Originally Posted by
Kagthul
Admission of your error by ad-hominem attack duly noted.
I didn't say a damn thing about wether they patched it or not. I said AMD broke the street date on the embargo, which is a plain old fact.
You were the one being "AMD was just tell da troof man!" and i pointed out that ... yeah, they were not, in fact, telling the truth.
Literally every CPU under the sun is vulnerable to SPECTRE.
ARM, Intel, AMD, IBM (POWER, formerly PowerPC, which has about a 60-70% market share in big iron), you name it. Everything.
The ONLY saving grace about SPECTRE is that the parts that cant be patched around require physical access to the machine to implement. Otherwise itd be a total shit-show (for -everyone-) from start to finish.
- - - Updated - - -
This goes to show why i both generally keep you on ignore (unfortunately, other people quoted you... bleh) and how a little knowledge makes dumb people dumber.
It has NOTHING to do with wether a game is "CPU heavy" or not. NOTHING. It has to do with what type of memory operations the software uses. Full stop. If it doesn't do a lot of memory swaps into the affected areas, it wont be affected at all. Not one bit. If you DO make a lot of operations like that, itll affect you a lot. Virtualization and server tasks make these swaps A LOT, so those areas (may) be heavily hit. I think Epic showed that their back-end server took a pretty severe hit after the patch.
Due entirely to the fact that WINE is virtualization. That simple. Nothing to do with "CPU intensity".
About as long as AMD did, as well. They aren't un-guilty here. Their CPUs of the same era use the same technology. The only reason that later architectures aren't vulnerable to MELTDOWN is because of the janky way they cobbled together their CPUs by stitching together multiple CPU dies. Same with Ryzen; the Infinity Fabric is what makes them not vulnerable, because of the way they had to staple multiple full-up CPU dies together to create the core counts they wanted on Ryzen. That's not a dig - with FX-series chips it was a failure but with Ryzen they got it working right. But it wasn't some intentional way of making their CPU not vulnerable to this since they had no clue this kind of vulnerability was even possible when they started to design Ryzen. Its basically a happy accident.
Uh.. K. Even if my CPU -DID- take a flat 30% performance hit.... it'd still be as fast or faster than a Ryzen chip. Since i definitely haven't taken anything near a 30% performance hit (or any perceptable performance hit at all; i dont have an pre-patch benches to compare to, unfortunately), im still doing better using Intel than AMD.
If you consider "fully vulnerable to the more dangerous of the two exploits" to be "minimal effects", sure. SPECTRE is far more dangerous than MELTDOWN. MELTDOWN can only read protected memory; SPECTRE can actually inject code.
How screwed am I? I'm not. Most people aren't. They aren't affected at all. This may affect their server division (heavy virtualization and server loads take the only significant hit)... but i hate to break it to you, most people wont turn to AMD to fix that problem. Theyll turn to POWER. A huge (majority) portion of the server infrastructure world-wide is already using POWER and has been for a looooonnnnggg time. POWER isn't vulnerable to Meltdown that i'm aware of, but it is fully vulnerable to SPECTRE. If they dont need a high-power server solution, theyll turn to ARM instead.
For.... what? Producing a product that is vulnerable to something that was so hard to find it took over 10 years to find it? Its not like theyve known all this time; they just found out last summer.
It does work as designed. Just like every CPU with branch prediction. Its why even some ARM chips (only a few lines) are also vulnerable to MELTDOWN. NO ONE in the CPU world thought this kind of thing was possible. Not Intel, not ARM, not AMD, not IBM. AMD's chips are immune to Meltdown by a happy accident, not deliberate design. The guy who designed Ryzen for AMD was pretty shocked when he found out. (Hes also done work for Intel and is responsible for a lot of their chip designs or large parts of them). So this guy who designs these cutting edge chips, had no idea this kind of thing would be possible.
... and that shows how out of fucking touch Linus is. The two major ARM64 lines used for servers? Yeah, vulnerable to MELTDOWN. And SPECTRE, like every other CPU in the world. Though the newer revisions of the ARM64 lines coming out mid-late this year should be immune to MELTDOWN.
And in what world is an Intel CPU "shit"?
Even if they took a flat 30% hit, most of the server CPUs Intel sells would still outperform their Ryzen-based EPYC counterparts because of basic IPC gains and higher clocks.
So, absolutely the worst-case scenario.... total worst case.... its still just as good as the competition.
Huh. Now, you'd have a point about buying EPYC chips because they're cheaper for the same performance at that point, but again....
They wont be looking to switch to EPYC if they leave Intel. Itll be to a real big-boy chip series like POWER if they need big-iron muscle or (if they dont need that kind of heavyweight) theyll be looking to ARMs newer server chips. Which are quite a bit cheaper, and less power consuming, and working in large clusters can perform just as well.