|
Paul MaudDib posted:It drives me insane that USB flashback isn't on every board nowadays. Why, have you had many BIOS flash failures? Trying to think but I can't recall any flashes that failed for me. I suppose for a 'bricked' mobo it could be handy. I really like the boards with 2 bios chips and a little switch you can toggle for boot/flash.
|
#
?
Oct 16, 2017 15:01
|
|
|
|
| # ? May 16, 2024 21:01 |
|
|
FaustianQ posted:Guy who works at ASUS on BIOS's says the following at Overlock http://www.overclock.net/t/1624603/rog-crosshair-vi-overclocking-thread/28240#post_26391661 More like Ayyyy420 Sorry. In other news, we have marketing materials for an R5 laptop from HP: http://www8.hp.com/h20195/v2/GetPDF.aspx/c05652570.pdf It's an Ultrabook type thing, so I'm not expecting much graphics wise, but it's nice to see RR progress.
|
|
#
?
Oct 16, 2017 16:04
|
|
|
NewFatMike posted:More like Ayyyy420 Also some confirmation on speeds (2.0Ghz base, 3.6Ghz boost) and 6MB L3 Cache (not 4MB as speculated). iGPU is called Vega M. This seems pretty mature so I'm expecting an announcement closer to the end of October and actual release, paper or not into late November, early December.
|
|
#
?
Oct 16, 2017 16:17
|
|
|
redeyes posted:Why, have you had many BIOS flash failures? Trying to think but I can't recall any flashes that failed for me. I suppose for a 'bricked' mobo it could be handy. I really like the boards with 2 bios chips and a little switch you can toggle for boot/flash. No, but I don't want to hang onto a first-generation chip for the sole purpose of flashing a BIOS upwards before I can use a later chip.
|
|
#
?
Oct 16, 2017 16:19
|
|
|
Also  HP, why are you using single channel TYOOL2017-18?!
|
|
#
?
Oct 16, 2017 17:38
|
|
|
Truly, we live in the post-apocalyptic future and mega-corporations worldwide must engage in cage matches for ocean barges full of DDR4 SO-DIMMs I'm pretty excited to see how Vega M
|
|
#
?
Oct 16, 2017 19:18
|
|
|
Single channel is ok if its a) DDR4 and b) for lovely $400 atom based stuff. That said, even dual channel DDR4 2400 is "only" around 32-35 GB/s which is well below the MX150's bandwidth which is the chip I'd immediately compare any ultrabook mobile vega variant to.
|
|
#
?
Oct 16, 2017 19:24
|
|
|
NewFatMike posted:More like Ayyyy420 Holy poo poo I beat WCCFTech to a story: http://wccftech.com/amd-ryzen-5-2500u-raven-ridge-apu-with-radeon-vega-m-gpu-spotted/ Also I think the 16GB SKUs will probably be dual channel. Doesn't help with 2400MHz, but in the mass market Envy line (instead of enterprise ones like the Elitebooks) I'm not totally surprised. Hopefully there will be SKUs where graphics performance is a priority.
|
|
#
?
Oct 16, 2017 19:42
|
|
|
It's probably HP trying to skim money off the cost of an AMD laptop, cause they don't think AMD laptops will sell at a higher price. This has been a problem with AMD laptops since the friggen K6-2 days. It's a shame because single channel DDR4 is going to starve the hell out of a 4 core Zen + 704CU graphics unit. For comparison, the Radeon RX 550 has 512CU but 112gb/s of bandwidth just to itself....
|
|
#
?
Oct 16, 2017 19:44
|
|
|
NewFatMike posted:Holy poo poo I beat WCCFTech to a story: wccftech just rips their info off twitter and reddit, but computerbase might have been first https://www.computerbase.de/2017-10/raven-ridge-hp-notebook-amd-ryzen-5-2500u/ 16.10.2017 9:36 Uhr
|
|
#
?
Oct 16, 2017 20:55
|
|
|
Cygni posted:It's probably HP trying to skim money off the cost of an AMD laptop, cause they don't think AMD laptops will sell at a higher price. This has been a problem with AMD laptops since the friggen K6-2 days. It's a shame because single channel DDR4 is going to starve the hell out of a 4 core Zen + 704CU graphics unit. For comparison, the Radeon RX 550 has 512CU but 112gb/s of bandwidth just to itself.... Except that the RX 550 is a 45-50 watt part and the whole Raven Ridge package is supposed to be less than that.
|
|
#
?
Oct 16, 2017 21:03
|
|
|
Seamonster posted:Except that the RX 550 is a 45-50 watt part and the whole Raven Ridge package is supposed to be less than that. Totally true! The RX 550 is clocked at 1183 boost and Raven Ridge looks to be at 800mhz. Also memory bandwidth doesn't make a huge difference to a lot of CPU based tasks... but it DOES make a huge difference to a lot of GPU tasks. Peak transfer rate for single channel DDR4 2400 is like 19gb/s... thats not a lot. AMD hasn't released a retail GPU with bandwidth that low since like the 6450 in 2011, and it only had 160 Caicos CUs to feed, not 704 Vega CUs AND an 8 thread CPU. Hopefully laptop makers don't all cheap out like this and hobble things.
|
|
#
?
Oct 16, 2017 21:22
|
|
|
I'm personally expecting regular Intel HD graphics to get smacked pretty hard if AMD can pump out enough Raven Ridge parts in volume. The Iris Pro equipped stuff with eRAM will get overtaken by a smaller margin but those parts' real downside is the massive cost for what they are anyway so gently caress 'em.
|
|
#
?
Oct 16, 2017 21:35
|
|
|
I got some water cooling in the past two weeks but it seems like I have a good launch bin, 1707. AMD  https://valid.x86.fr/9zvgxn 3.925 @ 1.31 https://valid.x86.fr/bfysv0 3.7 @ 1.2
|
|
#
?
Oct 17, 2017 04:12
|
|
|
Risky Bisquick posted:I got some water cooling in the past two weeks but it seems like I have a good launch bin, 1707. AMD Does seem you did. Does it pass Prime95? At 3.8 mine will go for 4-5 hours then fail on a single core.
|
|
#
?
Oct 17, 2017 11:20
|
|
|
What's a 'good' or at least OK temp as reported by Ryzen Master? Playing Bethesda Post-Apocalyptic Inventory Management Simulator 4 hard-locked my whole machine the other night and I brought up RM as fast as I could on boot to check the temp which was at 45 C then immediately dropped to 36. Thing is, it changed so fast that I assumed the first number was a placeholder that it shows before it gets real data, so IDK if that was real or not. Heck, the CPU might have cooled off during boot for all I know, even though it was really fast.
|
|
#
?
Oct 17, 2017 17:07
|
|
|
Munkeymon posted:What's a 'good' or at least OK temp as reported by Ryzen Master? Processors can heat up/cool down quite quickly, there's not a lot of "thermal mass" there. My 5820K (soldered, like Ryzen) can go from 70C under Prime95 to 40C in a matter of seconds with my AIO. Let's say it'll swing at least 10 degrees per 5 seconds, probably closer to double that. Any numbers you got after a reboot are effectively meaningless. If you're worried, I'd use the RivaTuner overlay so you can watch temps while you're playing. Have you run the kill-ryzen script to see if your processor suffers the "Performance Marginality Issue"? Contrary to the folk wisdom, issues can manifest in Windows (usually as hardlock or BSOD), they're just much more common under Linux for some reason. Paul MaudDib fucked around with this message at 17:17 on Oct 17, 2017 |
|
#
?
Oct 17, 2017 17:12
|
|
|
Paul MaudDib posted:Processors can heat up/cool down quite quickly, there's not a lot of "thermal mass" there. My 5820K (soldered, like Ryzen) can go from 70C under Prime95 to 40C in a matter of seconds with my AIO. Let's say it'll swing at least 10 degrees per 5 seconds, probably closer to double that. Any numbers you got after a reboot are effectively meaningless. I missed that but I'll try it hopefully tonight, though. E; hold on - isn't that a Linux thing? I'm on Windows here, which I guess is technically a VM since I never disabled Hyper-V, but I haven't seen anyone talking about making sure to disable Hyper-V with Ryzen. E2: oh, didn't see your edit. Is there a Windows version of that script or would I need to make a bootable USB to try it? Munkeymon fucked around with this message at 17:46 on Oct 17, 2017 |
|
#
?
Oct 17, 2017 17:16
|
|
|
We got Raven Ridge perf leaks: In short: An MX150 competitor. Kind of disappointing, really. https://www.pcper.com/news/Graphics-Cards/AMD-Raven-Ridge-Performance-Leaks-APU-GeForce-MX150-Performance
|
|
#
?
Oct 17, 2017 20:09
|
|
|
bobfather posted:Does seem you did. Does it pass Prime95? At 3.8 mine will go for 4-5 hours then fail on a single core. 3 hours prime stable & 3 hours game + obs cpu medium x264 encode stable
|
|
#
?
Oct 17, 2017 20:12
|
|
|
Munkeymon posted:I missed that but I'll try it hopefully tonight, though. E; hold on - isn't that a Linux thing? I'm on Windows here, which I guess is technically a VM since I never disabled Hyper-V, but I haven't seen anyone talking about making sure to disable Hyper-V with Ryzen. Make a bootable USB stick. It also needs a fair bit of RAM unfortunately.
|
|
#
?
Oct 17, 2017 20:24
|
|
|
SwissArmyDruid posted:We got Raven Ridge perf leaks: I mean, what were you expecting? Zen doesn't really exceed Intel's efficiency and Vega certainly doesn't exceed Pascal's efficiency (let alone when you're comparing an iGPU against a discrete chip that's using GDDR5). So MX150 performance is pretty much the upper bound there. Assuming it hits its TDP target, that's not actually terrible for 35W. The main design goal is to let cheapass system integrators avoid using a discrete chip, not top-shelf performance. Paul MaudDib fucked around with this message at 20:33 on Oct 17, 2017 |
|
#
?
Oct 17, 2017 20:30
|
|
|
.
sincx fucked around with this message at 05:50 on Mar 23, 2021 |
|
#
?
Oct 17, 2017 21:09
|
|
|
SwissArmyDruid posted:We got Raven Ridge perf leaks: You...you do realize that means the 640SP Vega iGPU is competing with the RX 550 dGPU with 128 bit bus and GDDR5, right? Did you expect it to compete with the RX 560? What's more interesting is that the 512SP of the 2500U net a score only 7-9% slower while having 20% fewer shaders (also physics is much better, so it's clear that the extra shaders are just being a power hog), kind of indicating that desktop Raven will do even better once outside thermal constraints. Seems like for Mobile, a cheap laptop with (dual channel because  HP) the 2500U will be the one to get, the 2700U seems to be for suckers for the likely poorer price to performance ratio.Also the best Intel can really muster in competition, the 650 is not only rare and expensive but scores only 2762. Sounds like even Ravens low end will be better than what Intel can offer for a premium. Only expensive, 1000+ laptops with at least a mobile 1050 and Coffeelake mobile will be worth it on the Intel side, as long as OEMs don't loving cripple Raven. Like I originally thought that Intel and AMD would at least be competitive but these scores for Raven kind of indicate it's a full on slaughter. EDIT: Iris Pro 580 scores 3220 in 3dMark11. Like LMAO, the 384SP variant is probably only 10% slower assuming perfect scaling in performance and that's going to be AMDs low end solution. Like, assume the Vega M 384SP is in fact 25% slower than the Vega M 512SP for having 25% less shaders and low end Vega beats the 650 and starts competing with the 580. EmpyreanFlux fucked around with this message at 21:36 on Oct 17, 2017 |
|
#
?
Oct 17, 2017 21:27
|
|
|
quote:The AMD Ryzen 7 2700U with integrated Vega-based graphics architecture That one slideshow urged me to expect quad-core CPUs.  I wanted graphics on an enthusiast CPU and hope springs eternal and all that, but I figured AMD would forever delay it because of room the die and also because they want you to at least buy a 550 from Raja. EDIT: I've been informed my fears are indeed the case, and this is why the chip has a "U" Craptacular! fucked around with this message at 21:49 on Oct 17, 2017 |
|
#
?
Oct 17, 2017 21:41
|
|
|
Mobile naming is different from Desktop FWIW. The 2700U is indeed 4C/8T, seemingly 10CU. The 2500U is 4C/8T, 8CU overall worse binning.
|
|
#
?
Oct 17, 2017 21:49
|
|
|
Question: iirc, NVIDIA Optimus can work on a per-window level, i.e. it will render a given window in 3D mode and then blit back to the iGPU across PCIe for compositing. Does AMD Hybrid Graphics have a similar feature?
|
|
#
?
Oct 17, 2017 22:21
|
|
|
Also the relative performance of the Vega iGPU kind of puts to rest a dGPU replacement of Polaris 12. I suspect 1024-1280 shaders is now the lower bound for any new dGPU product from AMD. They'll probably just iterate Polaris 12 until 2024 or so for people who need some kind of display capability so as not to totally cede that market to Nvidia though, so basically the new Oland/Caicos.
|
|
#
?
Oct 17, 2017 23:08
|
|
|
FaustianQ posted:Mobile naming is different from Desktop FWIW. The 2700U is indeed 4C/8T, seemingly 10CU. The 2500U is 4C/8T, 8CU overall worse binning. Haha, I spent a while looking around trying to find exactly this info and it was here all the time. So this is a high-end-ish iGPU cutdown but they can probably turn out a 1-CU die harvest in pretty significant volume. In terms of performance, the leaks have it about 20% under a MX150, which is basically a 1030. A 1030 is about 0-20% under a 750 Ti or an RX 550 depending on game/res. So, low-settings 1080p/mid 720p-ish territory. APUs are probably gonna be a good bread-and-butter seller for them, especially since APUs support freesync (note: actually since Kabini, it went in as part of the GCN 1.2 GPU upgrade). That'll be a super good experience in the bargain-basement <$500 PC market, low framerate compensation is ideal for weak-rear end PCs and console hardware. They can turn it out pretty cheaply for laptops too - Mobile GSync still hasn't taken off much except with gaming laptops. They can push that with Radeon Chill (framerate target) and cut power draw a decent bit, which vendors will like. (they'll plow it right back into a thinner/lighter cooling solution  )That's a waaaay better outlook than the discrete GPU market seems to be going for them. I guess the next big hope there is that Vega 11 ends up being a better balance of whatever, and/or has magical hardware fixes that produce speedups. They just need to stop shipping segfaulting processors and they'll be golden, because I can't see that poo poo flying with vendors or customers. I'm surprised it's flying in the desktop market as much as it has. Maybe put Raven Ridge on the B2 stepping if that has a fix... Paul MaudDib fucked around with this message at 00:47 on Oct 18, 2017 |
|
#
?
Oct 18, 2017 00:29
|
|
|
I think the performance difference between the Vega iGPUs and Vega 56 and 64 is kind of telling. AMD tends to design everything to be rather bandwidth hungry and latency sensitive, so improvements in those areas tend to greatly improve performance more so than pushing the core. Memory design just seems to be a really weak spot for them. Vega 10 might also be ROP bound. I'm guessing that Vega 11 will stronger than expected performance based on what we know of Vega 10 so far, like solidly better than the Fury X (10% is my guess) at stock and decently low power consumption, because of the better bandwidth to ALU ratio. I think something like GDDR6 will be critical to get the most out of Vega 11. Also the current 512SP A12-9800 does reasonably well in two extremely popular games, Overwatch and PUBG, so I'm expecting excellent 1080p performance in anything pre 2015 and pretty fluent 720p even in modern stuff, I mean based on some benches I've seen for BF1 using an RX 550 I think it can hit 1080p30 medium, that's loving impressive for an APU. Honestly if they can hit 500-600$ for a good quality laptop with a 2500U it sounds pretty drat perfect. I'm betting the dies with all 11CUs are going to the desktop or it's some halo tier product above the 2700U.
|
|
#
?
Oct 18, 2017 00:47
|
|
|
FaustianQ posted:I think the performance difference between the Vega iGPUs and Vega 56 and 64 is kind of telling. AMD tends to design everything to be rather bandwidth hungry and latency sensitive, so improvements in those areas tend to greatly improve performance more so than pushing the core. Memory design just seems to be a really weak spot for them. Vega 10 might also be ROP bound. I'm guessing that Vega 11 will stronger than expected performance based on what we know of Vega 10 so far, like solidly better than the Fury X (10% is my guess) at stock and decently low power consumption, because of the better bandwidth to ALU ratio. I think something like GDDR6 will be critical to get the most out of Vega 11. TBH delta compression is one specific thing you can point to that they fell behind in and is seriously dragging them down. This is another major ingredient in Maxwell's "secret sauce" and they got another gain with Pascal too. Vega attempts to catch up but iirc they're still a ways behind. Yeah I agree that sounds plausible. Basically a Polaris 10 replacement (like a RX 680 in terms of market placement) with Fury X/1070/1160 level performance. Vega is super bottlenecked in other places (definitely VRAM b/w, possibly ROPs, and I think geometry still isn't great), and there are a crazy amount of cores sitting idle right now (Vega 56 with V64 VRAM voltage flashed performs pretty damned close to a real V64). As always there's a lot of variables getting there, that would depend on RTG executing well. Hot take: Vega 10 would be a better chip if it had been designed around GDDR5. They are 100% reliant on HBM2 coming down to ever sell this chip, and the RAM companies are just choking production like crazy, of even commodity stuff let alone specialty products. In combination with fast RAM and the other gains they would be fine with a 512-bit bus on Vega 10. It's not like NVIDIA is getting off much easier, GP102 uses a 384-bit bus (the delta between them there, that's AMD's R&D deficit). But I guess a "more efficient" bus that can't actually feed the cores is better, having idle cores clocked up is awesome for efficiency right? And they ended up having to overvolt it to hell anyway just to get the clocks up a bit closer to the design spec... Hindsight is 20/20 but they drastically needed to leave themselves a little more safety margin. A little bit of overprovisioning like Tonga (384-bit card with only 256-bit ever populated) would have been helpful, like a bus that can support a third stack of HBM2 on a new interposer as a fallback plan. With more VRAM bandwidth, I think it'd probably be more competitive with a 1080 Ti, which would be a lot more marketable. I mean they actually reduced bandwidth from Fiji  IIRC NVIDIA can do geometry processing on the shaders, so they can scale it while AMD is trapped with a max of 4 fixed Geometry Engines, right? What's their deal there? They really need to fix that, or put it on its own uncore clock... Paul MaudDib fucked around with this message at 19:02 on Oct 18, 2017 |
|
#
?
Oct 18, 2017 01:06
|
|
|
 So release date confirmed? The wording makes it sound like AMD has secured a couple design wins only for 2017, and will just be releasing RR in general, with real OEM designs for 2018. *sigh* On the other hand, just a thought but since the old Steamroller and Excavator chips had a theoretical quad channel IMC could Raven Ridge have carried this over? Would be interesting to do a quad channel halo tier laptop or surface product.
|
|
#
?
Oct 18, 2017 13:59
|
|
|
What are the chances that Ryzen 2 will get substantially better with more IPC, higher clock speeds, etc and will make i5 8400 early purchasers regret their decision?
|
|
#
?
Oct 18, 2017 15:30
|
|
|
We have no idea. Like best speculation is the new process advancements boost clocks to 4.4-4.5Ghz, which is stronger but not in "regrets buying coffeelake" territory.
|
|
#
?
Oct 18, 2017 15:32
|
|
|
Otakufag posted:What are the chances that Ryzen 2 will get substantially better with more IPC, higher clock speeds, etc and will make i5 8400 early purchasers regret their decision? i5-8400 is still sparsely available, and no cheap motherboards are available. Meanwhile Newegg has a 1600x + Asrock X370 combo for like $250 after rebate right now. You might be able to do even better if you're near a Microcenter. I could see this becoming the new "normal" price for this kit in less than a month, with specials bringing it closer to $200 total for the kit. By the time cheap boards are available for the 8400, we'll be close to the Zen refresh. Should be interesting!
|
|
#
?
Oct 18, 2017 15:41
|
|
|
SwissArmyDruid posted:We got Raven Ridge perf leaks: Kind of a weird way to spell “amazing”. The MX 150 has a 25W TDP all on its own, a chip keeping all that in a 35W envelope rules pretty darn hard. Hopefully the higher-watt mobile SKUs will be able to handle medium 900p/1080p@30FPS. This one is awfully close.
|
|
#
?
Oct 18, 2017 16:47
|
|
|
Otakufag posted:What are the chances that Ryzen 2 will get substantially better with more IPC, higher clock speeds, etc and will make i5 8400 early purchasers regret their decision? Zen 2 isnt coming out until 2019, so I doubt people buying 1600/8400 now will regret it. Zen Refresh comes out in March-ish if the rumors are true, but there are no architectural changes and the only difference is a 'half-node' change with smaller libraries in places. No IPC changes, and likely not a huge amount of clocking differences. My personal all conjecture take is that the next few months will be a great time to buy either platform (once 8400 is in stock consistently and AMD cuts prices to match)
|
|
#
?
Oct 18, 2017 16:54
|
|
|
NewFatMike posted:Kind of a weird way to spell “amazing”. The MX 150 has a 25W TDP all on its own, a chip keeping all that in a 35W envelope rules pretty darn hard. Hopefully the higher-watt mobile SKUs will be able to handle medium 900p/1080p@30FPS. This one is awfully close. Yea, it can't be stated enough that if those 3Dmark11 scores for the Vega gpus are representative, then it really seems to be that Vega just scales like dogshit, both in ALU and frequency. I wouldn't be surprised to see 65W desktop Raven parts actually beating the GT 1030, free of the mobile variants thermal constraints and access to faster DDR4. AMD needs to get on the MiniSTX ball with Raven goddamn it's so loving perfect.
|
|
#
?
Oct 18, 2017 17:33
|
|
|
bobfather posted:i5-8400 is still sparsely available, and no cheap motherboards are available. But you have to buy good RAM ($$$becauseDRAMpricingisfuuuuucked$$$$) to get the better out of any Ryzen so don't think you can just plonk down a couple of bills and call it done. Still better than Intel pricing on mobo+cpu.
|
|
#
?
Oct 18, 2017 19:03
|
|
|
|
| # ? May 16, 2024 21:01 |
|
|
One other thing Zen+ may have is IMC maturity, so rather than having to hunt for Samsung B-Die or some poo poo you can just get the cheapest 3200+ and may natively support 3200Mhz as stock speed. Basically Zen+ may not promise much more than not having to RAM and a few more Mhz, but that actually means a lot.
|
|
#
?
Oct 18, 2017 19:15
|
|