• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

NVIDIA 5000 SERIES

The one air cooling fans have been waiting for

They're testing CPU temps with two GPUs

GPU: RTX4090 FE
9800X3D temps after gaming session: 73c

GPU RTX5090 FE
9800X3D after gaming session: 95c


The 5090 FE's double flow cooler cooks your air cooled CPU guys


They also tested what happens if you stick two massive 200mm fans in front of the CPU for extra air flow, and then the 9800X3D temps dropped down a lot. In summary be careful if you buy the FE card - you need a heck of a lot of intake airflow in your case to overcome the heat the 5090 is blowing into your CPU

Yea, I posted that yesterday. It's crazy isn't it? 20+c difference!
 
The one air cooling fans have been waiting for

They're testing CPU temps with two GPUs

GPU: RTX4090 FE
9800X3D temps after gaming session: 73c

GPU RTX5090 FE
9800X3D after gaming session: 95c


The 5090 FE's double flow cooler cooks your air cooled CPU guys


They also tested what happens if you stick two massive 200mm fans in front of the CPU for extra air flow, and then the 9800X3D temps dropped down a lot. In summary be careful if you buy the FE card - you need a heck of a lot of intake airflow in your case to overcome the heat the 5090 is blowing into your CPU

That's the only tactics Jensen have to kill AMD....
 
DLSS 4 in Performance mode appears to have the same image quality as DLSS 3.5 in Quality mode. Transformer models are clearly the way forward. Hopefully AMD does the same, but for now they are using the older CNN models, as does the PlayStation 5 Pro

 
Last edited:
It's worse the higher up you go. Optimum's video showed MFG4x has much worse flickering and ghosting than the old mfg2x

Yep - the text commentary accompanying the vid says there was less glitching at 2x when there is particularly fast motion ("2X improves, but still glitchy" - when he is showing driving around on a motorbike).

So it's a potential trade off, although Optimum himself did ultimately conclude, before making that comment, that he preferred 4x "every single time".

Here's the vid for anyone that's interested (timestamp to the bit with MFG):

 
Watching GN's tear-down, I'm really starting to wonder why Nvidia insisted on a 2-slot design for the 5090.

Yes the cooler design is very impressive in what they have achieved within this form-factor but it's far from perfect and is noticably louder and hotter than its predecessor.
I'm not saying that slapping a 4090 cooler on it would have produced better results, after all the 5090 is consuming way more power with faster memory, but imagine what they could have achieved with this new cooler design if they'd just stuck to the same dimensions as the 4090.

The new design is only 13mm thinner than the 4090. That's 78% of the size.
Was it really worth it? Does anyone actually care about having a 2-slot rather than 3-slot card? The whole "SFF ready" marketing is also total crap as it's just as tall and long and still needs ventilation space around it.

Because of their insistence on a 2-slot design, we now have an FE that will run hotter and noisier than pretty much all AIB cards and has already been noted to run the memory alarmingly hot.
 
A random thought from an uninformed idiot. I have looked at the PCB layouts from Techpowerup for the FE, Astral, Gamerock & Suprim cards, and there is little difference in the VRM phase count and types of controllers used. At least when compared to the entry vs high end 4090 cards. For example, the GPU VRM phase count is 22-24, the Memory phase is either 6 or 7. The controllers used are near identical (Astral is slightly different).

Theory: The reference PCB standards set my Nvidia are so good that the AIB partners don't feel the need to go above and beyond this time.
 
Last edited:
Watching GN's tear-down, I'm really starting to wonder why Nvidia insisted on a 2-slot design for the 5090.

Yes the cooler design is very impressive in what they have achieved within this form-factor but it's far from perfect and is noticably louder and hotter than its predecessor.
I'm not saying that slapping a 4090 cooler on it would have produced better results, after all the 5090 is consuming way more power with faster memory, but imagine what they could have achieved with this new cooler design if they'd just stuck to the same dimensions as the 4090.

The new design is only 13mm thinner than the 4090. That's 78% of the size.
Was it really worth it? Does anyone actually care about having a 2-slot rather than 3-slot card? The whole "SFF ready" marketing is also total crap as it's just as tall and long and still needs ventilation space around it.

Because of their insistence on a 2-slot design, we now have an FE that will run hotter and noisier than pretty much all AIB cards and has already been noted to run the memory alarmingly hot.


Everyone was oohing and aahing over the video showing the cooler design and how parts of the pcb are connected by ribbon cables for airflow purposes....but it doesn't really seem to be upto the job.

They done a whole lot of unnecessary bs when they could have went with a thicker cooler. Yes the small pcb is impressive but it's like a condensed hot spot and memory temps aren't great either, regardless of the memory modules being able to handle more heat, it's preferable to have them cooler.
 
Watching GN's tear-down, I'm really starting to wonder why Nvidia insisted on a 2-slot design for the 5090.

Yes the cooler design is very impressive in what they have achieved within this form-factor but it's far from perfect and is noticably louder and hotter than its predecessor.
I'm not saying that slapping a 4090 cooler on it would have produced better results, after all the 5090 is consuming way more power with faster memory, but imagine what they could have achieved with this new cooler design if they'd just stuck to the same dimensions as the 4090.

The new design is only 13mm thinner than the 4090. That's 78% of the size.
Was it really worth it? Does anyone actually care about having a 2-slot rather than 3-slot card? The whole "SFF ready" marketing is also total crap as it's just as tall and long and still needs ventilation space around it.

Because of their insistence on a 2-slot design, we now have an FE that will run hotter and noisier than pretty much all AIB cards and has already been noted to run the memory alarmingly hot.
Probably because Nvidia is catering this for business and AI customers who don't use giant gamer focused cases.
 
Last edited:
Watching GN's tear-down, I'm really starting to wonder why Nvidia insisted on a 2-slot design for the 5090.

Yes the cooler design is very impressive in what they have achieved within this form-factor but it's far from perfect and is noticably louder and hotter than its predecessor.
I'm not saying that slapping a 4090 cooler on it would have produced better results, after all the 5090 is consuming way more power with faster memory, but imagine what they could have achieved with this new cooler design if they'd just stuck to the same dimensions as the 4090.

The new design is only 13mm thinner than the 4090. That's 78% of the size.
Was it really worth it? Does anyone actually care about having a 2-slot rather than 3-slot card? The whole "SFF ready" marketing is also total crap as it's just as tall and long and still needs ventilation space around it.

Because of their insistence on a 2-slot design, we now have an FE that will run hotter and noisier than pretty much all AIB cards and has already been noted to run the memory alarmingly hot.

Or the development was not for the 5090FE specifically but the first step into future endeavours. An early test bed.
 
I didn’t really understand GN’s admiration for how efficient the card is for its size. It could be the smallest card in the world but if it’s not cool and it’s noisy then really what’s the point…? Two steps forward, two steps back.
 
Last edited:
I didn’t really understand GN’s admiration for how efficient the card is for its size. It could be the smallest card in the world but if it’s not cool and it’s noisy then really what’s the point…? Two steps forward, two steps back.
Agree with you and the prior 4 or so comments on this topic. They could have kept the flow through design and kept it at 61mm (3 slot) like the 4090 FE and it would been considerably more effective i'm sure.
 
The one air cooling fans have been waiting for

They're testing CPU temps with two GPUs

GPU: RTX4090 FE
9800X3D temps after gaming session: 73c

GPU RTX5090 FE
9800X3D after gaming session: 95c


The 5090 FE's double flow cooler cooks your air cooled CPU guys


They also tested what happens if you stick two massive 200mm fans in front of the CPU for extra air flow, and then the 9800X3D temps dropped down a lot. In summary be careful if you buy the FE card - you need a heck of a lot of intake airflow in your case to overcome the heat the 5090 is blowing into your CPU


Exactly. It doesn't make sense when a lot of these gpu reviews are being done on test benches rather than inside cases. There can be no proper consideration of thermals
 
Everyone was oohing and aahing over the video showing the cooler design and how parts of the pcb are connected by ribbon cables for airflow purposes....but it doesn't really seem to be upto the job.

They done a whole lot of unnecessary bs when they could have went with a thicker cooler. Yes the small pcb is impressive but it's like a condensed hot spot and memory temps aren't great either, regardless of the memory modules being able to handle more heat, it's preferable to have them cooler.

Also nvidia have removed the hot spot sensor
 
A random thought from an uninformed idiot. I have looked at the PCB layouts from Techpowerup for the FE, Astral, Gamerock & Suprim cards, and there is little difference in the VRM phase count and types of controllers used. At least when compared to the entry vs high end 4090 cards. For example, the GPU VRM phase count is 22-24, the Memory phase is either 6 or 7. The controllers used are near identical (Astral is slightly different).

Theory: The reference PCB standards set my Nvidia are so good that the AIB partners don't feel the need to go above and beyond this time.

All cards draw about the same power so no need for significant VRM changes between models. HUB did say the FE card has much more coil whine and electrical noise than partner cards so there may be quality differences between models anyway.

All cards are limited by the single 12 pin cable, and all cards when overclocked can draw a sustained load over 600w, causing the 12 pin connector to reach 80c
 
Last edited:
Watching GN's tear-down, I'm really starting to wonder why Nvidia insisted on a 2-slot design for the 5090.

Yes the cooler design is very impressive in what they have achieved within this form-factor but it's far from perfect and is noticably louder and hotter than its predecessor.
I'm not saying that slapping a 4090 cooler on it would have produced better results, after all the 5090 is consuming way more power with faster memory, but imagine what they could have achieved with this new cooler design if they'd just stuck to the same dimensions as the 4090.

The new design is only 13mm thinner than the 4090. That's 78% of the size.
Was it really worth it? Does anyone actually care about having a 2-slot rather than 3-slot card? The whole "SFF ready" marketing is also total crap as it's just as tall and long and still needs ventilation space around it.

Because of their insistence on a 2-slot design, we now have an FE that will run hotter and noisier than pretty much all AIB cards and has already been noted to run the memory alarmingly hot.

Anyone putting the 5090 FE into a SFF is gonna see brutal results, I can't wait for the first tests because some of these cases have no airflow and the 5090 and the CPU is gonna run hot
 
Last edited:
Everyone was oohing and aahing over the video showing the cooler design and how parts of the pcb are connected by ribbon cables for airflow purposes....but it doesn't really seem to be upto the job.

They done a whole lot of unnecessary bs when they could have went with a thicker cooler. Yes the small pcb is impressive but it's like a condensed hot spot and memory temps aren't great either, regardless of the memory modules being able to handle more heat, it's preferable to have them cooler.
Ahem... We called this when images and videos of the FE weeks ago. You cannot break the laws of thermodynamics, and however much people fawn over liquid metal and thd flow design, it's basically just copper, aluminum and fans.
 
Theory: The reference PCB standards set my Nvidia are so good that the AIB partners don't feel the need to go above and beyond this time.
I saw the same PCB pictures and agree. Base TDP is 575 and max OC is 600W so far. So they are very close together. Hopefully we´ll see some Galax HOF 666W Bios some time.

At the moment I favor the Palit Gamerock as a base to build up a watercooled GPU with the announced Watercool Heatkiller. https://watercool.de/en/heatkiller-v-pro-ultra-for-palit-gainward-rtx-5090-graphics-cards/
 
Last edited:
Back
Top Bottom