They don't really need to - the relationship between transistor switching speed and temperature is very well-understood. You only need to evaluate the speed at a couple temperature points to be able...
Type: Posts; User: iddqd; Keyword(s):
They don't really need to - the relationship between transistor switching speed and temperature is very well-understood. You only need to evaluate the speed at a couple temperature points to be able...
Overclocking in general is sorta dying, in no small part due to cheaper and better profiling tools available at the foundry. 15 years ago, these tools didn't exist, and nobody really knew how good...
IDK about you, but I'm not that passionate about whatever hardware my IT department decides to buy for work. It's whatever, y'know?
Who cares? It's not for gaming.
Don't you already get AA for free if you're using the AI denoisers? Since, y'know, aliasing is a type of noise.
4-5nm is probably still doable with trigate finFETs - although, the 4-gate variety ("gate-all-around") seems like it might become necessary around those dimensions. You could also do GAA with a...
Not sure about elsewhere, but at least in aviation, the DoD aren't huge fans of designing ASICs; it's all a bunch of FPGAs. Source: I once walked out of an interview for a DoD contractor that did...
Guys, 10nm is only 6 months away!!!
-Brian Krzanich, 2015, 2016, 2017, 2018
Sir, are you aware that this website is xtremesystems.org?
I hope AMD didin't really want to hang on to the zen IP anyways, since it's definitely going to be stolen.
In other non-news, fire is hot and water is wet.
*Nervously looks around the room* That's right, we simply can't uh... produce enough HBM2 to meet demand. 'Fraid you're gonna have to pay a premium, boys. Totally not artificially restricting supply...
I wrote a paper about similar side-channel exploits when I went to school a while ago... with the Pentium 4's Hyperthreading implementation. Nice to see that it's still just as broken 15 years later.
Seems like a pretty standard NDA to me. I'm no fancy big city journalist, but I've signed a bunch of these things... and actually, now I can't tell you anything else, I'm afraid.
You'll be glad to hear they're currently facing a class-action lawsuit over trying to fleece their customers $700 for replacing a (poorly-designed) $25 part. The tried-and-true way of getting Apple...
Did nvidia forget that they're an ASIC company? They could just make an ASIC that does the same thing and costs $10.
When you owe the bank $2000, you have a problem. When you owe the bank $2,000,000,000, the bank has a problem.
Just so - if you're Nvidia, and one of your AIB partners built up an inventory of 300K...
Ironically, Larrabee had more in common with GPUs of today than what it tried to compete (and lost,) against.
solution: replace the evacuated dewar with a vacuum pump, and the full LN2 dewar with... oh hell, just pipe it directly to an LN2-producing factory
Well, the point I was trying to make is that leakage always increases as channel length decreases. You can come up with more optimal transistor geometries, but that doesn't really change the fact...
this owns; you might actually be able to run a LN2 overclock 24/7 with this thing - as long as you keep buying LN2
That's why they had to come up with finFET - with planar transistors, the gate leakage and sub-threshold leakage would be unreasonable. But in general, as you make the channel shorter, it is more...
Isn't necessarily better? That's an interesting way of saying "always worse".
base clock is completely meaningless these days, since your CPU will never actually run at the base clock. what you really care about is the all-core turbo.
Probably just a rebrand
