its not sad that Zen 5 doesn't have an 40% IPC increase. its stupid anyone ever expected that in the first place. so no. Zen 5 isn't worse than we thought.
Zen 5 will be better than zen 4. People just have their expectations far too high. However my 7800x3d will match the 9700x in gaming so no need to upgrade until the 9800x3d in 2025.
I mean if you compare Ryzen 5000 and 7000 it's pretty close to that. Although it's very rare to see huge performance difference on the same plateform so it may not be comparable.
A particular program using AVX512 gets a 40% boost, Xitter leakers & click bait reporters create daft hype. IPC is just a claim, measuring a basket of selected programs at a specific frequency.
ill have 2 x 4090 and never had any issues with it . and yes , ill buy 2 more 5090 just cause i need more vram power for AI , but nvidia should make the gpu around a pc case so we can plug the other components to it
@@billlam7756 There is certain failure rate that's expected in anything. When you're this far over it you have an issue. And these are designed way to close to their limits.
I'm getting there. I just upgraded my gamer room wall outlet to a 20amp rated outlet using a single 12-gauge service wire connected to a dedicated 20amp fuse in my service box labeled "PC." Power-wise, I'm ready for my new 5090 GPU.
I dont get it. Power consumption efficiency is very valueable as a PC owner. If the 5090 is gonna take 600 Watt, no matter what power it holds...that is a massive increase in energy costs. Not to mention you will need a bigger PSU. And on top of that, it will probably release at around the $2000 mark. It is absolutely not worth owning such a GPU. Just red flags all over the place. We need to decrease power consumption while also increasing the power. Be more efficient. Or at the very least, only increase power consumption slightly.
Only $2000? I suspect NVIDIA will play the game it did for the 4000 series--initially only release the top end card for an insane price, and limit the supply of 4000 series card to keep their price up. It's not as if they care about gamers--the real money is in AI.
And in hotter climates, you need to get that extra 600W out of your room via air conditioning, too; so you pay extra money for the AC unit's electricity, too. In colder climates it's a an inefficient space heater at least. 😂
@@user-in2cs1vp6o I know but look at how much more efficient they are to a comparable Intel CPU/Nvidia GPU! The M chips also have GPUs so what difference does it make if they are ARM chips? They may not be as fast as a dedicated Nvidia GPU yet but look how much more power an Nvidia card needs, it's getting ridiculous.
The 4090 was also rated for 600w initially. That's why all the 4090 coolers are completly overkill and keeps the card so cool, because they were designing them for 600w.
I got curious, RTX2060 12gb, hitting it with the most intense game I have, 71C 1280rpm 165watts, 75F room temperature. A single small case fan and a power supply fan of about the same size to exhaust heat. The air filter in my room is still louder.
The Zen5 40% (probably less) is for AVX512 only from what I have read. I will be happy with 10-15% and move to Zen5 when my 7800X3D runs out of legs in a decade or two....
If I remember correctly, 2nd gen Ryzen was only about 10% better than 1st gen and 3rd gen was about 20% better than 2nd gen. . The real gains weren't seen until Ryzen 4.
Current motherboard designs are already obscolete due to oversized GPUs that block PCI slots. I hope motherboard and case manufactures finally deal with this ridiculous issue by coming up with new form factor that finally separates out the GPU from conventional PCI slots. We need a new connector specifically designed for GPU data and power along with a new efficient way to mount GPUs to cases.
The case would look weird. Unless the mobo gets shortened and still fits inside the case unmounted somewhere. They'll have to get smaller eventually, they got new tech they're just milking this old tech dry first
Maybe a change in PC case layout, have the case more horizontal than vertical, double that bad boy as a monitor riser and then we won't have to worry about Nvidia breaking PCIE slots because they have a boner for AI.
the tdp range is the same as existing 40 series products. So the coolers and their size wont change much. id love to see a trend in the other direction, but i think that wont happen any time soon.
@@AshtonCoolman not concerned about that, yeah those work well, I mean the pcie slots physically covered by the multi-slot width gpu’s make them useless.
High power draw is fine as long as it idles cheaply. My problem is the lack of balance for the consumer. Lovely GPU, but you're stuck with the 2.5 or 10G NIC on the motherboard because the top-end consumer CPU only has 24 PCIE lanes. Also, some of your motherboard NICs & SSDs will be bottlenecked through the chipset. Overspec'd GPU, but too bad about the storage and the networking and there are no spare high-bandwidth slots for anything innovative. 24 pcie lanes is too small. Threadripper 128 lanes is excessive for most use-cases. Where are the 42-64 lane options on high-end consumer CPUs? I think the answer is: They're gone because they would destroy the low-end server / workstation market, but it is preventing the consumer market from doing anything interesting. 50/100G networking is impossible. Could we do displayport over ethernet? There's no way to grow a new idea in the consumer market.
Current cpus basically have 48 lanes of data bandwidth, most stuff is using gen 4 while the cpu and chipset support gen 5. That's why the x670 boards are so nice, uses gen 5 to the chipset so all the pcie off chipset is running at full native speed since it's connected to the cpu with gen 5 x4.
@@PineyJustice Thanks for the tip. I've been hunting for the PCIE4/5 equivalent of x79 but I don't think threadripper or Xeon W are cost effective, so x670e dual x8 seems to be the way to go. Cheers!
i think the cards will stay the same size and then people will just make watercooling them more standard or board makers may have to just make more watercooled options but i doubt they would grow that much over the 40 series
There's a simple solution to the heavy card issue: desktop cases! The weight just goes into keeping the card in the slot, and the case brackets at the back support more of the weight from the direction it's supposed to go in. We seriously need to ditch tower cases and return to desktop ones. Also, more efficent for desk space, as you can put your monitor on top of it. Now, someone make a desktop case that has 6x 5.25" drive bays, so I have somewhere to put all my drives and I'll probably buy it.
Monitors on top of desktop cases is a huge ergonomic no-no. Eye level should be on the upper screen bezel. Show me how to do that with the typical 27"+ display.
@@marikann9073 I don't know how big that is, but if it's more than half a metre, it's too big. Why would I want my eye level matching the _top_ of the screen? That'd just make you look down all the time. You want it raised up a bit so eye level matches the _centre_ of the screen, so nothing requires looking too far away from your rest position. Well, that's what I do, anyway. I have my monitor sitting on some old books because it's too low for my liking when it's on the desk.
I was always hoping that they'd be able to slap a few heat pipes and a little fin (edit fam to fin) stack on the other side of the board to help dissipate some heat from the back. Maybe that'd lower the load on the front fins and pipes but I'm not sure.
I said it before, the wire and connector gauge is to small for the amperage that is being pushed through the connectors. Everyone said no, it's the way its connected, it will work now. Yeah, how is that working out for you now?
I'd say it sucks full time. Windows 10 is supported by Microsoft for one entire year when Strix will launch. But, the upsides is that it will probably help Linux market share a bit more, as people ask what to do and other people start recommending things like Linux Mint. Still not an excuse for AMD.
@@Psycho-Duck Oh, I don't know ... caring for your customers ? Because it's their job ? Because it's still a significant percentage of the total market share ? Because people will still using after the ENTIRE YEAR it will still be supported, just like Windows 7 was ?
The problem with nvidias power connector is that they are using cheap wire to push so much power through. Cheap wire increases resistance and thus creates more heat, which leads to melting over prolonged use.
Why would and GPU developer make a 4 lane wide card? How stupid! And it's certainly not going to require 600 watts! No one would buy it nor could they even use it.
For those couple dozen thousands that will pay the money to get the best GPU, the size or power draw won't be a major factor. If it can be done with off-the-shelve solutions, they'll do it.
If a GPU got that big, then using a GPU enclosure is reasonable. I wonder if anyone is brave enough to create this thing for PCs, not for laptops (or they can support both), which is basically an "advanced version" of the PCIe riser.
That's totally Not why. My 80 melted and I made sure it was fully seated every time I got in it. It happened after a few really long sessions. Regular game play wasn't enough to melt it.
That power connector needs to be completely redesigned, who ever designed it got he decimal point in the wrong place when calculating the power limit of those tiny metal strips 😂😂
I’m hoping/thinking Zen 5’s IMC will be much more capable than 4’s. People are already running 8000 speeds beating out 6000-6400 benchmarks on certain boards. Hoping new boards come out to match.
One of the best repair channels, northwestrepair, has claimed that 4090s have an alarming failure rate- not because of connectors, but with ripped pads. The huge gpu core is heavy enough to rip soldered pads during shipping and oxidation during usage. Connectors are repairable, ripped pads aren't. I would be really wary of buying the even larger 5090.
In general, A mini refrigerator uses between 70-90 watts and 280-360 starting watts; An average household refrigerator uses between 90-180 watts and 360-720 starting watts; A side-by-side refrigerator uses 180-200 watts and 720-800 starting watts. *Which one are you going to buy?*
The no-win10 driver will drive more people to use Linux. Maybe I could run Win11 custom in a VM to play games that only work in Windows.. like Destiny 2.
Anyone willing to do the math on whether water cooling is now actually smaller now? As in the sheer space of the cooler on a stock 4090 vs the piping and water block cooler of an AIO? How big is a PSU anyway? Couldn't one be built into a graphics card (don't get me wrong I hate this but it's feasible right?)
Hi Steve. The 5090's size and weight will require a special case made by-guess who-NVIDIA. It will sport three support brackets: side, bottom, and top. The rear I/O slots will be made of Titanium to prevent any flexing.
Why would they give a 40% improvement if they can keep the buyer occupied with 10% for 4 years, as has been the message for 10 years now , as long as the buyer continues to buy that nonsense nothing will change
Watch people claim they won't get 50 series NVIDIA cards because of their business practises or poor design etc. Then amd making an anouncement they have to scale down their gpu departement because they barely sell.
We all have seen that meme which mocks gpus size with 8090 being a ventilation system on a random ass building. Im kinda afraid that that might actually be the case with the size when next gen gpus come out....
Will rtx 5XXX get bigger? Absolutely yes- It does have to have higher performance- NVIDIA does nothing to logic structures since 3xxx series... So... Yes
The mobile GPU market is mostly Asia and since they had restrictions on selling to china backing out and focusing on the desktop market makes more sense.
The founders edition 4090, as well some board partner cards such as Gallax, draw up to 666W and these all have ample cooling, so if the 5090 really is "only" 600W then the cooler should be no bigger. The reason the cancelled(?) 4090ti/ ada titan was pictured with such as massive cooler was because it was supposed to be a 1000W card
It would be a shame if the RTX 50 Series didn't use TSMC 3nm. 3nm would allow for a larger number of SMs across all chips. TSMC 4NP is only going to be a minor upgrade over the 4N that the RTX 40 Series uses. Obviously, we can expect significant performance gains from the more enhanced Blackwell architecture. However, the process node also tends to be a big factor in the performance of GPUs and CPUs.
MLID has been saying AMD target for zen5 is around 15% ipc uplift from zen4. that is probably server products. Laptop could also see lower performance(not ipc) due to 4nm and lower clockspeed gain. Desktop would be around 20-30% PERFORMANCE uplift in MT workload and 10-15% in ST depending on app.
Is that the best nvidia can produce? Gpus sucking as much power as a space heater? With the price of energy as high as it is.. efficiency is priority number 1 here. Performance to power ration is the most important factor.
I am actually interested in Battlemage and glad the desktop gpus are still coming. I plan to build a pc for a friend and I want to make sure it has a decent gpu without breaking the bank. He isn't tech savvy so paying some insane gpu price isn't ideal.
I'd rather have game graphics stagnate that gpu size, power draw, and fire hazards increase further. 4090 can already do 4k60 ultrawide Cyberpunk with 4 rays, 4 bounces and insane modded textures and reshade profiles. We literally DO NOT need more than that.
Nvidia is going down the same path as Intel: Push more power to get more performance. I think AMD will gradually win some significant market share in GPU, especially when CuPBoP and HIPIFY matures with the help of the open source community.
Zen 5 in the Ryzen 9000 is more of an architecture test then a performance monster - gaining experience with their new improved chiplet design and its adapted interconnects, the enabler for several generations to come.
From everything I've seen, no, they basically took zen 4 and fixed a ton of internal bottlenecks, polished some stuff up and just made an all around better zen 4. Zen 6 and 7 are the big redesign generation.
At this point they really require water cooling. That heatsink is just ridiculous. Also better power connectors would help (lol.) It may get to the point they could consider moving to a higher voltage standard if this keeps up.
I really wanted to buy the 4090 once upon a time. Had the cash saved up. But… the power requirements and the reporting of connectors burning scared me off. That is a lot of money invested just to to burn away…. I wanted it for AI. I’m gaming good with 3090 TI, but read 4090 is significantly better for AI, and for $1500ish, is consider quite cheap for AI card at home. Just gonna keep saving for A1000 instead, or whatever AMD has that could be good for AI.
No Windows 10 support?.... Out with your ancient OS, you must embrace adverts in your menus.... maybe at some point there will be preroll full screen adverts when you boot up... ;)
The RTX 4090 has a target power usage of 450W. The RTX 5090 has at least 50% more cores overall over the RTX 4090 with a little die shrink and higher clocks. I would guess that the RTX 5090 would be targetting around 550W. My hypothesis on the melting connector problem, the connector is failing successfully. It is not that the connector is designed bad, but the power distribution system from the connector is the real problem. The 12+4pin connector can work, but it requires the video card to load balance the 6 +12v lines. Without something to balance the load, you need a safety factor for the resistance and amp variation. Nvidia will need a more expensive power system on the cards to properly support 600W through that 12+4 pin connector with minimal safety factor. If Nvidia does not properly fix this, I expect the RTX 5090 series to get a mandatory recall.
Friendly reminder that a performance increase up to 7% is not noticeable by the average end user and that we all should wait for independent benchmarks before judging new hardware.
People should skip 50-series just becasue of the connector, user error or not..i bet you they not gonna cover the warranty either. I'm skipping until i see a new design..this one SUCKS.
Would it make sense that they fixed the connector for the 50xx series? Or is that just wishful thinking? It would make sense business wise, as two gens with bad fame of catching fire is too much. Samsumg's exploding batteries did a real bad effect on the brand. Anyway... it does not happen with the lower models (series 60 and 70). Mostly the 90s... And the 4060 is indeed very power efficient. Or did I misunderstood something? These are questions I wonder myself, truly...not making a joke, here.
aside from massive 600 watt power consumption, you gonna need another air conditioning units and to solve this issue, you gonna need to live inside the freezer to extra cool your head for the electric bills ahead.
Just because i can make a 1000w GPU, does not mean i should, you reach a point where slapping more cooler and more watt becomes a problem in more way than one. It is showing that NVIDIA has reached the limits of it's design if the only answer they have is more power and a bigger cooler. They need to come up with something new rather than beating this dead horse, no wonder we are getting fake frame generation tricks instead of better rasterization (raw performance).
I have an honest question. Why are people upset with 10% uplifts in a CPU i understand we want more, but isn't 10 percent better uplift now days then back in the day? Ex. 100fps to 110 fps is 10% and 200fps and 220fps is also 10% what am i missing?
truthfully , it's going to become bigger to a point where people will be forced to buy water blocks to cut down the size, it's already started with the EVGA 3090 TI fTW3 and beyond. 5090 and below will see the same fate by its consumers
Nvidia needs to ditch single connectors for their 450w+ parts, as heat increases, so does electrical resitance, forcing PSU's to increase power throughput to maintain the wattage that the GPU is requesting. Setting power, clock speed and voltage limits should help to mitigate some risk. As for 5090 sizing, they'll probably use a vapour chamber within the cooling block, which should be enough to keep temps down. If Nvidia is concerned, hopefully they can encourage consumers to properly optimise fan configurations and fan speed tuning to assist in dissipation of heat within the chassis.
The 90 Series is for tech people, not for gamers; stop dropping your life savings on the 80's and 90's, because games have 5-10 year release windows. Plan for that instead. imo
AMD Won't be doing Win10 Drivers for Strix APU as Win10 is dead only getting security patches and that ends soon, should be using Win11on new hardware like this. Also I believe it's common practice for AMD and Intel to send out preproduction parts that are held back so as not to tip off the compition as to what you really can do.
NVIDIA: Bigger, bulkier, more powahhh, more powahh draw! We don't care if you need a whole field-generator to power our GPUs, more POWAAAAH! Oh, and be sure to buy a separate house to put it in. But guess what, you'll be the guy with the BADDEST GPU IN TOWN! Isn't that worth it? Btw, don't worry about melting connectors, shit happens... besides, just buy our next GPU! We promise, it's worth it... *wink wink* AMD: You do you, buddy. Our next gen, we're putting a discrete-level GPU into our CPU's, for a fully functional APU beast. And it'll need about 150-200w. In total. Plus faster cache-access and GPU to CPU bandwidth, all of it being on a single stacked die. You may not have the 'baddest' GPU in town, but you'll certainly have the most cost-efficient and stable one. And it all fits in a shoebox, with no extra cooling needed. Id, Ego, and Superego. Freud would have a field day trying to analyse the mindset of most average PC consumers, buying overpriced, overpowered, oversized Nvidia bullshit these days. For what? Bragging rights? Getting an extra 50 fps past 150, even though it makes literally zero difference to the naked eye? Myself, I can't wait to see what AMD cooks up with their APUs. Giving the green monster a wiiiiide berth as usual. And the blue one, for that matter. Team Red FTW!
its not sad that Zen 5 doesn't have an 40% IPC increase. its stupid anyone ever expected that in the first place. so no. Zen 5 isn't worse than we thought.
Zen 5 will be better than zen 4. People just have their expectations far too high. However my 7800x3d will match the 9700x in gaming so no need to upgrade until the 9800x3d in 2025.
AMD set the expectations...
That said 10% uplift in single core performance is promising.
I mean if you compare Ryzen 5000 and 7000 it's pretty close to that. Although it's very rare to see huge performance difference on the same plateform so it may not be comparable.
@@rustyclark2356 could be 20% in multicore since single core was over 10%.
I think its gonna be good performance uplift.
A particular program using AVX512 gets a 40% boost, Xitter leakers & click bait reporters create daft hype.
IPC is just a claim, measuring a basket of selected programs at a specific frequency.
Why would anyone buy a 5090 if they can't stop the 4090 from Catching Fire??
ill have 2 x 4090 and never had any issues with it . and yes , ill buy 2 more 5090 just cause i need more vram power for AI , but nvidia should make the gpu around a pc case so we can plug the other components to it
Theres hundred thousands of rtx 4090, a few catching fire is nothing. Look at Tesla, toyota etc...
@@billlam7756 There is certain failure rate that's expected in anything. When you're this far over it you have an issue.
And these are designed way to close to their limits.
The failure rate is about 0.5%
Why scaremonger
People are dumb mate, don't underestimate their stupidity.
If NVIDIA continues this trend, eventually, you're going to need a 220 volt outlet to power that thing.
The "Fixing the 4090's melting power connector" video meme will turn into reality, yaaay 😂
I'm getting there. I just upgraded my gamer room wall outlet to a 20amp rated outlet using a single 12-gauge service wire connected to a dedicated 20amp fuse in my service box labeled "PC." Power-wise, I'm ready for my new 5090 GPU.
We are already at 200 with the 4090.
And a car radiator to keep it cool
@@rustygear9370 Piped straight into HVAC.
I dont get it. Power consumption efficiency is very valueable as a PC owner. If the 5090 is gonna take 600 Watt, no matter what power it holds...that is a massive increase in energy costs. Not to mention you will need a bigger PSU. And on top of that, it will probably release at around the $2000 mark. It is absolutely not worth owning such a GPU. Just red flags all over the place.
We need to decrease power consumption while also increasing the power. Be more efficient. Or at the very least, only increase power consumption slightly.
Right! Why can't they make their GPUs like Apple's M chips with higher power and much lower power consumption?
Only $2000? I suspect NVIDIA will play the game it did for the 4000 series--initially only release the top end card for an insane price, and limit the supply of 4000 series card to keep their price up. It's not as if they care about gamers--the real money is in AI.
@justsomeguy1074 uhm the M chips are arm based processors lmao.
And in hotter climates, you need to get that extra 600W out of your room via air conditioning, too; so you pay extra money for the AC unit's electricity, too.
In colder climates it's a an inefficient space heater at least. 😂
@@user-in2cs1vp6o I know but look at how much more efficient they are to a comparable Intel CPU/Nvidia GPU! The M chips also have GPUs so what difference does it make if they are ARM chips?
They may not be as fast as a dedicated Nvidia GPU yet but look how much more power an Nvidia card needs, it's getting ridiculous.
The year is 2040: Let me just get my trailer out for my GPU.
You gotta have an external generator!🤣
The 4090 was also rated for 600w initially. That's why all the 4090 coolers are completly overkill and keeps the card so cool, because they were designing them for 600w.
that makes sense. most of these cards stay below 70c at sustained 450w. absolutely bonkers.
I got curious, RTX2060 12gb, hitting it with the most intense game I have, 71C 1280rpm 165watts, 75F room temperature. A single small case fan and a power supply fan of about the same size to exhaust heat. The air filter in my room is still louder.
The Zen5 40% (probably less) is for AVX512 only from what I have read.
I will be happy with 10-15% and move to Zen5 when my 7800X3D runs out of legs in a decade or two....
Hopefully the 5090 will use dual connectors to reduce the power draw per line.
Or use thicker gauges like the older pcie cables
@@ogremair803 And actually having safety factor because SF 1.1 isnt much of a safety factor
I thought u wanna say "dual connectors to have double chance of caching fire"😂
Or just buy a card that isn't a massive waste of money and energy.
Its not the power draw, its the connector not firmly connecting. My 4080 Melted its connecter. Its not even supposed to draw 350 watts..
I had a nightmare last night about buying a 5080 and the conector melted and they wouldn't replace it!!!😢😢😢
Any dream where I can afford a 5080 is a sweet dream to me! 🤣
@@socialenigma4476 haven't upgraded since 2080ti
@@Sheet123 I'm doing 3090-)5080. Ur upgrade will be massive noticeable
@socialenigma4476 AMD GPU is better these days. It is different from GTX era
User error… so no replasements!
😂😂😂
Are they including a fire extinguisher with each card
Hope the do 😂 if not let's hope the have home insurance 🙏 😂
No… that will cost extra!
With those stupid connectors that might be a good idea.
If I remember correctly, 2nd gen Ryzen was only about 10% better than 1st gen and 3rd gen was about 20% better than 2nd gen. . The real gains weren't seen until Ryzen 4.
And the won't be massive enough from now one ther is a limit how much the can do it 😂
Current motherboard designs are already obscolete due to oversized GPUs that block PCI slots. I hope motherboard and case manufactures finally deal with this ridiculous issue by coming up with new form factor that finally separates out the GPU from conventional PCI slots. We need a new connector specifically designed for GPU data and power along with a new efficient way to mount GPUs to cases.
The case would look weird. Unless the mobo gets shortened and still fits inside the case unmounted somewhere. They'll have to get smaller eventually, they got new tech they're just milking this old tech dry first
Maybe a change in PC case layout, have the case more horizontal than vertical, double that bad boy as a monitor riser and then we won't have to worry about Nvidia breaking PCIE slots because they have a boner for AI.
@@Rowdy_Raider You mean, like PCs were in the 80's - late 1990s?
the tdp range is the same as existing 40 series products. So the coolers and their size wont change much. id love to see a trend in the other direction, but i think that wont happen any time soon.
RIP my pcie slots again. Permablocked by gigacards.
Use the included anti-sag bracket and you won't have a problem
@@AshtonCoolman not concerned about that, yeah those work well, I mean the pcie slots physically covered by the multi-slot width gpu’s make them useless.
@@Ultrajamzah, I gotcha
Iirc the watercooled ones are single or double slot
@kitalthevali that is correct, they are however quite a lot more expensive as well
I think we're at the point we need cooling systems interconnectors
Or just design them correctly.
Or maybe don't buy massively overpriced completely unnecessary wastes of energy from Nvidia.
😅 reminds me of Inno3D with a 4th fan near the power connector
@@R3TR0J4N i happen to own one.
High power draw is fine as long as it idles cheaply. My problem is the lack of balance for the consumer. Lovely GPU, but you're stuck with the 2.5 or 10G NIC on the motherboard because the top-end consumer CPU only has 24 PCIE lanes. Also, some of your motherboard NICs & SSDs will be bottlenecked through the chipset. Overspec'd GPU, but too bad about the storage and the networking and there are no spare high-bandwidth slots for anything innovative. 24 pcie lanes is too small. Threadripper 128 lanes is excessive for most use-cases. Where are the 42-64 lane options on high-end consumer CPUs?
I think the answer is: They're gone because they would destroy the low-end server / workstation market, but it is preventing the consumer market from doing anything interesting. 50/100G networking is impossible. Could we do displayport over ethernet? There's no way to grow a new idea in the consumer market.
Current cpus basically have 48 lanes of data bandwidth, most stuff is using gen 4 while the cpu and chipset support gen 5. That's why the x670 boards are so nice, uses gen 5 to the chipset so all the pcie off chipset is running at full native speed since it's connected to the cpu with gen 5 x4.
@@PineyJustice Thanks for the tip. I've been hunting for the PCIE4/5 equivalent of x79 but I don't think threadripper or Xeon W are cost effective, so x670e dual x8 seems to be the way to go. Cheers!
@@fanshaw runs 4x gen 4 m.2 drives at full native speed. That's a big reason I upgraded. Sadly only the really expensive boards have 10g nics.
Liquid cooled 2 slot GPU cards.
i think the cards will stay the same size and then people will just make watercooling them more standard or board makers may have to just make more watercooled options but i doubt they would grow that much over the 40 series
Heheh you fork over 3500 for that ?😮
There's a simple solution to the heavy card issue: desktop cases! The weight just goes into keeping the card in the slot, and the case brackets at the back support more of the weight from the direction it's supposed to go in. We seriously need to ditch tower cases and return to desktop ones. Also, more efficent for desk space, as you can put your monitor on top of it. Now, someone make a desktop case that has 6x 5.25" drive bays, so I have somewhere to put all my drives and I'll probably buy it.
Monitors on top of desktop cases is a huge ergonomic no-no. Eye level should be on the upper screen bezel. Show me how to do that with the typical 27"+ display.
@@marikann9073 I don't know how big that is, but if it's more than half a metre, it's too big.
Why would I want my eye level matching the _top_ of the screen? That'd just make you look down all the time. You want it raised up a bit so eye level matches the _centre_ of the screen, so nothing requires looking too far away from your rest position.
Well, that's what I do, anyway. I have my monitor sitting on some old books because it's too low for my liking when it's on the desk.
I was always hoping that they'd be able to slap a few heat pipes and a little fin (edit fam to fin) stack on the other side of the board to help dissipate some heat from the back. Maybe that'd lower the load on the front fins and pipes but I'm not sure.
I said it before, the wire and connector gauge is to small for the amperage that is being pushed through the connectors. Everyone said no, it's the way its connected, it will work now. Yeah, how is that working out for you now?
3:41 "Win10 enthusiasts please note that staring from Strix Point, AMD will not provide Win10 drivers."
That lowkey sucks.
I'd say it sucks full time. Windows 10 is supported by Microsoft for one entire year when Strix will launch. But, the upsides is that it will probably help Linux market share a bit more, as people ask what to do and other people start recommending things like Linux Mint. Still not an excuse for AMD.
It’s pathetic for them not to support a current OS.
why should AMD put extra work into Win10 drivers if there are no more security updates next year? Better put work into Linux or Win11
@@sylviam6535no it is not it is economically smart and on the other hand safer for the user
@@Psycho-Duck Oh, I don't know ... caring for your customers ? Because it's their job ? Because it's still a significant percentage of the total market share ? Because people will still using after the ENTIRE YEAR it will still be supported, just like Windows 7 was ?
The problem with nvidias power connector is that they are using cheap wire to push so much power through. Cheap wire increases resistance and thus creates more heat, which leads to melting over prolonged use.
nVidia card will soon be called video processor unit / VPU for short.
If this is true I see no improvement, just an extremely overtuned 4090 sold as a new card...
give me five! 🤑
😂😂 we heard these stories before the 4090 came out. Patiently waiting for the 5090 announcement and release
you guys buy up those new Zen 5 chips, i'll sell my 7700X for a 7950X3D and be totally satisfied.
At this rate, the GPU will be slotted on the 5 1/4 bay, with a cable extension to the PCIe on the mainboard.
Why would and GPU developer make a 4 lane wide card? How stupid! And it's certainly not going to require 600 watts! No one would buy it nor could they even use it.
For those couple dozen thousands that will pay the money to get the best GPU, the size or power draw won't be a major factor. If it can be done with off-the-shelve solutions, they'll do it.
I'm surprised no class action suit
If a GPU got that big, then using a GPU enclosure is reasonable.
I wonder if anyone is brave enough to create this thing for PCs, not for laptops (or they can support both), which is basically an "advanced version" of the PCIe riser.
Well Minisforum has many mini PC with oculink support, and once thunderbolt 5 comes out , there will be no penalty for using eGPUs anymore.
4090's cooler is for 600W too so no upgrade in size on that
That Lenovo dude will get a call from AMD and from Lenovo HR very soon.
That's totally Not why. My 80 melted and I made sure it was fully seated every time I got in it. It happened after a few really long sessions. Regular game play wasn't enough to melt it.
That power connector needs to be completely redesigned, who ever designed it got he decimal point in the wrong place when calculating the power limit of those tiny metal strips 😂😂
Meh, those old Zalman cooling towers are starting to make sense.
Why does the name of this video change every time one is released. My notifications say one thing and the actual video says something different.
Nvidia need 12v Anderson connectors on their cards
Love your tech updates
What do you need to power this thing?
Has anyone ever asked these people what are they doing with their 4090 GPUs? Mining, AI , overclocking or combo of these things?
The ones that need it for that always say don't get it if y don't do the same stuff like me and I'll say it 2 if only foe games wast of cash 💯
I’m hoping/thinking Zen 5’s IMC will be much more capable than 4’s. People are already running 8000 speeds beating out 6000-6400 benchmarks on certain boards. Hoping new boards come out to match.
One of the best repair channels, northwestrepair, has claimed that 4090s have an alarming failure rate- not because of connectors, but with ripped pads. The huge gpu core is heavy enough to rip soldered pads during shipping and oxidation during usage. Connectors are repairable, ripped pads aren't.
I would be really wary of buying the even larger 5090.
Far too much current for such small pins and contact area with the crimp. Go back to the old pin size and the issue will go away
In general, A mini refrigerator uses between 70-90 watts and 280-360 starting watts; An average household refrigerator uses between 90-180 watts and 360-720 starting watts; A side-by-side refrigerator uses 180-200 watts and 720-800 starting watts.
*Which one are you going to buy?*
When you assume you make an azz out of you and me... I wouldn't assume anything anymore with how these companies are acting.
The no-win10 driver will drive more people to use Linux. Maybe I could run Win11 custom in a VM to play games that only work in Windows.. like Destiny 2.
Anyone willing to do the math on whether water cooling is now actually smaller now? As in the sheer space of the cooler on a stock 4090 vs the piping and water block cooler of an AIO?
How big is a PSU anyway? Couldn't one be built into a graphics card (don't get me wrong I hate this but it's feasible right?)
Hi Steve. The 5090's size and weight will require a special case made by-guess who-NVIDIA. It will sport three support brackets: side, bottom, and top. The rear I/O slots will be made of Titanium to prevent any flexing.
5090 is bout to be a beast or a possible disater?
It's a fucking nuclear reactor
Beast
Both
That why i am worried about the 5090s. Will they fix the power connector by then!
RTX 5090 will primarily be a fire hazard with secondary computing capabilities.
Why would they give a 40% improvement if they can keep the buyer occupied with 10% for 4 years, as has been the message for 10 years now , as long as the buyer continues to buy that nonsense nothing will change
Be prepared to take out a 2 year lob to afford a 5090. The 4090 is unaffordable right now and I can only imagine how much the 5090 will be
It's the board breaking under the weight of the gpu itself...go look at fix it channels and they actually state why it happens
9800X3D will be nice performance jump from my 5800x, but, i'll see next year with Zen 6 what performace will be with 11800X3D.
Watch people claim they won't get 50 series NVIDIA cards because of their business practises or poor design etc.
Then amd making an anouncement they have to scale down their gpu departement because they barely sell.
We all have seen that meme which mocks gpus size with 8090 being a ventilation system on a random ass building. Im kinda afraid that that might actually be the case with the size when next gen gpus come out....
i am SO happy that i chose a 4080. I knew how much weaker the card was but was also heavily suspicious of this adapter. VERY glad i didn't get a 4090
Will rtx 5XXX get bigger?
Absolutely yes- It does have to have higher performance- NVIDIA does nothing to logic structures since 3xxx series... So... Yes
The mobile GPU market is mostly Asia and since they had restrictions on selling to china backing out and focusing on the desktop market makes more sense.
maybe AMD open source move can help games to work better on linux?
The founders edition 4090, as well some board partner cards such as Gallax, draw up to 666W and these all have ample cooling, so if the 5090 really is "only" 600W then the cooler should be no bigger. The reason the cancelled(?) 4090ti/ ada titan was pictured with such as massive cooler was because it was supposed to be a 1000W card
if 5090 is gonna need 600 watts, i expect it to have 300% increased performance from 4090
the AMD 9000 APU's aren't a CPU increase, but a GPU one. it's the RDNA that is the advancement.
The 4090 cooler design was created with 600w in mind. The 5090 will be the same size as the 4090.
Using only cinebench to determine IPC gains is ridiculous. Also, the IPC gains are for the entire chip, not single core.
It would be a shame if the RTX 50 Series didn't use TSMC 3nm. 3nm would allow for a larger number of SMs across all chips. TSMC 4NP is only going to be a minor upgrade over the 4N that the RTX 40 Series uses. Obviously, we can expect significant performance gains from the more enhanced Blackwell architecture. However, the process node also tends to be a big factor in the performance of GPUs and CPUs.
They kept cheap price soyo 6600m 8g is 80 usd. Does it have encoder?
MLID has been saying AMD target for zen5 is around 15% ipc uplift from zen4. that is probably server products. Laptop could also see lower performance(not ipc) due to 4nm and lower clockspeed gain. Desktop would be around 20-30% PERFORMANCE uplift in MT workload and 10-15% in ST depending on app.
Power conaumption isnt a big deal as long as you dont have an nshittia firestarter
all these NVIDIA problems is due to basic rules of wiring is "broken" aka you never use more then 80% of the wires/transfer points captivity
Is that the best nvidia can produce? Gpus sucking as much power as a space heater? With the price of energy as high as it is.. efficiency is priority number 1 here. Performance to power ration is the most important factor.
I am actually interested in Battlemage and glad the desktop gpus are still coming. I plan to build a pc for a friend and I want to make sure it has a decent gpu without breaking the bank. He isn't tech savvy so paying some insane gpu price isn't ideal.
nVidia would do well branching into space heaters.
Maybe they should include a gpu holder in the box at this point, considering they already cost a (or more than a) premium.
I'd rather have game graphics stagnate that gpu size, power draw, and fire hazards increase further. 4090 can already do 4k60 ultrawide Cyberpunk with 4 rays, 4 bounces and insane modded textures and reshade profiles. We literally DO NOT need more than that.
Nvidia is going down the same path as Intel: Push more power to get more performance. I think AMD will gradually win some significant market share in GPU, especially when CuPBoP and HIPIFY matures with the help of the open source community.
Zen 5 in the Ryzen 9000 is more of an architecture test then a performance monster - gaining experience with their new improved chiplet design and its adapted interconnects, the enabler for several generations to come.
From everything I've seen, no, they basically took zen 4 and fixed a ton of internal bottlenecks, polished some stuff up and just made an all around better zen 4. Zen 6 and 7 are the big redesign generation.
At this point they really require water cooling. That heatsink is just ridiculous.
Also better power connectors would help (lol.) It may get to the point they could consider moving to a higher voltage standard if this keeps up.
I really wanted to buy the 4090 once upon a time. Had the cash saved up. But… the power requirements and the reporting of connectors burning scared me off. That is a lot of money invested just to to burn away…. I wanted it for AI. I’m gaming good with 3090 TI, but read 4090 is significantly better for AI, and for $1500ish, is consider quite cheap for AI card at home.
Just gonna keep saving for A1000 instead, or whatever AMD has that could be good for AI.
No Windows 10 support?.... Out with your ancient OS, you must embrace adverts in your menus.... maybe at some point there will be preroll full screen adverts when you boot up... ;)
I am so glad that I got a 7900xt with 20 gigs of vram. Runs like a dream with the 7800x3D.
The RTX 4090 has a target power usage of 450W. The RTX 5090 has at least 50% more cores overall over the RTX 4090 with a little die shrink and higher clocks. I would guess that the RTX 5090 would be targetting around 550W.
My hypothesis on the melting connector problem, the connector is failing successfully. It is not that the connector is designed bad, but the power distribution system from the connector is the real problem. The 12+4pin connector can work, but it requires the video card to load balance the 6 +12v lines. Without something to balance the load, you need a safety factor for the resistance and amp variation. Nvidia will need a more expensive power system on the cards to properly support 600W through that 12+4 pin connector with minimal safety factor. If Nvidia does not properly fix this, I expect the RTX 5090 series to get a mandatory recall.
The 4090 should be recalled and Nvidia knows it!
600W GPU.. wow! I think I don't need heater for sure..
IDGFA about the size of the GPU, as long as the price is right and the gfx power is there.
Friendly reminder that a performance increase up to 7% is not noticeable by the average end user and that we all should wait for independent benchmarks before judging new hardware.
People should skip 50-series just becasue of the connector, user error or not..i bet you they not gonna cover the warranty either. I'm skipping until i see a new design..this one SUCKS.
Would it make sense that they fixed the connector for the 50xx series? Or is that just wishful thinking? It would make sense business wise, as two gens with bad fame of catching fire is too much. Samsumg's exploding batteries did a real bad effect on the brand. Anyway... it does not happen with the lower models (series 60 and 70). Mostly the 90s... And the 4060 is indeed very power efficient. Or did I misunderstood something? These are questions I wonder myself, truly...not making a joke, here.
aside from massive 600 watt power consumption, you gonna need another air conditioning units and to solve this issue, you gonna need to live inside the freezer to extra cool your head for the electric bills ahead.
With how massive GPUs have become they should design the graphics card to be the base for the motherboard, or make GPU risers the norm
Just because i can make a 1000w GPU, does not mean i should, you reach a point where slapping more cooler and more watt becomes a problem in more way than one.
It is showing that NVIDIA has reached the limits of it's design if the only answer they have is more power and a bigger cooler. They need to come up with something new rather than beating this dead horse, no wonder we are getting fake frame generation tricks instead of better rasterization (raw performance).
I have an honest question.
Why are people upset with 10% uplifts in a CPU i understand we want more, but isn't 10 percent better uplift now days then back in the day? Ex. 100fps to 110 fps is 10% and 200fps and 220fps is also 10% what am i missing?
Most so-called adults are just "man-babies", crying over everything that isn't 100% perfect.
SMH
truthfully , it's going to become bigger to a point where people will be forced to buy water blocks to cut down the size, it's already started with the EVGA 3090 TI fTW3 and beyond. 5090 and below will see the same fate by its consumers
Nvidia needs to ditch single connectors for their 450w+ parts, as heat increases, so does electrical resitance, forcing PSU's to increase power throughput to maintain the wattage that the GPU is requesting. Setting power, clock speed and voltage limits should help to mitigate some risk.
As for 5090 sizing, they'll probably use a vapour chamber within the cooling block, which should be enough to keep temps down. If Nvidia is concerned, hopefully they can encourage consumers to properly optimise fan configurations and fan speed tuning to assist in dissipation of heat within the chassis.
The 90 Series is for tech people, not for gamers; stop dropping your life savings on the 80's and 90's, because games have 5-10 year release windows. Plan for that instead. imo
AMD Won't be doing Win10 Drivers for Strix APU as Win10 is dead only getting security patches and that ends soon, should be using Win11on new hardware like this. Also I believe it's common practice for AMD and Intel to send out preproduction parts that are held back so as not to tip off the compition as to what you really can do.
Time for a mini fridge PC!
NVIDIA: Bigger, bulkier, more powahhh, more powahh draw! We don't care if you need a whole field-generator to power our GPUs, more POWAAAAH! Oh, and be sure to buy a separate house to put it in. But guess what, you'll be the guy with the BADDEST GPU IN TOWN! Isn't that worth it? Btw, don't worry about melting connectors, shit happens... besides, just buy our next GPU! We promise, it's worth it... *wink wink*
AMD: You do you, buddy. Our next gen, we're putting a discrete-level GPU into our CPU's, for a fully functional APU beast. And it'll need about 150-200w. In total. Plus faster cache-access and GPU to CPU bandwidth, all of it being on a single stacked die. You may not have the 'baddest' GPU in town, but you'll certainly have the most cost-efficient and stable one. And it all fits in a shoebox, with no extra cooling needed.
Id, Ego, and Superego. Freud would have a field day trying to analyse the mindset of most average PC consumers, buying overpriced, overpowered, oversized Nvidia bullshit these days. For what? Bragging rights? Getting an extra 50 fps past 150, even though it makes literally zero difference to the naked eye? Myself, I can't wait to see what AMD cooks up with their APUs. Giving the green monster a wiiiiide berth as usual. And the blue one, for that matter.
Team Red FTW!
When will the 5090 be launched tho