Computer GPU/CPU/tech Mega Thread Based on new NVidia/AMD launches in 2025 and beyond. (Page 5 of 47)
|
I can run all the games in my steam library at 120fps with ray-tracing on my 2560x1440 monitor without issues or overheating using my 2080 Super in a high-airflow case. Are there any games coming out that really need much more than that? I have never noticed any improvement with graphics above 2560x1440 or faster than 120fps. |
Participation in the rights of citizenship presumes participation in the duties of citizenship
|
Originally Posted By CFletch: I can run all the games in my steam library at 120fps with ray-tracing on my 2560x1440 monitor without issues or overheating using my 2080 Super in a high-airflow case. Are there any games coming out that really need much more than that? I have never noticed any improvement with graphics above 2560x1440 or faster than 120fps. The modern games that are coming out are very hard on older graphics cards, especially with 8gb of ram. A good example is Indiana Jones and the Great Circle. Thats a very GPU demanding game. A 2080 Super with Medium setting/RT on in 1440p struggle to hit 60fps during some of the gameplay during testing, but needs 16gb of vram for full RT. Hell, that game recommends a 4090 to play in ultra settings 4k 60fps. But for most games, a 2080 is still GTG. Its just whats coming up now and in the future that may be an issue. |
|
Originally Posted By odiedodi: See the video above. It's all bullshit. Native rendering is actually even worse than I expected, and honestly my expectations were already tempered pretty low. Playing with fake frames will probably be an enjoyable enough experience, but there is no true "hi-fi" option in the GPU space, and at this point there probably never will be. The average consumer is okay with DLSS and frame gen. the simple fact of the matter is that path tracing is just not doable with current hardware in terms of pure rasterization, it is just too demanding |
|
Current performance figures show a weaker than expected performance uplift compared to 40 series. https://videocardz.com/newz/nvidia-official-geforce-rtx-50-vs-rtx-40-benchmarks-15-to-33-performance-uplift-without-dlss-multi-frame-generation Attached File |
|
Originally Posted By JTX23: Its looking more and more I might have to go with a liquid cooled version. Any advice or gotchas I should look out for? Make sure you have a place in your case that can accommodate the 360mm radiator while ensuring the tubing will be long enough to reach that position. The tubing length isn't as much of an issue if you use a vertical GPU mount with some adjustability and a riser cable with decent length. Also, I'm not a fan of orienting the radiator so that the tubing is on the top instead of the bottom, although some people are fine with this. |
|
Originally Posted By castlebravo84: Current performance figures show a weaker than expected performance uplift compared to 40 series. https://videocardz.com/newz/nvidia-official-geforce-rtx-50-vs-rtx-40-benchmarks-15-to-33-performance-uplift-without-dlss-multi-frame-generation https://www.ar15.com/media/mediaFiles/367135/blackwell_jpg-3435928.JPG It would explain the pricing for the 5070. |
|
Originally Posted By JTX23: Its looking more and more I might have to go with a liquid cooled version. Any advice or gotchas I should look out for? Have any of these AIO GPUs been user serviceable? Being unable to refill the coolant or having a dead pump render the entire card unusable because you can't replace it yourself would suck. |
|
Originally Posted By castlebravo84: Current performance figures show a weaker than expected performance uplift compared to 40 series. https://videocardz.com/newz/nvidia-official-geforce-rtx-50-vs-rtx-40-benchmarks-15-to-33-performance-uplift-without-dlss-multi-frame-generation https://www.ar15.com/media/mediaFiles/367135/blackwell_jpg-3435928.JPG As to the 5090, it is worse than that. It indeed has 33% more shaders than the 4090. However, see my prior post on the reduced core and boost clocks for the 5090. This generation is more about selling the DLSS software than it is an improvement in hardware, and it seems that the changes in chip architecture are primarily designed to provide support to DLSS. The 5090 does get more shaders and memory bandwidth that the 4090 in an attempt to meet consumer expectations for a flagship card, but either power or thermal limits caused by this change have reduced clock frequencies. As I noted above, if the limitation is thermal, the much larger surface area of the cooling solutions offered by board partner cards may significantly improve performance (assuming the BIOS is not nerfed). On the other hand, even using the full 675w that is theoretically possible with both the 600w cable and 75w PCIe power, the amount of power available to overclock is limited. I would say that Nvidia did learn from the 4080. The 4080 was unpopular because anyone who would spend $1,200 would spend $1,600 for the top card. The market segment below the XX90 series is presently at $1,000, as was shown by 4080 Super pricing. The only problem with the 5080 is that the RAM is too low, IMO. Nvidia could have charged a bit more for the card (more than the cost of the RAM upgrade) if there was 20GB or 24GB of VRAM. Supposedly these cards reduce the need for RAM, so we will see. Finally, after looking at more information released about the 5090 founders edition, I am disappointed and concerned that it will perform very poorly in many small form factor cases despite its stated purpose to accommodate such builds. It is going to pump a massive amount of heat right into the motherboard in many such cases. I previously misunderstood how the cooling was supposed to work. |
|
Originally Posted By castlebravo84: Current performance figures show a weaker than expected performance uplift compared to 40 series. https://videocardz.com/newz/nvidia-official-geforce-rtx-50-vs-rtx-40-benchmarks-15-to-33-performance-uplift-without-dlss-multi-frame-generation https://www.ar15.com/media/mediaFiles/367135/blackwell_jpg-3435928.JPG Weaker than expected? Really? Those numbers are about what I thought they were going to be. If anything, the 5090 has actually been a surprise to me, I thought it was gonna be worse. |
Twitter ID: @GattoSeh
|
Originally Posted By MeatBag: Have any of these AIO GPUs been user serviceable? Being unable to refill the coolant or having a dead pump render the entire card unusable because you can't replace it yourself would suck. Don't think so. If their PCB is any different then it's going to be impossible to find a good water-cooling block too. There is a place in my heart for AIO CPU coolers but I think if I were to need GPU cooling at that level custom would be the only way to go. |
|
Originally Posted By FLAJD11: As to the 5090, it is worse than that. It indeed has 33% more shaders than the 4090. However, see my prior post on the reduced core and boost clocks for the 5090. This generation is more about selling the DLSS software than it is an improvement in hardware, and it seems that the changes in chip architecture are primarily designed to provide support to DLSS. The 5090 does get more shaders and memory bandwidth that the 4090 in an attempt to meet consumer expectations for a flagship card, but either power or thermal limits caused by this change have reduced clock frequencies. As I noted above, if the limitation is thermal, the much larger surface area of the cooling solutions offered by board partner cards may significantly improve performance (assuming the BIOS is not nerfed). On the other hand, even using the full 675w that is theoretically possible with both the 600w cable and 75w PCIe power, the amount of power available to overclock is limited. I would say that Nvidia did learn from the 4080. The 4080 was unpopular because anyone who would spend $1,200 would spend $1,600 for the top card. The market segment below the XX90 series is presently at $1,000, as was shown by 4080 Super pricing. The only problem with the 5080 is that the RAM is too low, IMO. Nvidia could have charged a bit more for the card (more than the cost of the RAM upgrade) if there was 20GB or 24GB of VRAM. Supposedly these cards reduce the need for RAM, so we will see. Finally, after looking at more information released about the 5090 founders edition, I am disappointed and concerned that it will perform very poorly in many small form factor cases despite its stated purpose to accommodate such builds. It is going to pump a massive amount of heat right into the motherboard in many such cases. I previously misunderstood how the cooling was supposed to work. All of it is lackluster, but it is better than expected. At least from my point of view. Judging by how they are pushing the new Gen. Cards, I'm guessing they have some of the ideas from H100 and A100 cards built into them. If that is true then at least the new Gen. Cards will have sgignificantly better AI logic than the current 40 series cards. DLSS this, frame gen that, I don't care about those, Honestly, I think they are complete dead ends Unless they manage to get the input lag to negligible levels which I think will take a decade to do But the likes of reflex 2, frame warp and DLAA I see enormous promise with AI As well as for the classic game territory for AI taking primitive 1990's-00s textures and effectively remastering them. |
Twitter ID: @GattoSeh
Episches Leben
Joined:
Dec 2024
Posts:
807
EE: 0% (0)
OH, USA
|
Originally Posted By Archtaan: Anyone know when the partner board co's are going to be releasing pricing? 5090 review embargo is up on the 24th while the 5080's is up on the 30th (release date - strange) I think the 5080 review embargo date is designed to not draw attention to the 5090's reduced clocks. Everyone knows that the 5080 this generation is one-half of the 5090 as far as shaders and such (and price), and will be expecting that half performance. Given the clock differential, I think the 5080 is likely to perform more closely to 66%. It will likely will significantly outdo the 5090 on value-per-frame. |
|
Originally Posted By FLAJD11: I think the 5080 review embargo date is designed to not draw attention to the 5090's reduced clocks. Everyone knows that the 5080 this generation is one-half of the 5090 as far as shaders and such (and price), and will be expecting that half performance. Given the clock differential, I think the 5080 is likely to perform more closely to 66%. It will likely will significantly outdo the 5090 on value-per-frame. Bring back SLI.  My old 480 GTX pair used to function as a fantastic space heater. Performance benefit was... questionable. 
|
|
Originally Posted By Skoal7: PC Centric got his hands on a 5090, YouTube and did a review 2 days ago. With AI turned on 15ms more latency gaming than with AI created frames turned off. Snake Oil. The fake frame generation slows you down in the real world. Yeah I have no intension on using their magic. Half of me is considering sticking with my 3080Ti, but the other half wants a new card. I keep waffling on what, though. PROBABLY going to land on FE 5080. We'll see. |
Become prompt critical.
|
Originally Posted By JustinU235: Yeah I have no intension on using their magic. Half of me is considering sticking with my 3080Ti, but the other half wants a new card. I keep waffling on what, though. PROBABLY going to land on FE 5080. We'll see. Stock is probably going to be the biggest decider for me. I'm not going to play games again trying to land a card, I don't want one that bad. |
Episches Leben
Joined:
Dec 2024
Posts:
811
EE: 0% (0)
OH, USA
|
Originally Posted By JustinU235: Yeah I have no intension on using their magic. Half of me is considering sticking with my 3080Ti, but the other half wants a new card. I keep waffling on what, though. PROBABLY going to land on FE 5080. We'll see. 4080 Super makes my furnace run less heating my house. LoL 67c top temps but when it gets saturated with heat gameplay isn’t as smooth. I need to change coolant and improve radiator fans. |
|
Originally Posted By LittlePony: Weaker than expected? Really? Those numbers are about what I thought they were going to be. If anything, the 5090 has actually been a surprise to me, I thought it was gonna be worse. Some people were basing expectations on the 3090 to 4090 difference in performance. However, that transition was to a much smaller node, 4nm from 8nm. Both the 4090 and 5090 are 4nm. The 5090 gains much of its power from increased chip size. |
|
Originally Posted By FLAJD11: Some people were basing expectations on the 3090 to 4090 difference in performance. However, that transition was to a much smaller node, 4nm from 8nm. Both the 4090 and 5090 are 4nm. The 5090 gains much of its power from increased chip size. And since the 5090 is cut down it makes you wonder if they'll release a full die Ti variant in the future. |
|
Originally Posted By LittlePony: Weaker than expected? Really? Those numbers are about what I thought they were going to be. If anything, the 5090 has actually been a surprise to me, I thought it was gonna be worse. Well the 4090 was a HUGE increase in performance over the 3090 most were expecting a similar increase with the 5090. |
A tough man can take a bullet, but a wise man can dodge one. Stay focused my brothers.
|
Originally Posted By Klee: Well the 4090 was a HUGE increase in performance over the 3090 most were expecting a similar increase with the 5090. they are out to touch then, there is a world wide brain drain and competency crisis, no one is as good as they use to in most industries save for a few. expecting a 50% boost is unrealistic id say at this point. |
Twitter ID: @GattoSeh
Episches Leben
Joined:
Dec 2024
Posts:
822
EE: 0% (0)
OH, USA
He will have the testing done probably tomorrow: Worlds First RTX 5090 build on YouTube! |
Youtuber scuttlebutt is that the 5080 and 5090 supply is shit. Worse than the 3090 launch. Even Nvidia employees are saying little to no availability to their own employees. That blows if true and the bot scalpers are going to make bank  Nvidia RTX 5090 & 5080 Supply Leak | RX 9070 XT Benchmarks | AMD FSR 4 Support for RDNA 3 |
|
Originally Posted By LittlePony: this reminds me, if Nvidia really wanted to show there stuff in the game world they would make a new Titan. they already have 80gb cards to why the fuck not. oh sure it would be like $5000 but it would prove a point. The 90 series basically is the Titan. Nvidia fucked up their card naming scheme after 20xx series introducing a 90 series card and dropping the Titan name. Then in 40xx series they fucked up by having two 4080's with different VRAM numbers. They had to 'correct' that by renaming the lower tier 4080 as the 4070ti. They need regular cards for a new generation to launch with no ti bullshit and then the mid gen refresh be the ti like it used to be. Now, with how badly they fucked the naming convention up, most people have no idea what they're getting unless they do a lot of research. 'ti Super' as an example should never exist. 4080 Super should not exist. Should be 4080 and 4080ti. |
I'm not the one REEING, motherfucker! -FCSD2162
|
Originally Posted By Archtaan: Youtuber scuttlebutt is that the 5080 and 5090 supply is shit. Worse than the 3090 launch. Even Nvidia employees are saying little to no availability to their own employees. That blows if true and the bot scalpers are going to make bank  https://www.youtube.com/watch?v=bZ6NeSGad4IAlways a good idea to wait a few months or even 6 months anyways. Let issues get ironed out and avoid the early adopter bullshit. |
I'm not the one REEING, motherfucker! -FCSD2162
|
Originally Posted By Rebel31: The 90 series basically is the Titan. Nvidia fucked up their card naming scheme after 20xx series introducing a 90 series card and dropping the Titan name. Then in 40xx series they fucked up by having two 4080's with different VRAM numbers. They had to 'correct' that by renaming the lower tier 4080 as the 4070ti. They need regular cards for a new generation to launch with no ti bullshit and then the mid gen refresh be the ti like it used to be. Now, with how badly they fucked the naming convention up, most people have no idea what they're getting unless they do a lot of research. 'ti Super' as an example should never exist. 4080 Super should not exist. Should be 4080 and 4080ti. the 5090 does not, feel, like a titan moment to me ... or at least not like a EVGA Signature Titan. but yeah their naming is not as concise as it use to be |
Twitter ID: @GattoSeh
|
Originally Posted By Rebel31: Now, with how badly they fucked the naming convention up, most people have no idea what they're getting unless they do a lot of research. 'ti Super' as an example should never exist. 4080 Super should not exist. Should be 4080 and 4080ti. It is Super (  ) confusing. Then you have AMD coming up with a completely different naming scheme every other generation.
|
|
Originally Posted By Ayada: I really wish AMD would trade blows in the 5080 range again. Originally Posted By Ayada: Originally Posted By HecklerKac: Keep buying Nvidia cards...makes me rich. I really wish AMD would trade blows in the 5080 range again. AMD's N48 lineup seems like a very good match against the RTX5070 and 5070Ti. If AIBs get free reign with aftermarket cards/coolers and push N48 to 350-400w, they might be pretty competitive against the RTX5080 in the $600-$700 range. It seems like AMD's weird last minute changes are because they were surprised by how well they compete with Nvidia's lineup. That sucks because supposedly Lisa personally canned a ~600mm2 design that was being worked on after the chiplets were cancelled. I could easily imagine a 2.2-2.5ghz +120CU design with 28Gbps GDDR7 on a 384bit interface that could easily do +40% over N48 at ~500w. Even if they couldn't release it until Q2, it would have been a solid offering... price the cutdown version at $900 and the bigboy at $1200. |
|
Originally Posted By Archtaan: Youtuber scuttlebutt is that the 5080 and 5090 supply is shit. Worse than the 3090 launch. Even Nvidia employees are saying little to no availability to their own employees. That blows if true and the bot scalpers are going to make bank https://www.youtube.com/watch?v=bZ6NeSGad4IThe dude loves to blow his own horn when he is right. |
A tough man can take a bullet, but a wise man can dodge one. Stay focused my brothers.
|
Originally Posted By Kingdead: I'm ok with 1080p high frame rate gaming. My 2070s system will be updated sometime in 2025. It will be a big jump but I hope to last out until then. 1440p is all i wished for, i have no reason to go passed 2k, if i can get 120fps on 2k with any of my future settings im set for life. |
Twitter ID: @GattoSeh
 Embarrassing RTX 5090 Benchmarks, 9070 XT Hype Intensifies - Tech News Jan 19 2025 |
A tough man can take a bullet, but a wise man can dodge one. Stay focused my brothers.
|
Good video. I like that guy, subbed. |
Become prompt critical.
 RTX 5090 - Not Even Here and People are Already Disappointed |
Become prompt critical.
|
The size comparison with the 3080 was surprising. It makes 3080 looks downright tiny. |
|
Originally Posted By MeatBag: I wonder if the 5060 really will only be an 8gb card. Seems probable given that the standard 5070 is still 12. Have they released the standard and Ti versions at the same time before? I have RX 580 video cards that are 7 years old that came with 8 gigs. 
|
A tough man can take a bullet, but a wise man can dodge one. Stay focused my brothers.
|
Yeah, because it's shit. It's also the best consumer grade card money can buy, but at $2,000, it doesn't make much sense. It really ought to be the same price as the 4090, and even then it still would be a mediocre generational leap, but at least a fairly priced one. I'd like the additional ram, but I otherwise won't feel like I'm missing out on anything by getting a 5080, if I'm going to have to use multiframe generation and DLSS either way. The only way that it really starts to look appealing is if the latency on the lower teir cards is substantially worse with multiframe gen (this could very much turn out to be the case), but if that's the case, then the entire product stack is practically shit.
|