|
FaustianQ posted:Yeah, it's absolute garbage and it's just mindblowing they think it's acceptable. It's also in a 35W configuration, so it's basically running below UHD 630 performance. It's super dumb and I can't figure out how they think it'll sell at that price. To be special it'd need to somehow fully use the IMC to feed the 512 GCN3 ALUs, it'd have to be capable of running both DDR4 and GDDR5 simultaneously, and it'd need an adequate cooling solution to run at full speed. Instead it's just an obnoxiously expensive, underpowered toy. I wanted to chime in on this as say that I was a backer of their previous x86/arduino hybrid board, and they missed their ship date by 9 or 9 months, they didn't have some of the advertised features working like HDMI CEC (still not working completely correct), and their customer service is really really bad. My unit shipped without the screws for acrylic case that was included and it took them 7 months to even ship them (despite being repeatedly told that they had already shipped them or would that day). I know many other users had to wait months to get the antennas for their wifi modules. Basically, I wouldn't trusty these clowns to deliver what they promise or on time.
|
# ? Jun 1, 2018 15:38 |
|
|
# ? May 17, 2024 14:23 |
|
I wonder what kind of heatsink it takes to passively cool 25W (alu vs cu).
|
# ? Jun 1, 2018 17:55 |
|
Not much at all if there's some case airflow.
|
# ? Jun 1, 2018 18:00 |
|
Seamonster posted:I wonder what kind of heatsink it takes to passively cool 25W (alu vs cu). https://www.arctic.ac/worldwide_en/alpine-m1-passive.html This cools up to 35w apparently, I have one and it's enormous.
|
# ? Jun 1, 2018 18:21 |
|
So while we computer enthusiasts find the Atari VCS dumb and stupid and why woul- It sold 2 million in preorders on the first day and crashed Indiegogo. I'm loving baffled but *grumbles* Fishmech was right(?) What is interesting about this is they could still "fix" the Atari VCS since Bristol and Raven are socket compatible. A 2400GE and 4GB more of DDR4 @ 2933mhz would at least match if not sometimes beat the Xbox One S, while fitting into the same thermal envelope as the A12-9800E they intend to use. They still haven't finalized the design, that's still for Q3 2018 and 2 million in preorders means they actually have a loving market, it's not the Ouya again.
|
# ? Jun 1, 2018 19:10 |
|
FaustianQ posted:So while we computer enthusiasts find the Atari VCS dumb and stupid and why woul- It sold 2 million in preorders on the first day and crashed Indiegogo. I'm loving baffled but *grumbles* Fishmech was right(?) quote:“Hurry, while supplies last!” says the pre-order campaign
|
# ? Jun 1, 2018 19:26 |
|
I'd normally be all over a nostalgic Atari console -- and something about it is still kind of appealing to my inner child -- but I'm also old enough to know that crowdfunding something like this is a HUGE red flag. It's a method of raising money more suitable to DIY / vanity projects, not for a rollout of an ecosystem based on classic IP that will require ongoing support.
|
# ? Jun 1, 2018 20:09 |
|
Dadbod Apocalypse posted:I'd normally be all over a nostalgic Atari console -- and something about it is still kind of appealing to my inner child -- but I'm also old enough to know that crowdfunding something like this is a HUGE red flag. It's a method of raising money more suitable to DIY / vanity projects, not for a rollout of an ecosystem based on classic IP that will require ongoing support. Atari games at this point? I am too old to be entertained by Joust or Pong.
|
# ? Jun 1, 2018 20:16 |
|
Sounds a bit like an Ouya for nostalgics - instead of Android games.
|
# ? Jun 1, 2018 21:01 |
|
You'd think Atari would do what their history, Sony, and MS did... and use Jaguar.
|
# ? Jun 1, 2018 21:16 |
|
Crosspost from the PC building thread. Please keep in mind that I'm in Canada before you start giving me links to US retailers. Has anyone had any luck using RAM that's not on the official QVL to run at 3200mhz with a Ryzen 2000+ CPU? Specifically, I'm about to pull the trigger on this motherboard: https://ca.pcpartpicker.com/product/rpZ2FT/asus-tuf-b350m-plus-gaming-micro-atx-am4-motherboard-tuf-b350m-plus-gaming The QVL for 3200mhz memory is here: http://dlcdnet.asus.com/pub/ASUS/mb...7108.1527885469 I'm looking at the following RAM kits: https://ca.pcpartpicker.com/product/MYs8TW/team-night-hawk-rgb-16gb-2-x-8gb-ddr4-3200-memory-tf1d416g3200hc16cdc01 https://ca.pcpartpicker.com/product/ybrcCJ/gskill-tridentz-rgb-16gb-2-x-8gb-ddr4-3200-memory-f4-3200c16d-16gtzr But none of these are on the list. The Trident's 3000mhz version is, but not the 3200mhz. I know some people have luck getting RAM that's not on the QVL to run at 3200mhz but it's a gamble so I'm asking if anyone has been successful with these (or other) kits in a similar setup. I'm open to any recommendations or suggestions for motherboard/RAM picks if they're confirmed to work at 3200mhz on Ryzen as well.
|
# ? Jun 1, 2018 22:21 |
|
Running 3200 on a B350 motherboard is tough as it is. Either buy X470 or wait for B450. Old first gen Ryzen motherboards are over.
|
# ? Jun 1, 2018 22:23 |
|
Agreed, it's better to wait as B450 will likely debut along with the next wave of Pinnacle SKUs on June 5th.
|
# ? Jun 1, 2018 22:35 |
|
Yeah I don't mind waiting a little bit. I keep seeing that B450 will be available Q3 2018. Hopefully I don't have to wait till the end of September... Do you think boards will be available shortly after the June 5 date?
Streak fucked around with this message at 23:14 on Jun 1, 2018 |
# ? Jun 1, 2018 22:51 |
|
Mister Facetious posted:You'd think Atari would do what their history, Sony, and MS did... and use Jaguar. loving boooooooooooooooooooooooo.
|
# ? Jun 1, 2018 22:54 |
|
Looks like AMD's computex keynote will unveil Fenghuang. TUM APISAK has one (he works at an OEM partner and has leaked reliable benchmarks several times previously) Paul MaudDib fucked around with this message at 23:37 on Jun 2, 2018 |
# ? Jun 2, 2018 07:13 |
|
Craptacular! posted:Running 3200 on a B350 motherboard is tough as it is. Either buy X470 or wait for B450. Old first gen Ryzen motherboards are over. I have a low-mid tier MSI motherboard that runs my B-die at 3333 with tight timings. Confirmed stable with 24 hours of Prime and a couple full passes of MemTest. I’m merely saying that it’s not a pipe dream to overclock some on crappier boards.
|
# ? Jun 2, 2018 15:02 |
|
Paul MaudDib posted:Looks like AMD's computex keynote will unveil Fenghuang. TUM APISAK has one. Looks like 28CUs vs the 24 in Kaby Lake G. Also a pretty big memory frequency increase, although half the HBM.
|
# ? Jun 2, 2018 18:11 |
|
Paul MaudDib posted:Looks like AMD's computex keynote will unveil Fenghuang. TUM APISAK has one (he's an AIB partner who has leaked reliable benchmarks several times previously) Ya boy is thirsty for this one.
|
# ? Jun 2, 2018 18:34 |
|
Cygni posted:Also a pretty big memory frequency increase, although half the HBM. But everyone knows 3GB is too small for 1080p gaming in 2018 Guess we get to find out if HBCC can really "increase the size of the VRAM" in titles that aren't mismanaging their memory. It'll be fine either way though, 2GB is still alright especially for the esports that most people really play on these. Some AAA titles will probably need to run at medium textures, oh well. Paul MaudDib fucked around with this message at 23:51 on Jun 2, 2018 |
# ? Jun 2, 2018 23:44 |
|
Paul MaudDib posted:But everyone knows 3GB is too small for 1080p gaming in 2018 The big difference is that the fallback mode for a dGPU is the PCI-E bus instead of (hopefully) dual channel DDR4. This could be a big step towards making mid-range APUs price competitive, but it remains to be seen if AMD can stick the landing.
|
# ? Jun 3, 2018 19:19 |
|
Paul MaudDib posted:But everyone knows 3GB is too small for 1080p gaming in 2018 Yeah, this is my key interest point with this, the first real-world use case for HBCC. Well that, and the price - there's plenty of room beneath Kaby-G.
|
# ? Jun 3, 2018 19:40 |
|
I thought we already learned that HBCC didn't do much if anything?
|
# ? Jun 3, 2018 20:58 |
|
In the lead-up to the Ryzen 2X00 release, I had planned to replace my 1700 with a 2700, and put the 1700 in a partial build that I had put on hold. But some life stuff came up, and I had to put that cash elsewhere for a couple months. Yesterday I decided to finish out the build, but to just keep my desktop as-is and slot a 1X00 into the second machine. My local microcenter had restocked, and had 10+ R5 1600s at $150, so I snagged one of those. Now I have 28 threads, in two machines, crunching WorldComputeGrid workunits 24/7. Two years ago that would have cost me, what, $2500 in Intel CPUs? Total CPU outlay here: $450. Since neither of these machines is a gaming rig, I'll probably keep their mobos and memory and swap out CPUs when the Zen2s arrive. Finally, it looks like the continued BIOS and/or kernel upgrades have improved performance under Linux. When I first rebuilt my desktop with the 1700, all cores were running at their base clock of 3.0GHz. Yesterday after initial install, I checked the speeds of the 1600s cores, and the kernel was reporting 3.4GHz. I thought I remembered (and Wikipedia confirmed) that the base clock of a 1600 was 3.2GHz -- but the all-cores boost clock is 3.4GHz, and the CPU is continuing to run at that speed even after about 18 hours of 100% utilization. The reason I think this is due to BIOS/kernel improvements is that my 1700 is now running all threads at 3.2GHz, which it definitely was not doing before. I'll take it. Edit: interestingly, the 1600 is running all cores at 3393MHz, with just fractional clock differences. The 1700 has a bit more of a spread: 6 cores at 3193MHz, 1 core at 3187MHz, and 1 core at 3184MHz. mdxi fucked around with this message at 21:41 on Jun 3, 2018 |
# ? Jun 3, 2018 21:38 |
|
Paul MaudDib posted:But everyone knows 3GB is too small for 1080p gaming in 2018 Pubg definitely needs 4gb for High textures, I've run it through a few sampling runs and you'll start seeing the vram commit creeping up to the 3gb mark a few min in before it goes over, starts dumping in to ram and becomes a stuttery mess. That game has dogshit memory management though.
|
# ? Jun 4, 2018 14:52 |
|
mdxi posted:In the lead-up to the Ryzen 2X00 release, I had planned to replace my 1700 with a 2700, and put the 1700 in a partial build that I had put on hold. But some life stuff came up, and I had to put that cash elsewhere for a couple months. Yesterday I decided to finish out the build, but to just keep my desktop as-is and slot a 1X00 into the second machine. My local microcenter had restocked, and had 10+ R5 1600s at $150, so I snagged one of those. Well? Show us some sweet graphs of these monsters in action mdxi posted:Finally, it looks like the continued BIOS and/or kernel upgrades have improved performance under Linux. When I first rebuilt my desktop with the 1700, all cores were running at their base clock of 3.0GHz. Yesterday after initial install, I checked the speeds of the 1600s cores, and the kernel was reporting 3.4GHz. I thought I remembered (and Wikipedia confirmed) that the base clock of a 1600 was 3.2GHz -- but the all-cores boost clock is 3.4GHz, and the CPU is continuing to run at that speed even after about 18 hours of 100% utilization. The reason I think this is due to BIOS/kernel improvements is that my 1700 is now running all threads at 3.2GHz, which it definitely was not doing before. I'll take it. Have you tried overclocking? You're leaving a significant amount of performance on the table if you don't. My 1700 gets to 4.0GHz on 1.275V with 2x16GB ECC B-die at DDR4-2933 CL16. That's about ~30% more clock speed, to say nothing of the effects of high speed memory. The ECC RAM gets noisy throwing errors when I try to drive the DIMMs or the memory controller too hard, so it's easy to figure out how far you can push it. A cool dude wrote a script (https://github.com/r4m0n/ZenStates-Linux) that lets us overclock at runtime on Linux like Windows users can with the Ryzen Master app. I find it both easier and more reliable than trying to work around some of the overclocking-related UEFI bugs on my system. It won't do anything with RAM speed or timings though, so get those settings dialed in with the UEFI. code:
code:
SamDabbers fucked around with this message at 13:14 on Jun 5, 2018 |
# ? Jun 5, 2018 13:06 |
|
Gee, there's (probably unfounded) rumors about more than 16 cores for the upcoming Threadripper. https://www.pcgamesn.com/amd-threadripper-2-moar-cores While that'd be hilarious, it'd make latency worse, since it'd imply activating the other two dies, halving direct die to die bandwidth. --edit: Holy crap, that upcoming X399 Creation board from MSI is loving gaudy. Combat Pretzel fucked around with this message at 16:26 on Jun 5, 2018 |
# ? Jun 5, 2018 16:12 |
|
Combat Pretzel posted:Gee, there's (probably unfounded) rumors about more than 16 cores for the upcoming Threadripper. AMD should just rebrand an 32 Core Epyc part to ThreadRipper ( which it probably is ) and call it a day.
|
# ? Jun 5, 2018 17:54 |
|
If I recall correctly Epyc already uses sTR4, so they probably could straightforwardly hack together a 32-core "Threadripper" which would just be a full four-die Epyc instead of the 2-die+2-shim thing they're doing now. My main concern with that would be increased NUMA issues - we already have two channels connected to one die and two connected to the other, are we going to come up with a new memory controller design that connects one channel to each die or are we going to just tack on two new dies and force them to go through the first two for all memory accesses? I probably don't need to elaborate that each of those two approaches has some nasty performance implications. The alternative would be to use a >8 core die but that doesn't exist yet and I don't know that it's actually a good idea to go past 8 cores on the mainstream models yet. You could also go from four memory channels to eight but then you're just straight up selling Epyc as a consumer product and you'll need all new motherboards, I assume. e: 1 die != 1 CCX Eletriarnation fucked around with this message at 18:18 on Jun 5, 2018 |
# ? Jun 5, 2018 18:11 |
|
There was a video a while back where someone tried to get an Epyc to work in a TR4 socket. It is physically compatible but there is differences in the memory controller that prevent it from booting. So it is very possible for AMD to roll out a 32 core TR.
|
# ? Jun 5, 2018 18:16 |
|
SamDabbers posted:Well? Show us some sweet graphs of these monsters in action There's not really a lot to show, but here's some relevant WCG screenshots: SamDabbers posted:Have you tried overclocking? No. I'm running them at 100% utilization 24/7, and they're hot enough and drawing enough power as it is. Maybe this winter!
|
# ? Jun 5, 2018 18:31 |
|
So apparently Z490 was preemptively aborted due to shortage of PCI Plex chips if you were like me and holding off on X470.
|
# ? Jun 5, 2018 18:45 |
|
Xae posted:There was a video a while back where someone tried to get an Epyc to work in a TR4 socket. I don't think this will happen until 7nm but hey more power to them. Once 7nm TR does hit though and you've got 64 threads @ ~5 ghz holy shiiiiiiiiiitt
|
# ? Jun 5, 2018 18:51 |
|
The problem with a 32-core Threadripper on the current set of boards, or probably any upcoming board (due to routing complexity), is that only two dies will have access to memory. The others would be accessing memory by proxy. The current Threadrippers have double bandwidth between the two dies, that one will also go missing, the second channel would make a stop over one of the two additional dies. For creative stuff, a 32-core one on Zen+ would probably be awesome. But that's about it. Also, +33% TDP doesn't make up for double the cores.
|
# ? Jun 5, 2018 18:56 |
|
Combat Pretzel posted:The problem with a 32-core Threadripper
|
# ? Jun 5, 2018 21:39 |
|
Combat Pretzel posted:The problem with a 32-core Threadripper on the current set of boards, or probably any upcoming board (due to routing complexity), is that only two dies will have access to memory. The others would be accessing memory by proxy. The current Threadrippers have double bandwidth between the two dies, that one will also go missing, the second channel would make a stop over one of the two additional dies. For creative stuff, a 32-core one on Zen+ would probably be awesome. But that's about it. Also, +33% TDP doesn't make up for double the cores. I'd imagine most of the problems that benefit from more than 16 full x86-64 cores and can't be trivially done better on a GPU are not going to be memory bound anyway. Seeing as threadripper is essentially EPYC with two dead dies, it is definitely possible for 32 core threadripper to be a thing.
|
# ? Jun 5, 2018 21:50 |
|
Doubt it's 24C/32C, probably just extra TDP headroom for higher turbos.
|
# ? Jun 5, 2018 21:51 |
|
Eyes Only posted:I'd imagine most of the problems that benefit from more than 16 full x86-64 cores and can't be trivially done better on a GPU are not going to be memory bound anyway.
|
# ? Jun 5, 2018 21:55 |
|
Pablo Bluth posted:Computational Fluid Dynamics. You always want more cores than you've got but it's also massively memory bandwidth bound. CFD is still strongly single thread limited. Not nearly as badly as most things, but it's often better to have 10 cores at 5ghz vs. 20 cores at 3ghz. Some code sets scale very nicely, where each extra proc gives you 0.97x the previous performance, some code is poo poo and gives you only 0.71x. In some specific cases, it's better to run the code on some hilariously OC'd HEDT proc vs. putting it on the shiny new 90 core quad socket server.
|
# ? Jun 5, 2018 22:27 |
|
|
# ? May 17, 2024 14:23 |
|
The CFD software I use includes a lot of combustion and thermal radiation calculations. At the start of the simulation it is usually heavily memory bound until there is a lot of movement in the model then it tends to be heavily cpu bound. At least I think that's correct as recently I found out the cpu load reporting includes treating a cpu as busy with cache miss and the cpu is halted waiting for a memory read.
|
# ? Jun 5, 2018 22:59 |