|
Junior Jr. posted:If it does turn out to be really good, what's the best AM4 motherboards for it All we have are some comments from reviewers who got samples saying a mobo with a good power delivery system of some sort is key for overclocking the 8C Ryzen's. OC'ing the 4C or 6C Ryzen's could have much lower requirements. That and they're all saying there are some bugs with the BIOS's that are being fixed which is improving high speed memory support and performance but some things are still wonky.
|
# ¿ Feb 25, 2017 17:04 |
|
|
# ¿ May 15, 2024 12:40 |
|
quote:This is a Zen getting 33.99GB/s out of 2133Mhz memory, which has a theoretical maximum bandwidth of 34.128GB/s... meaning, epic efficiency.
|
# ¿ Feb 26, 2017 12:46 |
|
HalloKitty posted:Or buy the higher clocked stuff and enjoy the extra speed! Even for Intel's latest chips there are lots of work loads that are totally insensitive to bandwidth and see no improvement even with quad channel memory.
|
# ¿ Feb 26, 2017 14:03 |
|
Dante80 posted:Which - btw - may be a good thing. If some quirk of Zen's design makes it work just as well with the cheap rear end slower RAM as it would with the fast RAM that saves money. Really saving money is half the reason why people are interested in Zen at all and if it offers different ways to do that I'm all for it.
|
# ¿ Feb 26, 2017 16:47 |
|
I thought the NDA's lifted on the 28th? edit: WCCF says 28th some others say March 2nd.Paul MaudDib posted:Right now the AMD fanbase is a festering, oozing mass of hype and paranoia. Really judging anything by what you see on Reditt at this point is probably pointless. Exception would seem to be Trump voters but that is a whole other shitshow I'd prefer not to go into. Anime Schoolgirl posted:CL10 is unicorn status at that clock rate edit: WTF how is anything I posted fanboyish much less as rabid as what we're seeing on reditt?!\/\/\/\/\/\/\/\/\/\/\/ PC LOAD LETTER fucked around with this message at 06:49 on Feb 27, 2017 |
# ¿ Feb 27, 2017 06:08 |
|
Ihmemies posted:So Ryzen has 512K L2 while Intel has 256? How does that actually affect performance? Do programs need to be optimized for 512K? Only way to know is wait and see? Up until the last few years desktop work loads were still almost all about IPC and clockspeed for that peak single thread performance. Multi thread is starting to matter a whole lot more as more software is written to make use of it (4 thread CPU is probably the minimum you want to use right now, 8 threads will probably evolve to be the new sweet spot for peak performance vs cost over the next few years) but primarily single thread performance + clockspeed are still key for desktop work loads. Its quite possible some sort of software rewrite to take advantage of Zen's bigger L2 would help general desktop work loads but don't go expecting that sort of thing to happen.
|
# ¿ Feb 28, 2017 00:49 |
|
Dante80 posted:1T command rate? O_o
|
# ¿ Feb 28, 2017 13:04 |
|
HalloKitty posted:100Mhz above its stock boost (in a fully manufacturer supported manner) and it crashes? Ouch.
|
# ¿ Feb 28, 2017 15:58 |
|
Its not 2 seperate dies I think. That looks like that due to how they laid down the metal TIM in 2 different pads side by side. Also MCM's are real expensive.
|
# ¿ Mar 1, 2017 21:02 |
|
RyuHimora posted:Ryzen was never going to be better than Intel. But you're going to be hard-pressed to tell the difference between a Ryzen chip and an Intel chip in a game, especially in DX12 or Vulkan games. If you want something that will maaaaaybe go toe to toe with or even beat Intel's best you'll have to wait for Zen+. They'll probably sell it for a whole lot though. AMD was always perfectly willing to charge top dollar if they actually had a performance lead.
|
# ¿ Mar 2, 2017 11:29 |
|
SwissArmyDruid posted:Anandtech interview with Dr. Su: http://www.anandtech.com/show/11177/making-amd-tick-a-very-zen-interview-with-dr-lisa-su-ceo
|
# ¿ Mar 2, 2017 14:20 |
|
Boiled Water posted:Unless you happen to also need large amounts of memory or any amount of ECC memory. If you just want some ECC some of the AMD mobo's will support it. Asrock lists support for it in some of their AM4 mobo manuals at least so its unofficial but its there at least. BurritoJustice posted:It's really not a "small gaming performance loss", it's a serious tradeoff and a question of priorities for potential buyers. PC LOAD LETTER fucked around with this message at 17:15 on Mar 2, 2017 |
# ¿ Mar 2, 2017 17:13 |
|
feedmegin posted:Also, remember the last time AMD had worldbeatingly good chips? Remember how they did not end up in professional server/workstation type kit regardless? Intel has that poo poo on lock. At one point they had something like 20% of the server market (looong time ago now) so they were making some in roads there. It just takes a long time to happen. As a few other posters have noted generally about 2-3yr. Intel seems to be slow at updating their server stuff so Zen will actually have a decent shot at getting some lucrative server market share for AMD even with Purley coming.
|
# ¿ Mar 2, 2017 17:18 |
|
Twerk from Home posted:Where is AVX2 expected to matter and do any of these reviews look at AVX2 256-bit wide commands? HPC stuff. Doesn't matter really for games or desktop stuff. Maybe in several years that will change but not a whole looks to benefit from it on the desktop. Maaaybe physics engines in games? That could just as easily get tossed at the GPU though.
|
# ¿ Mar 4, 2017 12:06 |
|
Says the difference from DDR4 2133 (CL10) to DDR4 3466 (CL14) is about 10% increase in performance which is noticeable. Weird to see performance improve that much from RAM speed. I wonder if the firmware updates they're going to be doing will make a difference there. DDR4 3466 (CL14) isn't easy to achieve on Ryzen right now. Some AMD specific DRAM is coming for Ryzen from GSkill I think so maybe that will help. HalloKitty posted:This launch has kind of sucked really badly for AMD. Realistically they seem to be selling well if some of the rumors are anything to go by so people are at least giving them some benefit of the doubt. But yeah delaying another month or 2 to let things get fixed would've been a drat good idea. I have no clue why they rushed things. \/\/\/\/On the value stuff yeah it really could be that simple, the higher performance stuff might be a different story\/\/\/\/ PC LOAD LETTER fucked around with this message at 13:47 on Mar 4, 2017 |
# ¿ Mar 4, 2017 13:36 |
|
ConanTheLibrarian posted:AMD is still a loss making company. Cash flow problems could have dictated that they release Ryzen as soon as they could. I vaugely remember some stock guy saying AMD would fail by 2018 or so if they didn't turn things around but that was almost 2yr ago now I believe. edit: yeah, about a 3 month delay, another month or might've been OK. Certainly would've helped their PR.\/\/\/\/ PC LOAD LETTER fucked around with this message at 17:57 on Mar 4, 2017 |
# ¿ Mar 4, 2017 16:52 |
|
I love cats posted:Six months ago when RX480 was released, AMD touted that 1080p gaming is where it is at, and buying that, you buy into the future. Now, with Zen, they say 1440p and 4k is where it is that and it is buying into the feature. And people just eat that poo poo up and don't take a total of a second to consider that contradiction? What AMD said about 1080p gaming resolution was that its one of the most common used and they put out a product that does fairly well at that resolution for a good price (ie. 480 GPU). Which is indisputably true in of itself. What is happening now is totally different and being done for totally different reasons. People are using (so called, 1080p isn't really low to me but whatever) "low resolution" gaming benchmarks to alleviate any GPU bottlenecks and to bring out differences between different CPU's to show which is more powerful and/or as a indicator of performance in future games. Some are looking at these benchmarks and seeing them as indicators of Zen being too mediocre for gaming to consider buying vs a OC'd or even stock 7700K. Others are saying those benchmarks aren't a particularly good indicator of gaming performance for Zen from a practical standpoint (framerates are often so high that even a 240Hz monitor couldn't display them all at that resolution, other things like frame latency (smoothness) aren't being counted or considered, etc.) and that at resolutions many people would prefer to play at (ie. 1440p or higher) the difference is negligible now and likely to get smaller as time goes on and more software makes better use of more threads. IMO given the odd bugs that still need working out and that current benches are being done without proper windows support the results for Zen are pretty drat solid and generally a bit better than I expected. One of the guys in the review videos linked in this thread called the "Mario from Super Mario Kart" of CPU's (ie. good at everything but not the best at anything either) and I think that was spot on. Given the fairly reasonable price they're offering for it I think there is a solid value perspective (yes even for gaming but in particular for productivity software) to consider here and I'm totally at a loss regarding all the drama and strawman smacking that has sprung up around Zen's gaming benches. PC LOAD LETTER fucked around with this message at 08:57 on Mar 8, 2017 |
# ¿ Mar 8, 2017 08:37 |
|
AVeryLargeRadish posted:So in the end I don't think it makes much sense to switch to Ryzen if you already have a decent CPU for gaming I think its reasonable to say that software programmers in general are quite aware by now that much more clockspeed isn't in the cards and neither is lots more IPC but more threads will be available + console constraints will push the games to make use of more threads over time too. PC LOAD LETTER fucked around with this message at 09:15 on Mar 8, 2017 |
# ¿ Mar 8, 2017 09:12 |
|
Pryor on Fire posted:What the gently caress is wrong with you I'm p. sure 1600p users are the type to roll their own crypto
|
# ¿ Mar 8, 2017 16:43 |
|
Measly Twerp posted:A guy with very interesting hair talks about memory compatibility: That is actually a pretty interesting and informative video. If you're looking for higher clocking DRAM for Zen you should watch it. The tl&dw is there are only 3 mobo's right now that got a good shot at running DRAM that is clocked at 3200Mhz or higher and that GSkill actually had to do some real work on their DIMM's to make it work.
|
# ¿ Mar 8, 2017 19:26 |
|
ECC DRAM tends to lag regular old DRAM in speed significantly in release times. If someone like Micron is just releasing DDR4 2666 ECC DRAM I wouldn't expect to see DDR4 3200 ECC DRAM before the end of the year, maybe later.
|
# ¿ Mar 8, 2017 20:10 |
|
Gray Matter posted:the 1800X single-core performance was far enough below expectations that I lost faith in the lesser chips What performance were you expecting then?! AMD's own numbers for the longest time (ie. 40% better IPC than Excavator) and even those around launch (ie. 52% better than Excavator "beat our own goal!" etc.) were going to put it behind Skylake much less Kabylake. The cool thing about that though is that its still going to perform well enough for most things you won't see a difference in practical real world use. It was obvious from the get go they were going to be offering bang for the buck instead of going for the performance crown. If Zen+ really does achieve a consistent 15% IPC improvement over Zen then yeah they'll have a good shot at taking the performance crown back from Intel for the first time in a real long time but I doubt you're going to like the price they'll sell them for because it probably won't be cheap.
|
# ¿ Mar 9, 2017 07:34 |
|
Junior Jr. posted:If those 1800X benchmarks are that good on a 1080 Ti, I'm having even less hope for Vega. If they're really going to stay competitive at this point, they have to be at least on par with Pascal (especially with 1800X) and not be 10 or so frames behind.
|
# ¿ Mar 10, 2017 08:20 |
|
AMD tends to only use LGA's when they're forced to and the server sockets have a truly insane amount of pins that are needed due to all the memory channels, interpackage buses, and extra cores they have to support.
|
# ¿ Mar 19, 2017 12:50 |
|
If you don't mind DDR4 2133 or slower there are plenty of 32GB kits that will work. Any of them really. Its the faster stuff that has a hard time working. Generally getting to DDR4 3000 or higher is really hard if you fill all 4 DIMM slots. And the dual sided high cap (so pretty much all 16GB DIMMs right now) are very hard to get working at over DDR4 2666. If you give it a couple more months more kits should show up and the firmwares will mature too but right now you're kinda boned if you want a large amount of RAM + high RAM speeds with any Ryzen. Unless you're willing to try to overvolt the crap outta some "B" die DRAM like Buildzoid and other extreme OC'ers but I wouldn't trust that to last 24/7 for long.
|
# ¿ Mar 22, 2017 05:12 |
|
SwissArmyDruid posted:my disgust with this situation and how I keep seeing "DDR4 2400 MHz and higher memory module will only run at maximum of DDR4 2400Mhz due to AMDŽ chipset limitation when using 7 th Gen A series CPU" on every single mobo manufacturer's website. Its just right now Intel's BIOS is more mature + the DRAM market has had more time to tune the memory for Intel's stuff is all. Paul MaudDib posted:This poo poo isn't really something you're going to fix with BIOS updates. It's something that's pretty much inherent to the v1.0 silicon. I don't know why they set it like that and why they made it almost impossible to change right now but that is a big reason why its so hard to get higher clocked DRAM for Ryzen. Very hard to get high clocks + tight timings out of any DRAM period even for Intel too. PC LOAD LETTER fucked around with this message at 06:19 on Mar 22, 2017 |
# ¿ Mar 22, 2017 06:16 |
|
Lolcano Eruption posted:I'm curious. If that stick is rated for 3200 at 1.35 volts, why would we need to increase the voltage? For latency this is good and lower latency main memory is generally a good thing. For top DRAM clocks, its bad. Right now people are focusing lots on pumping up the DRAM clocks since that also improves the IF bus speeds which can give you a nice (~10%) boost in performance, particularly in games, if you can get it up to DDR4 3200. You'll still see nice performance scaling at DDR4 2933 though. Sometimes more volts can help with that issue. Sometimes it won't since lots of DRAM just can't run fast with a 1T command rate or other lower timings. Don Lapre posted:Any speed over 2133 is considered overclocking. PC LOAD LETTER fucked around with this message at 11:47 on Mar 23, 2017 |
# ¿ Mar 23, 2017 11:42 |
|
I would assume its still a driver issue at this point since they didn't even SATA drivers in that article to test with. If after a month or 2 of driver updates down the road performance is still lagging for NVMe SSD's then AMD probably screwed up somewhere in the hardware. That being said performance, while quite a bit worse than Intel at the moment, isn't actually bad for desktops. That is from a practical typical user perspective you might possibly see boot and program load times as maybe several seconds or so slower than on a modern Intel platform right now. That isn't good but calling it poo poo doesn't seem correct either. Its still a huge step up from their older socket AM3/FM3 platform SATA and I/O performance. Personally I've only messed with 1 AM4 system and while I didn't do any benchmarking it sure "felt" as fast and as smooth as any modern Intel system when loading up CS:GO, PoE, Libreoffice 5.3, and booting Windows 10 Pro.
|
# ¿ Apr 2, 2017 04:20 |
|
wargames posted:Didn't intel help develop the nvme standard, if so i would assume they would have better drivers for it. Given the way BIOS improvements are already making a impact on overclocking and memory support, as well as bug fixes, I think there is good reason to believe AMD is serious about and able to work out the current issues. Especially since the server platform Zen's should be launching within several months.
|
# ¿ Apr 2, 2017 06:18 |
|
Truga posted:I don't think CL value affects the speed of the fabric though? You're right it doesn't. The DRAM speed does so setting a CL 17 stick to CL 15 won't do anything to the IF clocks but setting a stick of 2133 or 2400 DDR4 to 2933 or 3200 will effect the IF clocks and apparently can reduce inter L3 caches latency significantly while also boosting inter L3 cache bandwidth but I believe its currently unknown if bandwidth is a big factor there. Going by the Pcper article on the subject recently it really seems like inter L3 cache latency is the issue.
|
# ¿ Apr 6, 2017 13:10 |
|
|
# ¿ May 15, 2024 12:40 |
|
Anime Schoolgirl posted:I think the scuttlebutt is that inter-L3 cache latency between each 1.5-2.5mb per-core L3 for Intel's post-Sandy Bridge uarch is a constant thing and can be accounted for, but AMD's CCX 8mb is actually fully shared between a four core block and it's probably not as transparent to the OS whether it switches between cache banks or not? Apparently MS's default power savings methods were also causing some issues (they assume Intel style power management) and the recent fix AMD released can make a noticeable difference too.
|
# ¿ Apr 7, 2017 11:26 |