non issue, intel would have issues stuffing a gpu into their i9 line, but you could fit probably 4+ more ccds under the zen IHS.

[@starmaster2010]

Having been a consumer electronics repair tech for years. The 4090 power connector in my opinion given the current draw and the size of the connector pins and wire this is a failure waiting to happen!

[@Sniper9773]

Can it be fixed by using better materials?

[@kaseyboles30]

Exactly. That wattage (450 in the case of 4090) at that voltage going down wires that small is technically o.k. if hardwired and not subject to abuse. But the margins are way to small for what amounts to a single clip and friction fit with LITTERLY to much wiggle room. Fi they went one gauge bigger and or added another pair of current carrying wires and upscaled to connector proportionately while adding a second clip opposite the first it would likely be o.k. Note is' mostly the 4090's and a much rarer few 4080's that have melting issues. Add a clip on the opposite side and it could be around 350w rated, but no way it can support 600 watts considering how often it fails at 450.

[@RinoAP]

@@Sniper9773 if you meant better material is ceramic, then yes

[@teddyvetter5256]

As an electrician, I’d say the size of the power pins and the current draw guarantee there will be fires. Wait till these plugs fail due to fatigue. When you put amperage over a wire, it heats up and expands. When current stops, it will cool down and contract. This will cause the metal in the connection to fail over time due to fatigue. I see this all the time in residential receptacles with push in connections. I have 10’s of thousands of infrared pictures of these connections failing on receptacles and hundreds of them burned up. This same issue would translate to these snap together GPU connections that carry high amperage. You think there is a lot of burned connections now? Wait about 2 more years… failure rate will hit around 80-90% by then.

[@carlosbelo9304]

@@Sniper9773 No, it's just too much current going in small wires. they will burn, eventually

[@getinthespace7715]

It will be interesting to see how the Ryzen 9000 APU's do. I'd buy one if it covers the performance of GPU's priced under $300.

[@arenzricodexd4409]

Doubt it will. Before this there were some kind of hype for 680M/780M to rival 3060 performance.

[@Richard_Cranium]

It won't lol😂😂😂😂😂

[@malik-mahdi]

@@arenzricodexd4409 there never was it was about them being as strong as a rx 590 which is almost the case

[@arenzricodexd4409]

@@malik-mahdi 780M at best is at 1650 level. RX590 is at 1660 non super level. Though in some newer games 780M can close it's gap towards RX580 due to 780M having new architecture. Games like AC Valhala for example being tailored specifically for RDNA2 (and above) architecture making them very fast in those tittle.

[@markymarc12]

I remember getting an AMD, Trinity series APU back in 2013. It was the A8-5500. I played Battlefield 4 on that thing. Lowest settings, 720p 30fps. It was quite a jump from my laptop's Mobile 4 Intel Series Express Chipset Family graphics. Back then it was the A10-5700k that had all the talk with integrated graphics.

[@sixthguardian6914]

my 8 year old amd a8 7650k also got built in graphics card. still works fine in opening adult sites.

[@danielbarnett5764]

I don't get why the 4090s aren't being recalled if they are melting and clearly a fire risk. Is there a percentage that is tolerated under electronics sales rules? Have we now exceeded this safety margin if using northridges as a sample of the real numbers melting? Does it affect only the old connectors or was the new connectors with short sense pins the industry's way to buy some time?

[@darthwiizius]

In consumer electronics a 4% failure rate is expected due to manufacturing tolerances on PCBs. If the failure is being caused by an inherent defect that changes the matter, in many countries that throws responsibility to resolve it back to the manufacturer, bypassing the retailers who customers have their actual contract with.

[@mikeb3172]

Northridge is not trustworhty. He has done at least two previous videos of 4090 connectors where the connector was clearly cut off, just for views. If he did a real burnt connector, it's still user error.

[@HFRG-zq1qm]

Because NVidia doesn't want to spend money to fix an intended feature of sucking a watermelon sized charge through a drinking straw sized contact.

[@HFRG-zq1qm]

@@mikeb3172 There is a reason AMD and INTEL refuse to use the 12VHP connector. It tries to force watermelon sized current down a drinking straw sized contact. The whole configuration of the connector is only safe for 250 watts or less. And NVidia is trying to send up to 900 watts down that.

[@jasondelarosa5457]

APU's is good for a small form factor build/travel PC but, the price/performance isn't there for a dedicated desktop PC.

[@skychaos87]

Strix point and Strix halo are most likely going to be primarily laptop chips.

[@adammagie4466]

Wait for Strix Halo - 40 GPU cores.

[@Droon_Jadhav]

🥶

[@michaeldisante5145]

Why not just add thumbscrews to retain the connector? I seem to remember older PC cables having this arrangement and I NEVER had one of the those come loose!

[@Veganarchy-Zetetic]

Plenty melted when they were fully connected all the way in. It's just an excuse to put blame on the consumers rather than Nvidia accepting responsibility.

[@ehenningsen]

Ive had mine for a long time, no melted cable.

[@UltimateGattai]

It really should have been a product recall, but Nigreedia would never. ​@@Veganarchy-Zetetic

[@Veganarchy-Zetetic]

@@ehenningsen Me too but that doesnt mean it wont in future. I took one look at that tiny connector when I received it and knew instantly that it would be a high chance of failure. They should have kept the big connectors to ensure safety and reliability.

[@kaseyboles30]

@@Veganarchy-Zetetic The standard 6 and 8 pin connectors have a huge sagety margin. You could build a 250 watt ten pin connect in the same style and pin/wire sizes and probably still have a wider safety margin than this thing.

[@xxiamlegendxx1390]

Bruh this kind of Clickbaiting should be punished with jail

[@JahTakeTheWheel]

I mean he's got timestamps 🤷‍♂️

[@xxiamlegendxx1390]

@@JahTakeTheWheel what has that to do with the fact that he claims the new ryzen processors will replace gpus? 💀💀💀

[@clem9808]

Rtx 3050's performance for rtx 4090's price. 💀

[@leorickpccenter]

The sense pin being shorter is a band aid solution to the main problem of these new connectors. I wonder if these new APUs will be priced reasonably well.

[@HFRG-zq1qm]

the shortened connections increased the problem, it's current vs impedance. Watermelon sized charge, drinking straw sized contact.

[@ACOnetwork]

Let's see what we have on the menu today 😁👍

[@Elkarlo77]

About the 12VHPWR Connector, the 8auer had a Feature about it, it got me curious and i got my old Tables out and there is a massive flaw in the wire gauges of the 12VHPWR Connector. The old PICE 8 Connector had a 30/50% Safety Margin in it, the new 12VHPWR has only 10% Safety Margin. Which is a massive problem. Depending of the Material the Resistance of a Conducter and it's connections increases. For the most Wires used in PC Application this is at a rate of 0.3 to 0.4% per °C in a range between -5°C to 100°C. The 10% Safety Margin is at 20°C/68°F. Which means that at normal System Temperatures of 30°C/86°F the margin is down to 7% or 6%. When the System got Cozy at 40°C/104°C it is at 3% on average. Which means as long as no Part of the Pathway of the Wires is over this Temperatur it will barrely hold on. The Connector has been meassured with 50°C/122°C which means it is now in failure Range. If the Card draws the full 650W the Connection is just not capable enough and will overheat further. Most of the Time the RTX 4090 doesn't run at that wattage, but at Peaks it does and it goes up to 65°C in Temperature when not properly seated and the contacts are cleaned.. moisture and corrosion. Which putt it down to 550W and then the Death Cycle begins. Power Peaks will put it again in again of overheating, which means lesser conductivity and melting Connections. TL;DR The 12VHPR Connector has not enough Safety Margin for Systems over 40°C/104°F which it will reach at the Card in non Watercooled Systems. Powerspikes together with Corrosion will result in Overheating Connections directly at the Card. Clean your PC with a RTX 4090 regularly and reseat the Connector to keep the connection optimal. Case/System Temps below 35°C/95°C are advisable.

[@LBXZero]

And the engineers behind the 12+4 pin connector family keep reducing that safety margin so they assume they can push more power. The problem that that safety margin handles is that nothing really guarantees the amps will be evenly distributed between those pins. It just starts with 1 pin drawing too much amps.

[@Elkarlo77]

@@LBXZero One of the reason some Connectors comes in from the cables have one ugly massive solder bridge and then goes to the pins. To spread it evenly. There is a reason why most electrical connectors are overbuild with a 30% margin. Somehow the commite of the 12VHPRW forget it. The world out there is not in a laboratory but with slight corrosion and thermal differences. 10% Margin means it is just enough for 1 factor while you have several factors out there in the Real world: Mechanical Problems prevent 100% seatings, temperatur increase resistance, Corrosion increase Resistance, Quality Control in the factory. Everyone can be addressed by a 15-20% Margin. Overbuild you have a 80% Margin. Underbuild you address one of it with a 10% Margin. Ideal you address 2 Factors which may happen with a 30% Margin. So it may get hot and not in perfect contact. Or it may get hot and little Corrosion. As it is likely to happen over time.

[@Dr-EV]

3:05 It's time for heating and fans to be mounted on connectors.

[@tatersalad6872]

The problem is the plug itself and how the pins connect. You have these 12 pins that are expected to carry a combined 600+ watts of power. Each one of those pins has a few tiny points of contact. It's not a solid connection the length of the pins. Putting that much power thru a bunch of teeny tiny connections, and if its not fully plugged in perfectly square you lose a bunch of those tiny little connections. All this mess can cause those tiny little connections to heat up.... Then POW! lol

[@HFRG-zq1qm]

not only that, fully and properly seated, those 12 pins are only MEANT to safely transfer 250 watts, 300 really pushing it. And NVidia insists on pushing up to 600 watts down it.

[@ctaylor960]

My 13900K had to be warranty replaced...Asus said for the new one to turn on Intel Fail Safe setting...its good now (only up to DDR5-7000) and its stable...

[@kaseyboles30]

A computers bios should have a default setting within the cpu's designed limits. NO motherboard should be overclocking as a default. This includes keeping max runtimes for short term boost frequencies and the like.

[@turboimport95]

The problem with the connector is 100% over power draw issue. My 4080 has this same plug and it pulls 300 watts and dont melt, my pc also runs 24/7, I had this card for about 1 year and i have not seen any 4080's melt at all anyway. Lower the 4090 to a hair under 300 watts and it wont melt. But i guess pushing 500 watts is more logical.

[@HFRG-zq1qm]

250 watts is the actual safety rating for the connector specs. 300 is feasible, but likely pushing it. Anything over that is just inviting disaster. NVidia knows this and knew it then. That is why EVGA no longer makes Graphics cards, because NVidia knew the problem and refused to allow EVGA to use additional connectors to correct the per connector safe draw limit. They seem to think burning down houses with their product will help them sell more of that product. I mean, logically, it checks out. You WILL need a new 4080/4090 after it burns your house down taking itself along with. But realistically, you'll need to buy a new house first.

[@slaynnyt8130]

Been using my 4090 with undervolt since day 1 that I bought it, it always stays below 300w, hopefully that will prevent or at least delay this melting connector mess...

[@karlos1060]

It's nice that APU's get more powerfull for people with less money or the ones that do not want to buy a dedicated GPU. For me i do game and need a lot of power. So an iGPU is not going to cut it not in a million years. Maybe in a distant future it can become powerful enough. Currently gaming with my I7-14700KF paired with my 4080 super.

[@XVa-uj8m]

Take out those AI cores. I don't want a borg like attack vector on my computer for a GPT sized AI. Who is to say my CPU AI cores can't be taken over then my computer is completely hijacked like a virus.

[@SchoolforHackers]

Oh, I can assure you, they can.

[@POLARTTYRTM]

This is why I don't want AI bs on phones and pcs, they have been pushing this AI so much since last year it makes me physically sick just from reading or reading AI-related something in PC or phones.

[@rodiculous9464]

This would be a great use for that 2nd pcie slot that most people leave unused imo. Have an add in AI card and maybe even a hard switch on the card itself when you don't want it turned on.

[@gummislayer1969]

@@SchoolforHackers ...AND prolly WILL!!! 🤖🤖🤖

[@samal90]

calm down, AI cores is just a fancy term for another type of CPU module. AI today is so basic, we are very very far away from anything scary.

[@JustHearMeOut_YT]

i can't go back to 1080p after playing ultra wide 1440p... i just can't do it

[@chrisward978]

I feel excactly the same. My first ultrawide was 2560 x 1080. I'm 3440 x 1440 now and I prefer it over 4k. My eyesight is terrible anyway lol.

[@JustHearMeOut_YT]

@@chrisward978 yeessss 3440 1440 and supersampling is gonna carry me for a long time before i go 4k

[@hdcomputerkeith]

I hope they make a better power c for the 5090. love North R vids on repairs.

[@vincentvoillot6365]

APU will never worried GPU in raw performance term, for one simple physical limitation : memory bandwidth. 128-bit bus shared with 12 cores CPU & 16 Compute units seem short to keep all these units fed, an RTX4060 alone have the same bus with DDR6 for itself.

[@selohcin]

Note to Gamer Meld: "Released" means released to consumers at regular stores, not "released" to OEMs only. Please stop with the misleading language.

[@NovaTurtlez]

Tru

[@humayoonxpro3220]

what about that leaked strix halo with 40 CUs?

[@henrydrago]

Probably to laptops

[@uros2909]

It will most likely come to laptops only, but if it somehow end up in desktop PCs, I'd be the first one to buy it, because that thing looks like a legitimate beast and a true dedicated gpu killer

[@glanguish9390]

If it's laptop only Minisforum may do a soldered ITX barebones of it. Finger's crossed

[@henrydrago]

And handhelds too

[@daneverdier1950]

My 4080 melted and he fixed it. Good work, horrible customer communication during the process. The rgb goes nuts now during gameplay and I can't get a word from him.

[@etchieSketchie]

I’m excited for what this might mean for the future of handhelds. Valve doesn’t want to release a Steam Deck 2 until there are chips available with a “meaningful” jump in performance. This just might be it! 22W is INSANE!

[@HFRG-zq1qm]

The 12VHP issue is plain and clear: The amount of power they are trying to force through too little surface area contact. They'd be better off trying to funnel that amount of power through an old fashioned MOLEX or SATA power connector.

[@legendaryz_ch]

I own a v1 4090 with the cooler Master v850sfx 12vHPWR adaptor cable working like a breeze

[@BeeFourFifty]

My 12700KF is unstable and likely suffers from similar issues to the ones mentioned here regarding the 13900K. Very frequently, in the middle of a game - it will randomly crash and reboot. I thought it was the GPU at first as it was happening with my 4070 Ti but after swapping it out for a 6950XT - I had the same issue. Both GPUs work just fine in my B550/5600X3D build.

[@DaltonP386]

Probably low SA voltage, had the same issue with my 13900k with a 6000 cl30 kit.

[@BobYeeterson]

Im gonna be so frikkin happy if apu's get as good as discrete gpu's or very close. It gonna help me save up for a high end gpu.

[@Housey29]

Bad connection can cause resistance, arcing resulting in heat hence the melting connectors.

[@LBXZero]

Doesn't need a bad connection. Any resistances from inside the PSU to the PCB on the video card can cause the amp delivery to be imbalanced across the 6 +12v wires, and the 12+4 power connector design is removing all those safety buffers that compensate for these imbalances.

[@johndoh5182]

No, next gen APUs are not going to replace GPUs, not even at the low end. The gen following this one will if priced right. The problem NOW for GPUs is they need AI cores, RT cores and then other cores to make the image. It's hard to shove all that in an APU. If a GPU ONLY needed cores to make an image with no RT then Zen 6 APUs would EASILY be good enough for 1080p gaming.

[@Anonymous-sb9rr]

I wonder in memory speed will be a bottle neck for iGPU's.

[@jamegumb7298]

It always is. Look up data rates for memory. GPU memory is 4× faster (DDR5 vs GDDR7). The second thing is power budget. AM5 PPT is around 240 watt with Intel not far behind. Yes I know they can take in more but in no way sustainable load, peaks and boosts and transients only. Then you need some reserve for boosting, some for safety, maybe OC and you get 120 watt or so for the GPU part, where the 3060 is at around 200-220.

[@Anonymous-sb9rr]

@@jamegumb7298 The biggest iGPU they've planned is only 16 CU's, which is half of what the RS 7600 has, so I don't think power will be an issue yet.

[@jamegumb7298]

@@Anonymous-sb9rr And that limit gets you a further limit on performance. There is only so muhc you can squeeze out of a chiplet gpu per core/CU for a given architecture. A slab of HBM/GDDR as chiplet will probably help, either dedicated vram (but adding 8GB let alone more will be very costly, especially if HBM) or as buffer to main memory like what they do with gpu's right now.

[@Anonymous-sb9rr]

@@jamegumb7298 That could all be very cool in the future. What I meant with my original comment, is whether RAM speed wil be a bottleneck in the iGPU's that were leaked and that this video talks about, which is at most 16 CU's of RDNA 3+, with regular 2 channel DDR5 memory. Can we even make proper use of 16 CU's?

[@jamegumb7298]

@@Anonymous-sb9rr It will be. Look up tests of 8K APU concerning memory scaling. DDR5-7200 goes quite a bit faster than DDR5-6000 (highest speed 1:1IF). Even 8000 speeds get better GPU performance. So memory is a bottleneck. Using the same memory and going from RDNA3 to 3.5 will do what exactly? Not a lot. Using more CU will give more performance but diminishing returns the more you go up. _Unless_ they do the same as AMD does now on its video chips: Infinity Cache. It would help quite a bit. Remember the 8700G being touted as the 1080P monster? It mostly plays 720P okay. I expect the same to happen to this chip, it will not live up to the hype.

[@POLARTTYRTM]

No more GPUs! Yes, if want to play CPU-bound games and still getting 30 fps on 1080p medium.

[@meibing4912]

My PC is around 40x the size of my wife's mini - she has never ever complained about performance. Still planning to have next PC with a gpu due to photo, video editing. There will probably also be AI programs relying on a GPU for a long while.

[@Zangetsu_X2]

What is in her mini? And what does she do with it?

[@meldmagic]

🤠 His PC is in her mini. Figure the rest out.

[@Zangetsu_X2]

​@@meldmagic Plot Twist!

[@Skarfar90]

What about the Strix Point Halo. The 32 CU part. Really curious about that one

[@koford]

When you say: If you are the owner of 4090, make sure to plug in properly. That's not the problem, it's the design that's the problem. Remember this has not been a problem until this new design came to the 40 series. Bet the 50 series will have new problems.

[@420IQPlayer]

lmao literally I've had a 4090 for over a year now and have yet to have any problems and I have the older original design, so again it's user error bud but keep hating on the best gpu on the market

[@Oliver-sn4be]

​@@420IQPlayerfor that amount of money the no way this should exist 😤

[@boastyy]

They know its a problem but have not done anything about it. A small percentage are probably user error the rest power connector problems as well as cracked pcie connectors. For the 50 series the connector has been redesigned, they would not have done this if they believed the original connector was 100 percent safe.

[@koford]

@@420IQPlayer its a "user error" because of the bad design..duhhhh. Iv been saying it all along. read my friend.

[@koford]

@@boastyy it a "user-error" because of bad design. duhh.. read man.

[@bgates275]

The massive APU jump is only indicative of how much of a leap the RTX 5000 series will also be.

[@ShaunaJagan]

I'm really hyped for "Strix Halo" with those delicious 40CUs! Gonna order a Laptop for sure Thank you, Dr. Lisa Su ♥️

[@harshivpatel6238]

Don't hype yourself, AMD has a history of disappointing it's fabs & missing it's targets

[@user-mz1if8oe9k]

That's good that we will have better iGPU, but it's still suitable (mostly for SFF) - only x8 lines and PCI-E 5.0 mobos are still not cheap (just a small look to the future when all new cards will have 5.0), performance improved but it still didn't beat RTX3050mobile (looks like somewhere around RX590/5500 or GTX1066/1070 in productivity some of which cost less than $100 used). Sad, but I don't think that we will get anything way better during AM5 lifetime, just like we got improved Vega8 after Vega11 - at best we may see "match desktop RTX3050-6Gb" results in some cases.

[@internetw4nk3r74]

That would free a pciex16 slot and improve (small) cases air flow performance. Big win for average consumers on a budget.

[@georginenov6779]

The thing with the "Cancel GPUs" has been ongoing for some time and has not happened as of yet. I highly doubt that an APU will ever be able to get close to a GPU.

[@samal90]

I think eventually, GPUs will be only for hardcore gamers that want the latest bells and whistles and crank up visuals to the max and options like RTX, etc. I mean, with Strix Halo coming with 40CUs, coupled with FSR 3, it could do RTX with playable FPS.

[@scotrick333]

Hmmm... the Steam Deck 2 may be coming sooner than we expected with a jump in APU's like Ryzen 9000.

[@kitnoman]

wow 22w? It seems AMD is gearing up for handheld PCs market as they are in the lead with igpu. I can see it now, a custom 6 or 8core with rdna 3+ chip on steam deck 2 next year lol

[@samserious1337]

9:52 hate to say it, but the graph is "flawed", since the ROG Ally (Z1 Extreme "Pheonix Point") gets 2800 Points with 30W in TimeSpy. 55W are NOT needed for this score.

[@nobmobbob21xd69]

Does anyone know why people take their 4090s to a repair shop? Like isn't it under warranty or anything?

[@chrisward978]

I'm just speculating, but maybe that guy is one of many used by Nvidia as a repair tech. Might be way off the mark here but 200 connecters in one shop seems really excessive unless they are covering a very wide area.

[@nobmobbob21xd69]

@@chrisward978 yeah most likely something like that, because as you said, there's no way they had that many repairs, but i was just wondering because i heard that if there's a problem with the connector u can just return the gpu for a replacement

[@johndoh5182]

I thought that Nvidia connector was destined for many failures as soon as I saw it the very first time. The notion of 600W going across that one connector is just too much. You need very durable pins, you need very durable wire, and that all adds up to a lot of bulk right there at the connector, to where if you put torque on it problems are GOING to happen. Any person with a lot of experience wiring houses or buildings could tell you that, because of the understanding of working with different types and thicknesses of wire. When this whole fire and melting thing started I simply expected that as you get closer to pushing the max of what that connector has to deal with. I was surprised it hadn't started happening sooner than the 4090, but with some of those bigger GPUs you get the issue of a wire bundle getting pushed into the side panel of the case, creating that torque. When Nvidia and some other companies said the problem is fixed I said in a few comments that it's FAR from being fixed because millions of these GPUs have been sold, and many people wouldn't be aware of the significance of that connector or the problems related to it because only some PC enthusiasts watch KZbin videos and tech news, and most the world certainly doesn't watch these videos. I mean, if that connector were in an aircraft, the VERY FIRST ONE that melted and caused some failure in the plane means that EVERY PLANE with that connector would have had it replaced. As it stands now, that connector is a house fire waiting to happen and I'm thinking that if it doesn't get phased out it's a class action law suit waiting to happen.

[@xzm996]

the more i hear about 4090 failure the more i get happy. International regulation parties have to set a 400W limit on TBP of each and every card released, even if it gonna be rtx 9090 ti super ultra. And set limits to cpus at 300W so that even i9-19900KS doesnt get above that

[@rodiculous9464]

Arent all the next gen nvidia cards going to have that connector? Oh boy. Idk if my poor 1080 can serve me through ANOTHER generation of cards.

[@212betto]

If they make ryzen 9000 apu quad Channel memory maybe mid end GPUs Will become relics of the past

[@bgates275]

Still have to deal with AMD drivers. For that reason alone I'd take an Nvidia card. It just works.

[@cagc7963]

if the best igpu wouldnt beconditionated to be with the top r7 they launch, apus could be more competitive, however they always crap themselves with their strategies

[@aakee87]

Hopefully we get refresh version of steam deck with 9xxxG.

[@DragoNate]

you say this every time a new cpu with integrated graphics comes out. gpu's aren't ded, you simply don't NEED one for decent performance at 1080p. gpus are still better and/or needed for so many other use cases like rendering, high end graphics, 3d stuff, vr.

[@wudruffwildcard252]

Yeah, current APU Gen was hyped up as a low end GPU killer. And yet it can't beat the GTX3050 and the most basic CPU MB combo in neither price or performance.

[@DragoNate]

@@wudruffwildcard252 Exactly. Like, on a serious budget or something, I'm sure it's good enough to get a playable experience in most things, which is fine to report on as such, but marketing it or talking about it like it's a pure replacement for a dedicated GPU is ridiculous. It's an option if needed/wanted under certain conditions. That's all.

[@pauloazuela8488]

@@wudruffwildcard252 Well the 3050 is basically rtx 1070. That would be very unrealistic with that size unless it's a console type apu like PS5 and Xbox X

[@skychaos87]

Obviously high end GPU will continue to exist and dominate. But the trend is that APU is creeping up closer and closer to upper tier dGPUs. Remember GT x20 or GTX xx20 GPUs? 20/30/40 lower end GPUs are totally dead, for a long time already because of the iGPUs closed the gap. Heck, nVidia didn't even make a 4050 GPU this gen. And the big thing about AMD's new APU, especially Strix Halo is that its another massive jump in iGPU performance. Strix Halo is most likely going to challenge and come really close to mid range RTX 5050 desktop and 5060 laptop dGPUs. Well of course 5060 desktop and above will continue to make sense but the need for upper mid and higher end GPUs are a smaller percentage of the gaming market. Another trend to note is that back then iGPUs cannot do 1080p gaming at all, with Strix Halo, 1080p on modern games is not going to be a problem anymore. Heck, with upscaler and frame generation tech, 1440p at over 60fps is very doable. Again, iGPU is not going to totally replace or overtake dGPU, but it is definitely displacing the lower end and challenging the mid end dGPUs.

[@wudruffwildcard252]

@@skychaos87 Mark my words, this and the last APU gen were hyped up to destroy low end GPUs but they couldn‘t compete on price or performance. It will be the same again when the new generation of dedicated GPUs comes out. Strix Halo will be a 550$+X CPU + AM5 Bord + the best DDR5 memory you can get because otherwise the GPU side of the APU is starved to death on the memory bandwidth. A dedicated low end GPU built will outperform any APU setup in performance and price to performance!

[@rickh8380]

That connector should be made from ceramic to combat the heat and melting of the connector?

[@HFRG-zq1qm]

That connector shouldn't be used for more than 250 watts per connector. Making it from ceramic will just cause the surrounding materials to combust instead,

[@rickh8380]

@@HFRG-zq1qm Thanks for the info. I thought the ceramic would handle the heat better.

[@texedomel01]

My next build can just be a Ryzen 9000G if I am only to play older games in my steam backlog.

[@LBXZero]

The problem with that 12VHPWR is not the connector. The problem is on the video card's PCB and a disregard for physics. The 12VHPWR connector family is "successfully" failing. The problem involves how the power "paths" from the 12v rail(s) to the power distribution system on the GPU. In single rail PSU, all the power sources for 12v comes from a single post. On the 12VHPWR connectors, all 6 +12v lines merge at a single VRM. This creates 6 routes for 12.5amps (150w), 25 amps (300w), 37.5 amps (450w) , or 50 amps (600w) to split between. Amps per area of conductor versus local resistance is what results in heat generation. Due to the design of the entire path from the 12v rails in the PSU to the power distribution system on the GPU, there is nothing regulating the amperage division between the 6 wires, placing faith that the random resistance tolerances won't add up to a significant difference. The same is true for PCIE 8-pin and 6-pin connectors, but due to their low combined amp limit compared to total amp safely possible in a perfect world, there is a large safety buffer for uneven distribution of amps. For the 12VHPWR connector, these only have an equivalent safety buffer when running at 300W (25 amps). All the variations in wire quality, wire length, and number of connection points from the 12v rail(s) in the PSU to the video card create a total resistance for the path. This total resistance is what ultimately determines how much amps passes through each wire. A total path with lower resistance will conduct more amps regardless of the resistance at individual points. If the total resistance is imbalanced enough between the 6 +12v wires, it will be much easier to load over 10 amps (exceeding spec of the connector) through a single pin on the 12VHPWR connector. The adapter dongles that input 4 PCIe 8-pin connectors and output a single 600w 12VHPWR connector make it vastly easier for resistance imbalances. The melting process will be slow, as once the plastic starts to warp and contaminate the conductivity of one pin, it will alter the total resistance, passing the amps unsafely through another pin, and it keeps going back and forth until catastrophic failure. Want to test this? Using a 4 PCIE 8-pin to 1x 12VHPWR adapter dongle, add PCIE 8-pin extension cables to 3 of the PCIE 8-pin routes. Each length will increase those resistances, giving the 1 PCIE 8-pin a better conductive route that should melt first. Stress test for at least 24 hours or more. This is basically exaggerating the problem to make it visible quicker.

[@remisteele8904]

AMD also said that they don't expect to make any APU's with anything but RDNA 3.5 until after 2027, which blows my mind. I can't help but think that theres far more issues with RDNA 4 than previously thought.

[@VictorMartins239]

now we just need a G-ITX (G for integrated graphics) mobo and even smaller cases

[@LovelyAlanna]

My consolized computer has an RX6400, and now you say it could be faster with just an APU... tech just doesn't stop

[@sannyassi73]

Once a CPU can run VR at at least 1440p (that means 1440p in each eye), then we won't need GPUs. That's a long way off though. I do suspect that a decade from now there will be no more low end GPUs and we'll get tech that allows us to sort of crossfire/SLI our integrated graphics with our dedicated stuff to give it a boost. Tech like this has popped up every now and then, I remember a few Years ago this was a thing that some Motherboards supported, you could even mix AMD's integrated graphics with an Nvidia GPU or Intel integrated with either of them (if I remember right). That might not have been CPU graphics though, it may have been graphics integrated into the motherboard, which used to be pretty common.

[@kylenorris9585]

How did they all the sudden forget how to build GPU connectors?

[@SethonanGaming]

They didnt , its propietary bs to force you to buy their replacement at inflated prices. Its just literally blowing up in their face

[@T3chWarrior]

why not add retention clips for the opposite side of the connector so that both sides are secured?

[@LBXZero]

I think the point being made in this video is that the repair data is demonstrating that the problem is not the connector disconnecting.

[@narrativeless404]

APUs limit not just GPU power They cut down CPU as well And it's not easy to balance CPU to GPU performance

[@tactrix1h]

I'm not waiting on this upcoming APU set, I'm waiting for the next one or 2, because by then they will have made it all the way to mid GPU range properly and THEN you're gonna see a huge chunk of the market stop buying GPUs and switch exclusively over to APUs.

[@herbertholland924]

APU's are impressive these days, but are stll not close to "cancelling GPU's" They may be able to replace the low end, but mid range and high end, have nothing to worry about.

[@jonasmartinez1739]

that new GPU connector will still not work since it is just way too small.... compare that to the regular 8 pin GPU power connectors.... yeah... wont work

[@jadedforlife17]

So I love your vids first off but do you actually game or just report on GPU'S? If so what are you running and do you ever do videos on you gamin??

[@olivr_wtf]

OMFG! THANK YOU. I WAS ABOUT TO BUY A HANDHELD LMAO. I WILL WAIT.

[@zameerbe]

by which year 9000g coming?

[@XVa-uj8m]

WTF will the Strix Point Halo look like Gamermeld?

[@enchantereddie]

The 12vhpwr connector is a joke. It only works under a perfectly ideal condition. When being pushed close to 600w, which is its rated limit, self-destruction is just a matter of time.

[@amante762]

It's not that crazy, it's normal and expected. Apple silicone has had competitive graphics for 4 years now. It's called catching up. The Macbook air and pro are the best all around laptops. My Asus G14 battery just died on me while working in a GUI for 90 minutes, not gaming, after a full charge.

[@DaltonP386]

My 13900k wasn't stable out of the box with 64gb ddr5 6000 cl30 ram. I upgraded from a b760 to a z790 so that i could up the SA voltage by 10mv and worked like a charm.

[@ARH0101]

Why were you even running a 13900k on a b760 to start? Trash VRM, limited oc potential. If you are gonna drop damn near 1K on a cpu, get a motherboard with the power delivery to handle it 😂

[@DaltonP386]

@@ARH0101 But that's the thing, it COULD handle it really well. I had the same symptoms of crashes whether it was the b760 or z790. I also never planned to OC it just UV it.

[@kingtuckgesus8879]

I can tell you this from 100% personal experience overclocking your ddr5 ram on the apus yields massive gains on 8700g. I can almost play most games at 1440 medium or low and these are predicted to be much better ryzen has me excited for this launch!

[@misterPAINMAKER]

I expected RDNA 4 igpu on the APUS.

[@thseed7]

Very much think Nvidia needs to go back to traditional connectors or a more robust design than the 12v. Nothing like paying a company $2000 dollars to destroy your computer or burn down your house.

[@belarius2189]

Can't wait an all in one motherboard that only require power to work , no gpu and no ram to buy .

[@gamingtemplar9893]

The APU is great, but motherboard and memory is part of the cost, needs to be compared to other setups in price/performance. Still looks great if you want a very fast CPU but don't need a powerful GPU. My only issue with this is that it is AMD, I prefer Nvidia, but I guess gaming, AI stuff, and other things are not something you would use with a IGPU.

[@bionicseaserpent]

the APUs are something i need being priced really nice and low. poor guy like me can't spend an extra 1k AUD on a mid range graphics card frick that

[@greebuh]

If you pay attention to the channel about the 4090 connectors, you will see that it's not just 4090s.

[@tech4u2022]

If they hit the RTX3050 at stock then i am all-in for it ! The iGPU's from AMD are truly rock solid the last few years no matter what people say. If you tried it you know what i am talking about. The biggest advantage is that you won't have to worry about VRAM thus making it 4K capable out of the box at 65 watt, that's a no brainer to me ...

[@garmar704]

Bottom line is there's too much power going into GPUs now. It's out of control.

[@uggZymoroon]

maaaan am I glad I built my most recent pc completely on amd. I'm just chillin, no issues. I fucking knew why not to buy the 4090!!

[@trusoljia87mrjcazz76]

Guys should be in 2160 with aa. Or 8k 1440p lol I'll never game with same resolutions as a consoles setup.on pc,

[@HuntaKiller91]

1080p100 igpu is good enough IMHO Improve on AFMF and it's perfect

[@samserious1337]

Its only 16CU which is a good upgrade, but nothing life changing!

[@gaza.primera]

another hype! unless those cpu's will increase 1000 usd in price for 4090 performance, it won't be replacing any discrete gpu soon

[@tatersalad6872]

It seems to me the 4090's are basically going to be a 100% failure rate given enough time...

[@andrewsolis2988]

People been saying this since the 5kG series 😂 What? Maybe we can FINALLY game at 1080p on balanced upscaling without severe frame stutters at almost 60fps on medium settings?? Bc my higher end Rog Ally sure cant!

[@FireCestina1200]

I can only imagine one mfer having his APU running at 250W (And now for a reason unlike intel CPU) casually beating the 4090

[@abdulhkeem.alhadhrami]

It's not the connector that is bad but Nvidia's stupidity to push 600w through thin pins every card from 4070 ti and below are fine, never exceed 80% of any rated capacity of a cable\fuse\connector!

[@Robotrik1]

I would be weary about AMD news coming on on the 1st of April .

[@danielbarnett5764]

I don't believe it's the connector - i recon it's simply drawing too much power - i.e. perhaps there is a logical board issue that's been overlooked thats pushing the draw past the connector capacity. Also, why not just have three big pins for the power connector? It works for power sockets - why do we need lots of tiny fragile ones connectors? Plus, if it is "not quite connecting" well just make really long (ideally thick) pins. Or just recall all the 4090s until the problem is resolved.

[@gamingtemplar9893]

It is not about belief, it is about the people who understand these things and tested it, and the problem only present with that specific connector and not others. Whatever Gamers Nexus said was obviously false, they wanted to protect Nvidia blaming users, or it was absolute incompetence and ignorance.

[@420IQPlayer]

@@gamingtemplar9893 lmfao keep whining about a user error issue, I've had my 4090 for over a year now with the original design and have yet to have any issues with the connector because I know how to plug in a connector 🤣

[@UltimateGattai]

I partially agree, I think the wires are just too thin for the current, but using thicker wires would just ruin the "aesthetics". Plus, the new connector has a razor thin safety margin compared to say, an 8 pin. They should just go back to using multiple 8 pins even if we end up with 3 or 4 of them sticking out of the card. I'd rather not have a fire hazard waiting to happen. I don't think it's a connection issue either, because they've remodelled that connector several times to ensure a better connection.

[@Techno4more]

​@@420IQPlayerthe connector is a stupid design. The margin of error is small. I laughed when I first saw the connector. If you don't understand why, its ok. You're just uneducated.

[@Techno4more]

It's the connector.

[@pankothompson5903]

If APUs had HBM ram yeah they could be good but they wont do that

[@Balu-rk7ys]

Small PCs will have a much higher use rate :3 I am tuned for those CPUs