|
Balliver Shagnasty posted:What brand RAM did you buy? I heard how finicky Ryzen was when it came to RAM sticks so I bought 16GB of AMD-optimized DDR4-3200 G.Skill Trident Z RGB. Had absolutely no trouble getting those two sticks to 3200Mhz with A-XMP. I went with: RAM: Corsair Vengeance LPX 16GB (2x8GB) DDR4 DRAM 3000MHz C15 Desktop Memory Kit - Black (CMK16GX4M2B3000C15) ($119.99 @ Amazon) I'm booted to it at 3000MHz right now and almost the advertised timings, but it's reporting at CAS 15 instead of the 16 that's set in BIOS for whatever reason (screens and junk here https://forums.somethingawful.com/showthread.php?threadid=3774409&userid=0&perpage=40&pagenumber=630#post491736471). It's sounding like I should probably return it and get some 3200 for :tenbux: more instead though. This is on my board's QVL and sitting in my cart, 12 hours to pull the trigger and have it show up by Saturday: Patriot Viper 4 Series Extreme Performance DDR4 16GB (2 X 8GB) 3200MHz Kit (PC4-25600) PV416G320C6K
|
# ? Jan 17, 2019 18:00 |
|
|
# ? May 17, 2024 00:03 |
|
GutBomb posted:I get better performance and thermals doing the XFR/PBO stuff in games than I did with manual overclocking with my 2700x. The performance in synthetic benchmarks is actually worse but in game benchmarks and just general game performance it was better. I might go back to XFR/PBO at some point. I just wanted to see how far I could push a manual overclock. For context, I had my old i5-3570k clocked at 4.2Ghz with a 1.16v underclock. I mentioned in the OC thread about how I tried BCLK overclocking, only to end up with boot looping. Haven't tried it again for fear of what it might do to my NVMe drive.
|
# ? Jan 17, 2019 18:04 |
|
teh_Broseph posted:I went with: What you have said here is in direct contradiction to your other post? A lower CAS is better. Around or past the DDR4-3000 mark, slightly slower RAM with a lower CAS may prove equal or better to faster ram with a higher CAS. SwissArmyDruid fucked around with this message at 19:47 on Jan 17, 2019 |
# ? Jan 17, 2019 19:43 |
|
SwissArmyDruid posted:What you have said here is in direct contradiction to your other post? A lower CAS is better. Around or past the DDR4-3000 mark, slightly slower RAM with a lower CAS may prove equal or better to faster ram with a higher CAS. Groovy, thanks! Knowing that I think I'll just stick with what I have and not eff with the whole return/swap process. Now to just try and figure out why setting tCL to 15 in the BIOS is appearing to show up as 16 in CPU-Z/HwInfo64. (I'm re-reading the posts and not seeing what contradiction I said?)
|
# ? Jan 17, 2019 23:12 |
|
teh_Broseph posted:Groovy, thanks! Knowing that I think I'll just stick with what I have and not eff with the whole return/swap process. Now to just try and figure out why setting tCL to 15 in the BIOS is appearing to show up as 16 in CPU-Z/HwInfo64. (I'm re-reading the posts and not seeing what contradiction I said?) From what I’ve read Ryzen won’t run with odd cl timings unless you disable a certain memory setting, might be “geardown” but I’m not 100% sure.
|
# ? Jan 17, 2019 23:28 |
|
teh_Broseph posted:Groovy, thanks! Knowing that I think I'll just stick with what I have and not eff with the whole return/swap process. Now to just try and figure out why setting tCL to 15 in the BIOS is appearing to show up as 16 in CPU-Z/HwInfo64. (I'm re-reading the posts and not seeing what contradiction I said?) teh_Broseph posted:but it's reporting at CAS 15 instead of the 16 that's set in BIOS teh_Broseph posted:I set the CL to 15 as Corsair says it'll do, but CPU-Z is still showing it at 16..? Those are two diametrically-opposed scenarios.
|
# ? Jan 18, 2019 00:11 |
|
SwissArmyDruid posted:Those are two diametrically-opposed scenarios. I'm setting CL15 in the BIOS, but once it boots it shows as CL16 in utilities in Windows. Apparently cause (with some googling to back it up B-Mac posted:From what I’ve read Ryzen won’t run with odd cl timings unless you disable a certain memory setting, might be “geardown” but I’m not 100% sure. (so BIOS allows me to put 15 in the field but it just goes to 16 anyway) which puts me back at maybe I'll cross my fingers and see if I can either hit CL14 or 3200Mhz, and/or just return it for some 3200 cause 10bux to get CL16 3200 over CL16 3000 when I'll probably be using it at least a few years sounds like the way to go. Little more on my 2500k->2600x trip report, Overwatch deffo feels smoother even though my 2500k was exceeding my 96hz refresh. Few minutes of Asscreed was smoother in a town/hub. It's not screaming fast, but Frame Times Matter and so far I'm happy with the purchase for smoothing out the hiccupps I was running in to more often with the 2500k, looking forward to trying more games. teh_Broseph fucked around with this message at 01:23 on Jan 18, 2019 |
# ? Jan 18, 2019 01:14 |
|
B-Mac posted:From what I’ve read Ryzen won’t run with odd cl timings unless you disable a certain memory setting, might be “geardown” but I’m not 100% sure. It will run, it just rounds the number up to the closest even number. And my motherboard (gigabyte aorus x-470 gaming 7 WiFi) refuses to boot with geardown turned off on the most recent bios so I just deal with 16 cas instead.
|
# ? Jan 18, 2019 04:31 |
|
My RAM conundrum seems solved, my CL15 3000Mhz rated memory booted up and ran Asscreed for an hour or so no prob at CL16 3200Mhz (wanna say it was 16-18-18-18-36).
|
# ? Jan 18, 2019 05:46 |
|
CPU-Z is reporting the tRC of my 3200MHz RAM is 76 clocks, but when I tab over to SPD it says the tRC is rated at 54 when the RAM is running at XMP-3200, which it is. Is that particular timing value significant enough that I should manually change it in the BIOS?
|
# ? Jan 18, 2019 07:31 |
|
Balliver Shagnasty posted:What brand RAM did you buy? I heard how finicky Ryzen was when it came to RAM sticks so I bought 16GB of AMD-optimized DDR4-3200 G.Skill Trident Z RGB. Had absolutely no trouble getting those two sticks to 3200Mhz with A-XMP. I got flare x, cas 14. It just ran it stock everything with xmp.
|
# ? Jan 18, 2019 07:32 |
|
As a reminder most motherboards can and will change timings on boot up if something fails. Even if you have all the settings dialed in.
|
# ? Jan 18, 2019 16:16 |
|
Tech Jesus on X570 Video version: https://www.youtube.com/watch?v=LQHMUJXhxlc
|
# ? Jan 18, 2019 19:37 |
|
Has AMD said anything about Epyc chipsets being backwards compatible with Zen2? Everything I'm seeing talks about Ryzen specifically
|
# ? Jan 18, 2019 19:55 |
|
I know getting rid of ASmedia sounds great, but it also means x570 will be a reset to effectively a first-gen chipset for all the mobo makers and that doesn't fill me with confidence if I was thinking about putting together a new system for ryzen 3000, and I didn't expect to need the additional bandwidth of a pcie 4.0 connection from CPU to I/O, I might think about grabbing an x470 board for cheap
|
# ? Jan 19, 2019 01:19 |
|
Klyith posted:I know getting rid of ASmedia sounds great, but it also means x570 will be a reset to effectively a first-gen chipset for all the mobo makers and that doesn't fill me with confidence
|
# ? Jan 19, 2019 04:39 |
|
Porting down the Epyc chipset seems like a decent move in concept, especially if X570 will end up being the platform to try and tackle the I/O coming from 16 cores whenever that second chiplet space gets filled in. Who knows when it'll even have an appreciable effect though....
|
# ? Jan 19, 2019 13:16 |
|
I always thought ASmedia's involvement was for USB 3.0. Which is to say, they and Intel share patents and Intel sure wasn't going to help AMD develop a chipset.
|
# ? Jan 19, 2019 20:15 |
|
Klyith posted:I know getting rid of ASmedia sounds great, but it also means x570 will be a reset to effectively a first-gen chipset for all the mobo makers and that doesn't fill me with confidence It's based on a server chipset. I think we'll be alright.
|
# ? Jan 19, 2019 20:27 |
|
Craptacular! posted:I always thought ASmedia's involvement was for USB 3.0. Which is to say, they and Intel share patents and Intel sure wasn't going to help AMD develop a chipset. Aren't all the USB patents in an open license scheme?
|
# ? Jan 19, 2019 21:04 |
|
I don't think Epyc... has... much of a "chipset", does it? IIRC, it basically runs as a SoC in a lot of installations, with a lot of stuff directly connected to CPU lanes and/or FPGA. An advantage of everything being PCIe these days, and Epyc have 10 billion PCIe lanes. Like here is the block diagram for one of the popular 1 socket Supermicro boards for workstations: And heres one for a Gigabyte board: The quote was something like they wanted to leverage "Epyc I/O", whatever that means. In a Ryzen system, cause its got less lanes, the chipset works as a PCIe switch to hang various PHYs and such off of (Intel LAN, Wifi, etc), USB controller, and SATA controller. Here's it's block diagram: So I guess thats the gap they have to close, unless they are willing to integrate more stuff on to the package. Ryzen already has sound onboard, a 4 port USB3.0 hub, and non-intel gigabit LAN, but obviously theres still a good bit needed. (disclaimer that im just a dumbass that likes computers as a hobby) Cygni fucked around with this message at 23:53 on Jan 19, 2019 |
# ? Jan 19, 2019 23:51 |
|
Klyith posted:Aren't all the USB patents in an open license scheme? Cygni posted:I don't think Epyc... has... much of a "chipset", does it? The quote was something like they wanted to leverage "Epyc I/O", whatever that means.
|
# ? Jan 20, 2019 00:45 |
|
Cygni posted:I don't think Epyc... has... much of a "chipset", does it? IIRC, it basically runs as a SoC in a lot of installations, with a lot of stuff directly connected to CPU lanes and/or FPGA. An advantage of everything being PCIe these days, and Epyc have 10 billion PCIe lanes.
|
# ? Jan 20, 2019 01:01 |
|
Combat Pretzel posted:The Threadrippers are quasi-Epyc and they also have a chipset (X399, which also has the codename Promontory for whatever reason). Yeah, its the same Asmedia silicon as all the rest of the Ryzen chipsets. I imagine its the cheapest way to add the USB/SATA/Intel LAN/WIFI that desktop users expect to TR, while leaving enough PCIe lanes for multi-GPU.
|
# ? Jan 20, 2019 01:36 |
|
Threadripper should be batshit.
|
# ? Jan 20, 2019 01:42 |
|
Gamers Nexus is reporting their sources have informed them that Zen2 core count will be 8, 12, and 16 for the Ryzen 3000 series.
|
# ? Jan 23, 2019 03:39 |
|
OhFunny posted:Gamers Nexus is reporting their sources have informed them that Zen2 core count will be 8, 12, and 16 for the Ryzen 3000 series. Does their source sound... Scottish?
|
# ? Jan 23, 2019 03:55 |
|
Risky Bisquick posted:Does their source sound... Scottish? I think Tech Jesus is more reliable than that Scottish guy.
|
# ? Jan 23, 2019 04:06 |
|
OhFunny posted:16 for the Ryzen 3000 series. Do NOT expect these to be cheap.
|
# ? Jan 23, 2019 05:49 |
|
Seamonster posted:Do NOT expect these to be cheap. It could be if AMD opt for a low power/lower frequency process again.
|
# ? Jan 23, 2019 06:23 |
|
OhFunny posted:Gamers Nexus is reporting their sources have informed them that Zen2 core count will be 8, 12, and 16 for the Ryzen 3000 series. Edit: Never mind, I'm a dummy and misinterpreted the source info. spasticColon fucked around with this message at 10:03 on Jan 23, 2019 |
# ? Jan 23, 2019 07:02 |
|
That's not what he said. He said there will be 12 and 16 core processors in addition to the 8 cores we already know about, not that 8 cores is the baseline. Those are going to be the ultra-high end super expensive ones, maybe the TR binnings that clock 100mhz higher.
|
# ? Jan 23, 2019 07:06 |
|
So it's actually going to be 4, 6, 8, 12, and 16 for the core counts then? That still puts the 8-core CPUs in the mid-range offerings doesn't it? A possible Ryzen 5 3600 with 8C/16T for ~$250 would still be pretty sweet.
spasticColon fucked around with this message at 07:15 on Jan 23, 2019 |
# ? Jan 23, 2019 07:11 |
|
https://www.youtube.com/watch?v=oLlKDJfRRaY&t=150squote:Our understanding: the CPU core count should be 16, 12, and then your usual 8 for example, 8 4 whatever.
|
# ? Jan 23, 2019 07:18 |
|
spasticColon posted:If these "sources" are correct then holy poo poo. I'm now imagining an entry level Ryzen 3 3300 CPU with 8C/16T. If the sources are correct, it'll be just like the leaks. If so, that'll be flat out amazing.
|
# ? Jan 23, 2019 08:33 |
|
spasticColon posted:So it's actually going to be 4, 6, 8, 12, and 16 for the core counts then? That still puts the 8-core CPUs in the mid-range offerings doesn't it? A possible Ryzen 5 3600 with 8C/16T for ~$250 would still be pretty sweet. Packaging and IO die might make 4 core parts too expensive to bother with. I'd expect the pricing structure to remain closer to the 2000 series, with 12 and 16 slotting above that.
|
# ? Jan 23, 2019 10:02 |
|
Yeah, if a chiplet is defective enough to go down to only 4 cores being worthwhile, it might be better to just chuck the drat thing, or go 4+4 with another equally bum chiplet to make lower-spec eight-core parts. Of course, I also dream of 4 + 8 core poo poo specifically for running VMs, but I'm almost absolutely certain that anything above eight cores will have symmetric core counts. Anything intentionally designed to have 4 cores or fewer will probably going use Zen+ 4-core CCXes die-shrunk to 7nm.
|
# ? Jan 23, 2019 10:28 |
|
I'd expect them to repeat what they've done with the 2000 series and sell the APUs (now "12nm") for four cores and below.
|
# ? Jan 23, 2019 10:32 |
|
SwissArmyDruid posted:Yeah, if a chiplet is defective enough to go down to only 4 cores being worthwhile, it might be better to just chuck the drat thing, or go 4+4 with another equally bum chiplet to make lower-spec eight-core parts. Depending on exactly how the cache is laid out, a 4+4 part might have more L2/L3 than a single 8 core part, and because of how the IMU interfaces with main memory, might actually be better than a regular 8 core part, even if you burn a bit more power to get it.
|
# ? Jan 23, 2019 10:40 |
|
|
# ? May 17, 2024 00:03 |
|
Llamadeus posted:I'd expect them to repeat what they've done with the 2000 series and sell the APUs (now "12nm") for four cores and below. Honestly, I need a 7nm Zen2 APU with non-GCN graphics also on 7nm like needing to pee so bad your back teeth start floating. Just the sheer potential of such an idea has me anxious.
|
# ? Jan 23, 2019 11:41 |