Official NVIDIA GeForce RTX 40 Series & H100 Thread

Any reason to keep changing cards or just for fun?
This time I didn't bother at all and am pumping the OG day 1 TUF DUD 4090. :bleh:
 
So, the CableMod 90 and 180 degree adapters are up for pre-order now. They are $39.90, so not exactly cheap, but I guess they are filling a need. Shipping is free if you get economy (12-21 days), or $15 for standard (3-7 days).

The order page was a little confusing since standard is selected first and economy option is just blank versus stating 'free' or '$0.00', so at first I thought you had no choice but to pay $15, which would put the cost at over $50.:nuts: That was really pushing the limit of what I was willing to pay for an adapter. Fortunately, it turns out economy is actually free, so it's only $39.90 if you're willing to wait.

Interestingly it seems they didn't collect any tax on the purchase either.
 
KAC said:
Any reason to keep changing cards or just for fun?
This time I didn't bother at all and am pumping the OG day 1 TUF DUD 4090. :bleh:
Click to expand...
Mainly just for fun and to figure out which AIB card had the best memory temps which ended up turning out to be the tai chi.

Been staying with the gigabyte OC 4090 for now and I don't see myself switching around again if the Ti card comes out. Don't need anything more than this and earliest upgrade would be dependent on what the next gen ends up looking like.
 
Nascar24 said:
4090's work fine on 850W power supplies.
Click to expand...

I replaced my 1000W EVGA with a 1000W ROG THOR with the power draw display on the side and the highest I have seen it go was 749W and that was running Port Royal at +106 +760. I was able to move the power slider back up to 133 with no issues and I also noticed that instead of being all over the place, the GPU was running at 2910mgz constantly under load. with no fluctuation during any run of Port Royal. Put me over 27000 for the first time. Jedi Survivor was maintaining more consistent frames as well.
 
Man last time I used anything for my main rig that was less than 1050 watts was in 2013. Been on mostly EVGA 1300 Watts for about 8+ years now.
 
KAC said:
Man last time I used anything for my main rig that was less than 1050 watts was in 2013. Been on mostly EVGA 1300 Watts for about 8+ years now.
Click to expand...

I was looking at the EVGA 1600W and the platinum 1200w from Corsair but both required buying a cable for $30 and $40 so I decided to wait on a PCI-e 5.0 PSU and when I saw an opportunity to pick this one up at an insane price, I read some reviews and I think the Tom’s hardware one had it stable and efficient still at over 1360W so I figured I’d give it a try and with the real time display, I would be able to see if I needed more. Turns out, it didn’t need the extra power, just a better delivery of the power. Looks so clean now with the single cable running to the GPU too.

Honestly, I’m really disappointed in EVGA for a 1000W not being that stable in a system that is pulling less than 750W.
 
Ran across a lab analysis of the connectors for problems of the 4090 that is interesting. Very detailed lab analysis. TLDR: connectors have too much malleable copper instead of being tougher (brass) and may not have enough surface contact area even if inserted correctly on bad headers.

https://www.igorslab.de/en/smoldering-headers-on-nvidias-geforce-rtx-4090/2/

Now, it would be tedious to weigh electrical conductivity against mechanical durability, but the PCI SIG certainly did not make these material specifications without reason. From an electrical standpoint, copper is probably the best solution, but brass, depending on the zinc content, is harder and also has a higher shear strength, tensile strength, and also a higher yield strength. The fact is that pure copper bends, compresses, stretches, and twists much more quickly, which can become extremely dangerous with these connectors.

The headers tested here from the defective cards are thus, most likely even with good intentions, relying on the wrong material! Please remember this fact until we come back to the causal connection in which pin width, twisting, positional tolerance, and clamping surface, unfortunately, stand.
​...
his very process is pivotal for the quality of the square wire and its dimensions, and along with the electroplated coating and its tolerances, the final outer dimensions are determined. We will soon see a pin whose wire was not perfectly rolled and still shows remnants of the round wire’s curvature. If such a wire is processed anyway, it means the Quality Control Inspection (QCI) of the connector manufacturer has failed. And we also recall the relatively soft copper and its mechanical disadvantages compared to a more suitable alloy.
...

The near-ideal pins of the 3rd party headers

Yes, they do exist. The pins in the headers from CableMod (at the adapter output to the power supply) all conform to standards regarding pin strength, alignment, and positioning. It can be done when the header manufacturer really wants it to be so! This should be emphasized, because aside from the tolerances PCI SIG allows for emergencies: The required 0.64 mm should be the norm, not the exception! However, CableMod also uses headers with pins made of pure copper instead of brass. But the wire used could not be more precise, and even the edges are only slightly rounded. I tested the headers from a total of 18 unused CableMod adapters from various batches and found the same, nearly perfect measurements in all 12 pins of each header.
​
...

The pins on the original headers: Fishing in the tolerance range

I have also examined headers from Asus, because I had many of them, providing a certain statistical reliability. The wires used are often uneven, with slight bulges (compression) instead of straight edges and also often uneven dimensions. Yes, it’s all in the micrometer range, but when it comes to borderline high currents, the effective usable clamping surface is important and it certainly suffers from such deformations.
​...
The measurements here are the absolute maximum values, which again show that the contact surfaces of the pin are partially even slightly bent and not truly flat. Let’s quickly recall the sentence at the beginning of the page about round wire and the problems in manufacturing.

And another thing: I have not been able to find a pin with 0.64 millimeters or more in outer dimensions that would have been damaged in this way! It was always only the thinner ones, with 0.63 mm and significantly less, or those that were not really rolled smoothly!
​...
In fact, there are pins that are twisted internally! A 3-degree deviation of the axes from the vertical on this non-charred pin is already significant. This brings us back to the crimped cable plugs. Such a spring contact, since it sits loosely in the housing, can absorb and compensate for this twist to a certain extent, thus adjusting the positioning. However, if the counterpart is fixed-wired, such as a true socket like the one we find on the CableMod adapter and whose pins are soldered onto a PCB, no compensation can occur, and the clamping surface is also significantly minimized in part because the contacts are askew and therefore do not lie flat completely.
...

[h=3]ntermediate Conclusion[/h]
It is quite obviously the sum of pin size, positioning, and possibly twisting that negatively affects the contact surfaces (“clamping surfaces”). Such a borderline construct in terms of currents to be delivered demands the highest manufacturing quality to really be considered safe. And it is not acceptable that I can short circuit three CableMod adapters in series at a limited 600 watts on the laboratory power supply and nothing happens, while a single inferior header is enough to cause damage to the expensive graphics card.

I can only advise manufacturers of these connectors to adhere exactly to the specifications and to check this continuously and more often within a series. Yes, these are higher costs, but the customer will surely appreciate it. Part of the problem certainly lies in the wire used for the pins, as the incoming goods must be inspected more closely by quality control from the start. Since there are many manufacturers of connectors, the quality of the products varies greatly. While the plastic housings were all cast very correctly, it becomes more problematic with the square wire and its initial quality.

...

This concerns both the material, i.e., the appropriate alloy instead of pure copper, as well as the definition of the galvanization layer thicknesses, the processing of the wire pieces (cutting, bending), and their positioning (pressing) in the housing. In addition, I miss a strict specification for the minimum contact area and spring pressure on the spring contacts. All these are important factors that are primarily essential for the right connection and should not be left to chance or to the individual manufacturers (which amounts to the same thing).
Click to expand...
 
You must log in or register to reply here.
Back
Top