• 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.

Is this the end of the internal high-end GPU?

Soldato
Joined
1 Apr 2014
Posts
19,073
Location
Aberdeen
New high-end GPUs are hugely power-hungry. The 40 series is looking for 450W to 600W, maybe more. The form factor imposes considerable restrictions on cooling and noise control. This is a quandary. But we now have high bandwidth external connections like USB 4 and 100 Gb ethernet. So the question arises: does the GPU really need to be an internal device any more? Why not make the GPU its own device external to the PC? Not just a standard card on an adapter in a box like current eGPU solutions. Making the GPU external would allow for innovative design, cooling, and power solutions while removing the heat and noise problems from the PC.

It also frees up PCIe lanes in the PC.

Low-power GPUs can still go in the PC.

What do you think?
 
Consider the fact that SLi is dead, they are now using up the space where 2 or 3 GPU's could have been.
As long as you can exhaust the heat from the room your PC is in even in summer I think it's fine, nothing to worry about. If PCI space is an issue you would have gone with the appropriate setup to ensure you always have PCI availability such as those boards with 20 PCI slots.
 
Last edited:
They should put the graphics card slot on the backside of the motherboard then the GPU would just click onto the side of your case :cry:
 
The further the distance, the higher the latency. The further the distance, the more advanced and expensive the signal packaging required to make up the latency difference (if even possible). You can see the issues that arose for even pci-e 4.0 riser cables; don't want to think about what will be needed for 5.0.
 
The further the distance, the higher the latency. The further the distance, the more advanced and expensive the signal packaging required to make up the latency difference (if even possible). You can see the issues that arose for even pci-e 4.0 riser cables; don't want to think about what will be needed for 5.0.

I think you’re mixing latency with coherence. That was a problem as far back as parallel port printer cables. And with serial links like USB4 and Ethernet coherence isn’t an issue.
 
1000W should be fine, but what about all the recommendations of the new standard? Plugging 4 PCI-E into an adapter is quite scary, specially in a small case.
600W is quite a lot of power.
Spikes for the 3000 series was an issue. Apparently spikes for the 4000 series should be higher, one of the “issues” that the new PSUs should address.
 
1000W should be fine, but what about all the recommendations of the new standard? Plugging 4 PCI-E into an adapter is quite scary, specially in a small case.
600W is quite a lot of power.
Spikes for the 3000 series was an issue. Apparently spikes for the 4000 series should be higher, one of the “issues” that the new PSUs should address.

Nvidia states they have "fixed" the transient spikes.

MQw04Hs.jpg

Nvidia unveiled the PCB design and other data for the GeForce RTX 4090 on Wednesday. To maximize GPU overclocking, the business upgraded the card's voltage regulating module (VRM) by adding a PID controller with a feedback loop.


The PCB includes a single 12+4-pin 12VHPWR extra PCIe Gen5 power connector capable of delivering up to 600W, while the default BIOS settings limit the GeForce RTX 4090 FE to 450W. The board's 20+3-phase VRM (20 phases for the GPU, 3 for memory) is an improvement over the 18+3-phase power circuitry found in the GeForce RTX 3090/3090 Ti FE. Nvidia also incorporated a PID controller with a feedback loop to reduce current spikes and drops. Nvidia claims that their new VRM decreases power management reaction time without offering specific figures. A clean power supply is required for stability under severe loads and overclocking. The green team employs 'power transient management' to increase the speed of its AD102 graphics processor to 3 GHz without overvolting or using harsh cooling.

 
Last edited:
Let’s hope.
The only “good” feature for the 4090 so far is that most cases where people would choke their cards against the side glass panel, won’t be able to now.
 
I think it's the price that could kill the high-end PC market segment, especially so if it goes external. So you buy a good-spec PC, and then you pay double that again for an external GPU? For video games? It's not an easy sell.
 
Back
Top Bottom