Link? I don't see any 64G UDIMMs on NewEgg. This is the only one that says UDIMM in the title, but then in the description it say LRDIMM, which I'm apt to believe over the title.

Certainly larger DIMMs are broadly available, but only in RDIMM and LRDIMM formats, and those don't work with Threadripper.

https://hardware.slashdot.org/comments.pl?sid=15704014&cid=59687840

I suspect a lot of it is a result of the chronic sickness in technology of chasing taillights. It's especially common in frontend JS right now, just look at the bazillion frameworks that are the new hotness for about 3 minutes each before everyone collectively forgets about them and moves on to the next new hotness. A lot of frontend development is done by younger developers and a lot of it is "throwaway" work, in that nobody is ever going to maintain it and they'll simply start from scratch in a year or two when they decide they want a new design. It kinda sorta "works" there, sometimes, and that just reinforces the cycle.

More along the lines of cloud though... it makes perfect sense for the cloud vendors to be pushing it hard. What could be better than recurring revenue from your customers who volunteered to be held hostage within your proprietary services? I still can't figure out why so many people that aren't cloud vendors are moving in that direction though.

I've seen lots of 60 fps video. It is enormously better than 24 fps video.

The "too real" arguments have been applied to every resolution bump since the beginning of time and yet we're still bumping resolution and watching movies. Maybe you would prefer to go back to VHS?

So basically every single film maker is actively choosing to shoot in 24 fps after carefully considering other framerates? Yeah, right. Obviously it can't be the immense (unjustified) inertia behind 24 fps that's driving it as the defacto standard where any deviation from that is cause for a media frenzy.

...and you know what the best fix is? Stop filming movies at 24 fps and we can finally get beyond this idiocy. I propose 240 fps as the new standard for everything.

Taskbar? That's what auto-hide is for. I might use the taskbar once a month. Every window (or at least enough of it to easily bring it to the top) I'm using at any particular time is visible all at the same time. Note that if you're one of the many many many users afflicted by the "every window must be maximized all the time" disease, you'll have to cure that first.

Basically, this. Npm and friends are all raging dumpster fires when it comes to security. Install one package from npm and you're likely pulling in hundreds or thousands of other packages as deps, and due to the fact that there's no review of any of them, you may as well be going to 1000 random websites and running random executables from them. Most people wouldn't do the latter (I hope...) because you'd have to be outrageously stupid to do so, but the former is now the industry norm and nobody seems to care.

The Linux distro model of distributing software is infinitely better, but has failed to catch on outside of that realm. The software development industry is generally extremely myopic and this is no exception.

Actually, no, AMD invented x86_64, which is the vendor neutral name for AMD64. Intel originally came up with ia64 (the aforementioned Itanic) and had to adopt x86_64 years after AMD came out with it, dumping ia64 into the wastebin in the process.

Perhaps you're thinking about 32-bit x86?

Good grief, no SSE2 support in 2019? The newest AMD CPU to not support SSE2 was the Athlon XP, released in 2003! You could probably dumpster dive something 10 years newer than that for free. The power efficiency gains alone would make it a highly worthwhile upgrade. Surely you jest.

Current NVMe SSDs are already posting sequential performance numbers of 3.5G/s on their spec sheets. The maximum theoretical speed with 4x lanes of PCIe 3.0 is 3.94G/s. Considering overhead that's either already bottlenecked by the bus or very close to it (recall that SATA 3 runs at 600M/s, but practically speaking you can only get ~540M/s of data out of it).

If that were true then you would know that network cards are specced with PCIe lanes such that they can max out their interfaces. 100G ethernet needs 16x PCIe 3 right now, but only 8x PCIe 4. 10G ethernet can be had in 1x PCIe 4 instead of 2x PCIe 3 or 4x PCIe 2, and so on.Only a foolish vendor would do otherwise. So of course you haven't seen PCIe lane bandwidth issues from them, they are designed to not have them. That doesn't mean PCIe 4 is useless, far from it.

Since you're talking about machines with hundreds of PCIe lanes, in the x86 space that means 8+ socket Intel, or lots and lots of PCIe switches (all of which are expensive now, and will be a compromise on bandwidth if you're actually saturating your network interfaces), but perhaps you're not running x86 at all. Regardless, with PCIe 4, you could run double the NICs in the same system with PCIe 4 instead of PCIe 3, or you could use a system half the size (and significantly less than half the cost - 8+ socket systems are super rare and correspondingly expensive). In fact, you could probably run the same workload on a 1P Rome EPYC system, with 128 lanes of PCIe 4 for less than 1/10 the cost.

But you know, totally uninteresting. Oh, and feel free to send me 90% of your paycheck since that would also apparently be not noteworthy to you.

True only up to 2nd gen Ryzen. 3rd gen Ryzen is beating Intel here, at least based on what we've seen so far. Third party benchmarks are coming in under a month, when the CPUs launch.

Hilariously untrue. Have you forgotten how long Intel was pushing quad cores (and later, 6 cores) as their top consumer level CPU vs 8 core Ryzens? Intel was getting absolutely destroyed on multithreaded performance until they launched the 9900k late last year.

Also, if you want to talk about "ever", let's talk about the Athlon 64 (and Athlon X2) era, where Intel had nothing competitive at all for many many years.

Maybe to you. I'll take doubled throughput NVMe SSDs and doubled throughput Infinity Fabric every day of the week. Doubled throughput for PCIe also means you can effectively run twice as many expansion cards (once they catch up) without loss of speed. Four GPUs at PCIe 4.0 with four lanes each is unlikely to bottleneck on the bus if that's your thing, or you can throw in a 100G ethernet card using only eight lanes!

Only with Threadripper (60) vs Intel HEDT (44) and EPYC (128) vs Xeon (48 per CPU) does AMD offer more PCIe lanes than Intel. Both regular consumer level platforms have essentially 20 (16 + 4 connected to the chipset) PCIe lanes available. Of course, at PCIe 4 speeds, AMD will have double the bandwidth despite having the same number of lanes.

Not true unless you're cherry picking benchmarks that heavily favour Intel. Of course, if you want to play that game, I can cherry pick benchmarks where the 2700x soundly beats the 9900k. But back in the real world, the performance delta is only in the 10-15% range, and gets much much closer after fully patching for all of the currently known Meltdown/Spectre/MDS/L1TF/etc vulnerabilities.

That aside, when you add in the clock speed increases to the 15% IPC uplift, we're looking at more like a 25% increase in single threaded performance for the 3800x vs the 2700x, and that is huge.

Based on AMD's numbers, the 3800x will beat the 2700x by ~25% on single threaded performance. We'll have third party benchmarks to confirm when it launches July 7th.