« first day (3730 days earlier)      last day (44 days later) » 

12:09 AM
@Hennes Well aware of that, consumer processors top out at four ranks per channel (two ranks per module). With server processors that support registered or load-reduced memory, you could have eight ranks on a single DIMM and 16 ranks per channel.
@Ramhound That typically means the click switch is failing (it's not reliably maintaining contact). Usually, that means it's time to get a new mouse.
 
12:37 AM
Was restoring a 2011 AMD HP laptop that I've had laying around for a while. There's quite a bit of smoke and mirrors on its stickers, which tout it as this amazing CPU that has quad cores and "Ultra Radeon" graphics. Just did some searching and found out that it only performs as well as a Pentium Dual-Core of the era in multithreaded tasks, and the Pentium beats it in single-threaded tasks.
 
 
@bwDraco Once I had to replace my mouse because it kept registering two clicks on one physical click randomly. I opted to replace it with a more reliable mouse, an opto-mechanical one
 
Eh, I had a Roccat gaming mouse wear out after 3+ years of heavy daily use.
But price is not a guarantee of longevity. It's true that high-end gaming mice are generally designed to withstand heavy usage by eSports players, and the same is true for premium productivity-oriented mice like the Logitech MX Master series. But some (like the original Logitech Anywhere Mouse MX) had poorly-designed click switches that failed prematurely. (I've dealt with this firsthand. Newer MX mice are far more reliable.)
Similarly, we've had cheap mass-produced mice (of the sort that comes with your typical desktop) last over a decade.
@gparyani Phenom II? These chips weren't exactly bad but they weren't that good, either.
My first laptop had a Phenom II N930.
It did not help that the processor had no L3$.
It's AMD Family 10h adapted for mobile.
Fast forward to 2020 and AMD now has these wicked-fast eight-core laptop processors. They can't make enough of them.
But yeah. AMD used to just suck.
TBH it's a bit of a miracle that they survived with this crappy Bulldozer architecture.
 
1:13 AM
@bwDraco A8-3500M
 
@gparyani ...ah. That generation was still K10-based.
And when the next generation of AMD processors came out, it turned to be a shitshow.
Horrendously inefficient, and AMD's vision of GPU compute as the primary driver of floating-point performance working in conjunction with lots of small integer cores never really came to fruition.
 
But that HP Compaq machine with a K8 CPU I was talking about earlier, turns out the competition from Intel wasn't that great either. The fact that it's 64-bit means I can install modern Lubuntu on it (the latest release is 64-bit only). Intel's competition was still only 32-bit at the time (Core 1), and the Core 2 64-bit CPUs that came out a couple months later still only had the same performance level.
 
@gparyani Eh, Pentium 4 was crap, too. Intel was chasing clock frequency, even at the cost of IPC.
Core 2 Duo had something in the ballpark of 90% more IPC than Pentium 4. Seriously, almost double the IPC.
 
I've seen NetBurst compared multiple times to PowerPC G5
 
Intel basically cheated to prevent AMD's superior Athlon 64 processors from gaining a foothold, which ultimately resulted in a big antitrust lawsuit.
 
1:21 AM
Similar things happened for both. Apple's laptops continued to use older G4 CPUs with clock upgrades, and the Pentium M mobile CPUs used the older Pentium III microarchitecture
 
And guess what? Today's Intel Core processors ultimately trace to the Pentium M, rather than the absurdly inefficient NetBurst architecture.
 
It seems like the Pentium 4 is what drove Intel to rebrand Pentium as a low-end CPU line and push it away from their flagship line
 
Yup.
 
Pentium Dual-Core CPUs were branded as having similar performance to Pentium 4's, and the Core series had higher performance
 
FWIW I'm now using a ThinkPad with an AMD processor. It's hard to imagine that AMD was once this far behind Intel...
And if these new Ryzen 5000 Series desktop processors match up to AMD's claims, Intel will be in serious trouble.
 
1:32 AM
It'll take more than a year for 10nm SuperFin to get to desktop CPUs, according to current plans
 
Yeah. Q4 2021 is what I've heard.
They're going to backport their new Cove architecture to 14nm, but I'm seriously concerned about power consumption.
And even that may not be enough.
 
Currently Intel still holds an advantage in gaming and single-threaded performance, but the 5000 may make that disappear.
 
Speaking of the K10 APU laptop I mentioned earlier, it originally came with a 1366x768 15-inch screen. Will upgrading it to a 1080p screen put too much strain on the graphics?
 
@gparyani The internal display interface (not sure if it's eDP) might not even support it.
@gparyani My instinct tells me that Rocket Lake might just be able to edge out Zen 3. But it'll be by a tiny margin, and it'll draw far more power in doing so.
 
1:38 AM
@bwDraco It does support it. I've seen forum posts describing a successful upgrade for my exact model. Only thing is, the internal video cable doesn't support higher resolutions, so I'll have to replace that, but it should work.
@bwDraco It's possible that mainstream pre-built desktops may use the lower power Intel T-series CPUs for a while
 
It'll probably work, and I don't think you'll suffer poor performance for everyday use.
AMD is claiming such a colossal improvement in IPC for the 5000 series that backporting Cove and pushing it to 5 GHz (which is what the rumors say) is not likely to be sufficient. Of course, this assumes the claims are true, but...
It's getting late here so I'm going to bed in a bit. TTYL :)
One last thing, though: I just remembered that AMD was claiming absurdly large improvements in CPU-bound eSports game performance. You're looking at 50% faster than Zen 2 in League of Legends at 1080p high, and 21% faster than the Intel Core i9-10900K in the same.
(using GeForce RTX 2080 Ti graphics)
I cannot see how Rocket Lake can close a gap this large.
The next few months are going to be fun.
In any case, good night :)
 

« first day (3730 days earlier)      last day (44 days later) »