|
Case/cable management and fans are p much the best part of pc building
|
#
?
May 29, 2019 22:17
|
|
|
|
| # ? Jun 6, 2024 02:41 |
|
|
Statutory Ape posted:Case/cable management and fans are p much the best part of pc building *looks at her case* Nah.
|
|
#
?
May 29, 2019 22:47
|
|
|
NewFatMike posted:Extremely excited for Threadripper, I will absolutely be getting that enormous loving Noctua cooler if it's hitting 4.8 with PBO. YOU ARE HUGE! THAT MEAN YOU HAVE HUGE PROJECT SIZES! RIP AND TEAR! (I too am super loving excited for threadripper and all the delicious AVX goodness it possesses.)
|
|
#
?
May 29, 2019 23:12
|
|
|
Really want that passive cooler.
|
|
#
?
May 29, 2019 23:18
|
|
|
Currently sitting pretty on a 8700k but all this AMD news is nice. The market needs competition and AMD is bringing it. I can use what I currently have for some time to come but when the time comes I may just go for AMD this time around. I hope nvidia gets a similar kick in the dillz for pretty much the same reason.
|
|
#
?
May 30, 2019 02:09
|
|
|
Mindblast posted:Currently sitting pretty on a 8700k but all this AMD news is nice. The market needs competition and AMD is bringing it. I can use what I currently have for some time to come but when the time comes I may just go for AMD this time around. I hope nvidia gets a similar kick in the dillz for pretty much the same reason. This is my position as well, despite every machine I own currently having an AMD CPU in it. The key word is "currently". Two years ago, my "hot" desktop was an i3-6100T with a GTX 1050 Ti in it. I happened to fall down the rabbit hole of grid computing projects just as the original Zen hype was getting going. I had a fever for more cores, and there was this slim chance that AMD was going to unfuck themselves and give them to me. The rest is history. But this has all happened before, and it will all happen again. If we go back to 2002, my "hot" desktop was a dual-socket Athlon MP -- but every machine I had between then and now was an Intel, because AMD decided to throw themselves in a dumpster after the runaway success of the Opteron line. I'm currently loving what Ryzen is doing for me, but as soon as the cost-benefit analysis is back on Intel's side, that's where I'll be. And my compute machines will have Nvidia GPUs in them until AMD delivers something that performs better while running inside the 75W power envelope. I don't mix religion and corporations. Everyone should use the tools that work best for them, with the problems they are currently trying to solve. But all that said, it is super-fun to watch Intel gently caress up at every opportunity right now.
|
|
#
?
May 30, 2019 04:12
|
|
|
K8.0 posted:I'd assume it's a 12c part based on the cinebench vs 9900k.
|
|
#
?
May 30, 2019 04:27
|
|
|
I think his point is that Cinebench scales well with cores. Intel CPUs still beat AMD ones core for core.
|
|
#
?
May 30, 2019 04:53
|
|
|
shrike82 posted:I think his point is that Cinebench scales well with cores. Intel CPUs still beat AMD ones core for core. If AMD's IPC increase is to believed then Zen2 matches or slightly beats Intel CPU's core for core though too. That is why Zen2 is such a big deal. Its not just more cores. For desktop (and server, very different market though and more cores matter lots there) Intel can still clock higher but the difference is no longer all that big and clocking those Intel chips high enough to make a noticeable difference outside of synth benches presents a real cooling challenge if you're interested in 8C or more parts. Which many if not most people here are.
|
|
#
?
May 30, 2019 05:37
|
|
|
I have a 1950X so you're preaching to the choir
|
|
#
?
May 30, 2019 05:47
|
|
|
K8.0 posted:I'd assume it's a 12c part based on the cinebench vs 9900k. AMD isn't suddenly beating Intel 13% clock for clock in cinebench.. Those clocks are pretty solid looking for a 12c part, especially given that it's a (probably late) ES and the retail products will probably be slightly better. If these slides are to be believed, the 3800X is about on par with the 9900k in Cinebench. 
|
|
#
?
May 30, 2019 06:17
|
|
|
It’s nice that Cinebench made AMD remember AVX but don’t trust numbers that aren’t independently proven.
|
|
#
?
May 30, 2019 08:05
|
|
|
I just want info on x570 itx boards. Hoping for a good one with 2 * m2.
|
|
#
?
May 30, 2019 10:04
|
|
|
GRINDCORE MEGGIDO posted:Hoping for a good one with 2 * m2. There you go. EDIT: On the "DTX" thing, TechPowerUp clarifies: quote:The board sticks out a good 1-inch longer than mini-ITX specifications. The idea behind this being that most mini-ITX gaming cases have room for 2-slot thick graphics cards anyway, so it could make sense to stretch the PCB a little lengthwise, so the precious real-estate gained could be used to cram more devices.
|
|
#
?
May 30, 2019 10:37
|
|
|
That should fit my case! How weird, it has a sodimm2 slot with a caddy for 2* M2. How does that work?
|
|
#
?
May 30, 2019 11:07
|
|
|
Anybody get a look at those white Chromax fans? I know what I'm blowing my money on....
|
|
#
?
May 30, 2019 11:11
|
|
|
Craptacular! posted:On the "DTX" thing Maybe it can grow breadthwise to "cram more devices" since most people actually use full length GPUs, oh wait mATX called and said hi
|
|
#
?
May 30, 2019 11:51
|
|
|
More details leak on Zen2 memory latency. Scroll down to see the relevant part. Latency appears to be 10ns lower than Zen1 on average going by that. (edit) And that is with decent DDR4 2666. So latency numbers with good DDR4 3000 or 3200 will be better. So it looks like rumors of Zen2 actually having lower memory latency than Zen1 or Zen+ despite having a off die memory controller may be true. I guess most of the latency increases from the old North Bridge based memory controllers was from them being off package + a relatively long distance from the CPU core itself? Maybe this means we'll see more off die but on package memory controllers in the future since there doesn't seem to be any real performance degradation by using this approach? edit: there has been a kind've low key but constant susurrus of FUD about Zen2's off die memory controller with claims that it'd cause latency to spike significantly and therefore kneecap its performance in anything that liked low latency (ie. games) and while others have mentioned before (months ago) that the likely hit to latency was going to be very small at worst (2-3ns with DDR4 2666) that got largely ignored. So its good to have some evidence that was BS. \/\/\/\/\/\/\/\/\/\/ PC LOAD LETTER fucked around with this message at 18:55 on May 30, 2019 |
|
#
?
May 30, 2019 17:47
|
|
|
PC LOAD LETTER posted:More details leak on Zen2 memory latency. Scroll down to see the relevant part. The amount of additional trace run to hit a memory controller inside the socket is minuscule compared to the distance to get to the memory itself, this isn't really surprising so long as they managed to keep the frequency of the core to io die high enough. It's extremely good for home users as applications no longer need to be numa-optimized to avoid pulling from a non-local memory channel and suffering the penalty of traversing the core to core interconnect. Also good for most datacenter customers on Epyc, though some people doing high-frequency trading and whatnot would probably find this sub-ideal as they were already already optimizing around numa to ensure they were pulling from local memory channels. The lowest latency measurements are likely to be a little higher than Zen1, with the average being considerably lower.
|
|
#
?
May 30, 2019 18:04
|
|
|
GRINDCORE MEGGIDO posted:That should fit my case! How weird, it has a sodimm2 slot with a caddy for 2* M2. How does that work? Your other option is this Gigabyte board which is true ITX and does the more traditional(?) thing of putting the second M2 on the back of the motherboard. IMO the Asus mesh looks better than what Gigabyte has on that board for cooling the chipset.
|
|
#
?
May 30, 2019 19:28
|
|
|
PC LOAD LETTER posted:edit: there has been a kind've low key but constant susurrus of FUD about Zen2's off die memory controller with claims that it'd cause latency to spike significantly and therefore kneecap its performance in anything that liked low latency (ie. games) and while others have mentioned before (months ago) that the likely hit to latency was going to be very small at worst (2-3ns with DDR4 2666) that got largely ignored. So its good to have some evidence that was BS. \/\/\/\/\/\/\/\/\/\/ It's good to know it doesn't do the bad thing when compared to zen1. Khorne fucked around with this message at 19:59 on May 30, 2019 |
|
#
?
May 30, 2019 19:56
|
|
|
pokin around that website and looking at a bunch of 2700X + DDR4 2666 results, it seems to me that the pre-production sample Zen 2 might be slightly behind the better 2700X results. But very slightly, probably to the point of being insignificant and within the range of timing tweaks and such, so I would chalk that up as a win for AMD (even if they are still 10ns or so behind Intel). My primary takeaway from that site was that greater than 50% of the builds that had a 2666 kit of memory were running it at 2133. Reminder to check your memory settings, it gave me a sad.
|
|
#
?
May 30, 2019 21:01
|
|
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
|
#
?
May 30, 2019 21:33
|
|
|
No because PCI 4 eats too much power.
|
|
#
?
May 30, 2019 21:34
|
|
|
Active cooling (fan or liquid) is an AMD requirement, apparently.
|
|
#
?
May 30, 2019 21:35
|
|
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
|
#
?
May 30, 2019 21:38
|
|
|
sincx posted:The Aorus Xtreme is fanless. Huh! You are right. I hadn't seen that one specifically with the millions of boards coming out. Odd that literally every x570 board except that one has em? Maybe AMD gave em an out because they put those functional heatsinks on it? Dunno. Although with the size of that cladding, it would be pretty hilarious if there is a tiny fan buried in there (GN says there isnt a fan but still)
|
|
#
?
May 30, 2019 22:01
|
|
|
PC LOAD LETTER posted:Latency appears to be 10ns lower than Zen1 on average going by that. (edit) And that is with decent DDR4 2666. So latency numbers with good DDR4 3000 or 3200 will be better.
|
|
#
?
May 30, 2019 22:58
|
|
|
What was the most recent previous chipset that used ~20w?
|
|
#
?
May 30, 2019 23:11
|
|
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
|
#
?
May 30, 2019 23:17
|
|
|
Does the X570 chipset run on 7nm as well? I can't seem to find that detail about the chipset.
|
|
#
?
May 30, 2019 23:20
|
|
|
spasticColon posted:Does the X570 chipset run on 7nm as well? I can't seem to find that detail about the chipset. Doubt it.
|
|
#
?
May 30, 2019 23:29
|
|
|
IIRC GF is making the chipsets which is at least part of how AMD intends to fulfill the amended WSA.
|
|
#
?
May 30, 2019 23:35
|
|
|
redeyes posted:No because PCI 4 eats too much power. Just a question: I know tech gets cheaper over time, but is this cooling challenge something that changes with time and manufacturing improvements, or are we all looking at motherboard fans in our future?
|
|
#
?
May 31, 2019 00:00
|
|
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
|
#
?
May 31, 2019 00:02
|
|
|
GRINDCORE MEGGIDO posted:That should fit my case! How weird, it has a sodimm2 slot with a caddy for 2* M2. How does that work? the sodimm2 slot is wired up as a custom connector for NVMe and Sata, the caddy plugs into it and makes the wires go to the right place? sodimm sounds like a good solution for that, it lies flat so you can still fit the thing under a 2-slot video cooler. if you plug laptop memory into the slot the ram almost certainly goes up in smoke and possibly the mobo as well. spasticColon posted:Does the X570 chipset run on 7nm as well? I can't seem to find that detail about the chipset. chipset is being made by GloFo so no
|
|
#
?
May 31, 2019 00:04
|
|
|
What's the earliest that we'll see widespread DDR5? 2024? Am I completely off-base on that? Cuz it took like, three, four years for DDR4 to hit critical mass, I figure it'll take at least five, considering diminishing returns.
|
|
#
?
May 31, 2019 00:27
|
|
|
SwissArmyDruid posted:What's the earliest that we'll see widespread DDR5? 2024? Am I completely off-base on that? Cuz it took like, three, four years for DDR4 to hit critical mass, I figure it'll take at least five, considering diminishing returns.
|
|
#
?
May 31, 2019 00:36
|
|
|
Klyith posted:the sodimm2 slot is wired up as a custom connector for NVMe and Sata, the caddy plugs into it and makes the wires go to the right place? sodimm sounds like a good solution for that, it lies flat so you can still fit the thing under a 2-slot video cooler. Understood, but I'm wondering why they went to the trouble of a caddy / convertor. It can't save much any board space?
|
|
#
?
May 31, 2019 00:50
|
|
|
|
| # ? Jun 6, 2024 02:41 |
|
|
Current M.2 mounting means you're covering something up. So if there's something that needs to radiate its own heat ambiently, or has pins, it can't be put there. A caddy + slot frees up space, and moves the M.2 drive out to somewhere it can also get air for cooling, which the SSD controller will gladly take wherever possible.
|
|
#
?
May 31, 2019 00:55
|
|