With Zen 2 moving to 7/5nm they say it boosts existing IPC by 15%, and with Intel having issues manufacturing their own 7nm chips I'd say AMD has a good chance to surpass Intel next year
People say this every time, I'll believe it when it happens.
Please no. Ryzen's lack of onboard video is a great example of their commitment to the enthusiast niche, and I don't want them wasting the die space. I especially think it's unnecessary because the Ryzen-Vega APUs are so damn good.
If you used the same exact design, other than clock improvement, the node change shouldn't give you IPC improvements.
IPC is an architectural thing, and something AMD should be working on improving.
AMD doesn't follow Intel's tick-tock pattern. They do incremental uarch design improvements every time, they don't insist on keeping the design the same with die shrinks.
And the IPC with Ryzen is neck-and-neck with Intel anyways.
Just doing video output, no actual graphics processing more advanced than a Color Graphics Adapter, is useless in today's world. The Windows UI is full of effects too heavy to do CPU-side without dragging everything down, not to mention GPU compositing in web browsers, video playback, etc. which all need some level of actual GPU capabilities.
There are cases where you need more CPU power than GPU. That's when you buy an APU. If you need more than that, you buy a Threadripper and a GeForce 705. You don't make every single Ryzen cost an extra $5 to slap a bunch of useless transistors in it.
The post I quoted was only talking about the 7nm node. While Zen 2 is going to be on 7nm, that doesn't mean the IPC improvement is because of 7nm (it's because of uarch improvements).
Ryzen should seek to exceed Intel Core in IPC, not match it.
At this point I don't even know what you're trying to say. You definitely don't know as much as you think you know. For instance, Intel also sells server chips with no GPU, because most people don't need graphics on their servers. And the ones that do, have a full-blown iGPU, on par with the Core i5/i7 series - and those are not "just a display block", but a full (if small) GPU that takes up about two to three cores worth of die space. And for servers, most of us would rather have another core or two, rather than an unused iGPU.
Except for very specialized boards, the majority server boards have working VGA out - you sorta need video out when you need to hop on a crash cart to fix shit.
Yes, they're the same die as the mainstream desktop parts (E3).
The issue on the AMD side is that, you don't really have an e3 alternative. I wouldn't mind too much if you could get 8-core PGA-1331 "server" boards and "server" Ryzen chips at roughly Ryzen desktop clocks, but you can't - it's either high cores and low clocks, or go Intel.
So, if you want to build a Ryzen server, you have to include a dedicated GPU, which is a fucking waste of space and PCIe slot. (You could probably go headless on some motherboards, but then you're up shit creek if stuff goes wrong, since you don't even get an IPMI-like interface on consumer boards).
This would be the most worthless thing ever. I can't wait to be forced to run Windows in potato-effects mode and watch YouTube videos with software decoding
https://i.imgur.com/EEnq3SC.jpg
I thought I was having a stroke for a second when I was trying to link a friend the steps for reinstalling windows
I agree with all of you in saying that AMD should make Zen perform as well as they can without too much compromise. If they're just throwing cores around like nobody business, it'll be like the FX series all over again.
https://files.facepunch.com/forum/upload/634/3e6c583b-b1dd-486a-8026-3088be6dae67/image.png
I wouldn't worry too much about that. The only thing they're at all weak on is peak clock speed, and not by much as of Zen+. And judging by how quickly Intel started putting out more cores, it's safe to say AMD hasn't gone overboard with them yet - rather, Intel had been shortchanging everyone severely when it came to core counts, and AMD righted things.
I don't see them actually releasing 16-core parts as normal desktop chips. Some kind of prosumer, Threadripper-type deal, sure, or maybe some marketing-wank lies like smartphones do ("we totally have twelve cores... a DSP for the camera counts as a core, right?"). But I don't see a benefit in making the normal high-end-consumer-desktop part have that many cores, which is why I don't think that rumor is all that likely.
The "4-8 cores for mainstream, gamers and light enthusiasts/workstations" and "above that for HEDT" paradigm isn't going to change any time soon and I don't think it needs to, either.
I think they'll chase 64-core Epyc. Question is how.
With the density increase, and scaling of 7nm, they may just add more IFIS/∞-fabric + IFOP SerDes and try and hit 8 dice per package.
Then don't buy it.
Just going to drop this in here because I thought it was cool that he showed towards the end of the video how his sponsor - a vpn - works.
Also because he's fucking hilarious and the shit he builds is just so fucking weird.
https://www.youtube.com/watch?v=Q8QlNuTUe4M
oh my god it's the kid who made the remote-controlled involuntary dab machine
https://www.youtube.com/watch?v=5QYpD428hAQ
That's the first video I had found of his a while back and it made me instantly subscribe to him. Dude is hilarious and quite smart.
Bought my friends unopened GPD Win 2 for $500. Posting from it now. This thing is niiiiiiiiiiiice!
Is that the guy who built a robot to shine a laser pointer in people's eyes?
Yep, and one that was a modification of it to shoot energy drink at people's faces.
Most of those implement it on the motherboard, not the CPU.
There's a couple oddballs out there that are marketed as "entry-level workstation" CPUs like https://ark.intel.com/products/52275/Intel-Xeon-Processor-E3-1260L-8M-Cache-2_40-GHz but they're very rare.
And people said the same thing about dual cores, then quad cores, and then oct cores too when they first came out.
640K is more memory than anyone will ever need
Why can't I delete this windows.old folder?
https://i.imgur.com/IJo3212.png
The only thing in it is a nest of 7 empty folders, and once I navigate far enough, I get this:
https://i.imgur.com/SQGb5HU.png
I've tried disk cleanup with admin privileges as well as booting up a live linux environment and deleting it that way (it didn't work). This is odd as fuck.
windows dot jpeg
Did you try via command line with sudo in Linux?
I used the linux subsystem to remove the files.
Don't know how it worked. ntfs lol
If AMD wanted to implement a very limited iGPU+Display+MME onto the chipset, I would be totally OK with that. Dealing with the wiring for APUs in the situation will probably be less ideal, and obv chipset cost will go up which isn't great for mainboard partners.
Eventually you get a dichotomy of things that are embarrassingly parallel (raytracing, rasterization, etc) that will just be done on specialized hardware (GPGPU), and tasks that are either somewhat parallel (LZMA2), and then tasks that aren't really parallel at all. The CPU is obviously going to run those last two, so you're eventually going to see decreasing returns per each core you add - AMD is doing the right thing by targeting 8-cores, and implementing great technology like PBO and XFR2. Server CPUs are a different matter generally, since the majority of applications there can be scaled (apache, nginx, etc).
Just in my experience dealing with applications, anything that could eat up all those CPU cores is either GPU accelerated now (Blender), or a legacy application that isn't going to scale past 16 logical processors anyway (VRAD).
I have a script that gives me ownership of files and it sometimes works.
don't you ever disappear on me again, motherfucker
https://files.facepunch.com/forum/upload/230258/13ea0be6-9575-4739-b6b9-bb60bf948f34/IMG_20180722_132828024.jpg
Sorry, you need to Log In to post a reply to this thread.
CIPWTTKT&GC 0x2E - Abort, Retry, Fail?
