Recent content by TurboGLH

This is the type of comment that I'm talking about. 1/10 of 1% sounds ridiculously low to almost anyone, but in discussions about safety factors and failure modes, it's not. Tsumi said it above, engineering controls vd administrative. A robust design would NOT fail in this manner, even when...

This has been covered ad nauseum, in the end a connector that lacks the robustness to be used by the average consumer, including being used incorrectly, is a bad design. People who work in fields where design issues can cause the loss of life understand this. People who work the front desk at a...

That doesn't make sense, why would they replace the connector for something that is 100% user error and not at all related to a substandard design.

What was the point of arguing if you weren't going to read anything that was shared? The FTAIA made changes to the Sherman antitrust act, it's main purpose was to improve foreign trade by limiting it's use. But it still allowed the US to apply legal action against companies for actions in...

That's absolutely incorrect. Nvidia is an American company and is subject to American laws everywhere in the world. Now, they might try and limit it to the Chinese market, so maybe it has no direct effect on us consumers, or maybe it doesn't violate antitrust laws. But if it did, 100% able to...

Remember everyone, pci-sig will redesign power connectors based on "user error" alone.

A design that allows for user error to cause a failure like this is a bad design, that's what you're not getting. A robust design will allow for some error while still functioning or failing gracefully, which in the case of a gpu would look like excess power draw without enough heat to cause...

Remember that every connector can fail if used incorrectly, which explains the v2 versions of the ATX, SATA, pcie 6/8 pin and good old molex.

Yes. They move the reference plane from the rear of the connector socket to the front, then made the power pins slightly longer while also moving the target contact point out by a 1.05mm AND moving shortening the sense pins 1.25mm at the same time. This requires you to insert the connector an...

Did you do that with the 24pin, 8pin CPU or any other connections? If no, then you're acknowledging that you have more concern about the new connector than you do about the older designs.

Pretty sure I'm on the block list, but regardless. This is the argument I'm talking about. It doesn't help prove it's a good design, but instead helps the argument that it's a bad design. A robust design fails gracefully, a bad design does not. The connector here does not give good feedback...

100%. The number of people throwing themselves on the sword to defend this design is nuts. A robust design MUST be idiot resistant, if it allows for normal users to make mistakes that have dangerous consequences, it's a bad design.

I think we're at two different meanings for availability. I think there will be more 7900 cards available overall. You mean that they will sell well at $999 and will be harder to get. Both are valid, and can be true at the same time.

Typical hot take from zerobrains. AMD is TSMCs second largest customer behind Apple, and only has two products on 5nm, zen4 (which is selling just fine in enterprise) and rdna3. Zen3 availability was great for almost all of covid, except a spike early on when scalpers thought there was money...

I think the pricing is off on the 7900xt, at least relative to the xtx. To keep the same bus width and ram count, your bom cost has to be the same as the xtx, you would have all 6 MCDs with the same core die. This negates one of the primary benefit of chiplets, the ability combine cheaper cores...