|
Subjunctive posted:Then why do you want updated games? Are they CPU-limited at 1440/4K now, such that Ryzen optimizations would show up? It doesn't hurt if you can go even faster theoretically. Tbh some games like Ashes of singularity has terrible performance with Ryzen, would be interesting to see if an update would help with those.
|
# ? Mar 5, 2017 19:41 |
|
|
# ? May 22, 2024 04:49 |
|
Dante80 posted:Yes. My C2D still runs games valiantly. Wish I did get the Q6600 back then though. The Westmere X5690 (i7-990X equivalent) in my 2009 Mac Pro still does just fine in every game I play with a GTX 980.
|
# ? Mar 5, 2017 21:10 |
|
K8.0 posted:Twitch and youtube aren't at all alike. Twitch gives you a ton of leeway with your settings without reencoding, which means you can just do your comparisons locally, and it's readily apparent that GPU encoding blows. Twitch has recently started recoding many streams (used to be for partners only), though you can always still choose source when watching, which is nice. And yeah, I think I mentioned this earlier, but gpu encoded videos can already suffer pretty hard in high motion scenes, and then youtube/twitch does another pass over that, and while they do have pretty decent transcoding going on, the shittier your source, the extra shittier the output, since the blockyness/blur just adds on top of blur and rip your stream.
|
# ? Mar 5, 2017 21:21 |
|
ufarn posted:Would be cool if someone did the OBS test again and included the AMD encoder so we could just compare the videos side by side on YouTube. NewFatMike posted:Wouldn't the YouTube encoding gently caress that? https://www.youtube.com/watch?v=PK6IVEfJLqk
|
# ? Mar 5, 2017 21:23 |
|
Dante80 posted:Yes. My C2D still runs games valiantly. Wish I did get the Q6600 back then though. These days used CPUs of any generation older than Sandy Bridge are rather cheap, to the point that you could get something like a Q9400 or X3350 for $20 if you're still using that system and care. Bloomfield/Westmere and Lynnfield systems are an especially good value proposition; I got an old 1156 board the other day and was able to push an X3440 that I picked up for $25 up to ~4GHz, not really much worse than my main system's 2500K at 4.4.
|
# ? Mar 5, 2017 21:25 |
|
Cardboard Box A posted:You tell me I can't reproduce such good results even with 4000kbps doing slow x264 encoding, yet that guy claims he's doing 2500? I'd like to see that, I really would, but it seems bs. e: downloaded the vid, it's a 1080p 4Mbps video, 4Mbps live streamed with a modern card or even x264 will look like absolute balls, so that video is definitely captured to disk and then reencoded, or maybe just uploaded to youtube as is and let it do the processing. Truga fucked around with this message at 21:47 on Mar 5, 2017 |
# ? Mar 5, 2017 21:34 |
|
MaxxBot posted:I was curious so I took a look at some historical Steam HW survey data. I'd put this down to Intel pushing dual cores in laptops. The dual core i7s are great marketing, using the desktop performance brand to push their mobility skus.
|
# ? Mar 5, 2017 23:25 |
|
How does them pushing dual core affect quad core numbers?
|
# ? Mar 5, 2017 23:27 |
|
Two hyperthreaded cores appearing as 4?
|
# ? Mar 5, 2017 23:31 |
'I need a new laptop. *buys dual core U or Y to replace higher power 4 core*' or 'I need a laptop. *buys recent dual core laptop due to cost* Oh, whats this Counterstrike/DOTA/Bindings of Isaac/etc? *makes steam account*'
|
|
# ? Mar 5, 2017 23:34 |
|
eames posted:PCGH has some interesting graphs on 720p core scaling, too bad they didn't add min fps. Quadcores aren't looking so hot.
|
# ? Mar 5, 2017 23:34 |
|
Aesculus posted:Two hyperthreaded cores appearing as 4? Ah, I thought they were measuring C not T, but that makes sense.
|
# ? Mar 5, 2017 23:37 |
|
Aesculus posted:Two hyperthreaded cores appearing as 4? You'd think the steam hardware survey just identifies the processor model and already knows the difference. I mean, it knows I have a GTX 1080, not a "1500 whatever shaders device." Surely it just asks the operating system "What is your CPU model?" Otherwise we'd be seeing a lot of "8-core" i7's, right?
|
# ? Mar 5, 2017 23:43 |
|
Subjunctive posted:How does them pushing dual core affect quad core numbers?
|
# ? Mar 5, 2017 23:51 |
Combat Pretzel posted:Err, what does the X axis indicate? Y says framerate, X is what? Time? At first I was thinking # of threads at once considering it peaks at 16 and worsens quickly after that, but the colors have enabled cores or something, so I'd guess that if it were the case, they'd all be at different #'s
|
|
# ? Mar 5, 2017 23:54 |
|
Col.Kiwi posted:I think the point was most people only buy one system at a time, if they buy a dual core they're not buying a quad core because they're choosing one or the other. Oh, hmm. So the dual core numbers are growing instead? I should dig out the survey but it's so painful on my phone.
|
# ? Mar 5, 2017 23:54 |
|
Subjunctive posted:Oh, hmm. So the dual core numbers are growing instead? I should dig out the survey but it's so painful on my phone. edit: punctuation Col.Kiwi fucked around with this message at 00:06 on Mar 6, 2017 |
# ? Mar 6, 2017 00:02 |
|
Subjunctive posted:Ah, I thought they were measuring C not T, but that makes sense.
|
# ? Mar 6, 2017 00:05 |
|
Laptops are a very big segment of shipments. Quad core would've grown faster if they pushed quadcores on mobile is what I'm getting at. Just look at the gpu skewing to Intel instead of discrete. We've tapered off at near 50% quadcore.
|
# ? Mar 6, 2017 00:42 |
|
Laptops make up most of new PC shipments but probably not Steam users, based on 16% of Intel GPUs, maybe 20-25%? Most of which would be dual core, so I'd guess that like 80% of desktops are quad core (the survey refers to physical cores, not HT). Anyway, would be interesting to see if AMD gets over 20% share , and helps 8 cores reach a full percent.
|
# ? Mar 6, 2017 01:13 |
|
The steam hardware survey is meaningless at this point when it comes to westerners because it's probably >80% Russian/Chinese/Brazilian DOTA 2 players with computers from 2005.
|
# ? Mar 6, 2017 02:15 |
Hahah Anandtech's review has zero gaming benchmarks? They must have given them a pallet of cash to push that out a few weeks to a part 2.
|
|
# ? Mar 6, 2017 02:57 |
|
Pryor on Fire posted:Hahah Anandtech's review has zero gaming benchmarks? They must have given them a pallet of cash to push that out a few weeks to a part 2.
|
# ? Mar 6, 2017 03:29 |
|
K8.0 posted:The steam hardware survey is meaningless at this point when it comes to westerners because it's probably >80% Russian/Chinese/Brazilian DOTA 2 players with computers from 2005. I've noticed they're still heavily slanted towards old GPUs as well. Not that if you bought the high end in 2011-12 you'd need to run out and get the best because they'll still play medium-high for new titles, but the hardware survey is still loaded with Fermi and VLIW cards. They are also very heavily slanted by mobile results.
|
# ? Mar 6, 2017 03:34 |
I'm sorry but if you don't think there was a reason behind the premier cpu review site on the internet deciding to exclude all gaming benchmarks at launch for the first time ever then you're just wrong. This is insane. It's loving Anandtech. "Deep dive" indeed.
|
|
# ? Mar 6, 2017 03:46 |
|
Pryor on Fire posted:I'm sorry but if you don't think there was a reason behind the premier cpu review site on the internet deciding to exclude all gaming benchmarks at launch for the first time ever then you're just wrong. This is insane. It's loving Anandtech. "Deep dive" indeed. You mean like the fact that they did almost the entirety of their testing while attending press events (like actually in the hotel) and therefore just ran the scripted CPU tests? He also put up a poll on twitter and only 26% of ~200 said that gaming results would be more interesting.
|
# ? Mar 6, 2017 03:59 |
Oh ok sounds like Ian is on the up and up sorry my bad I'm sure that interview came with no strings attached my bad bro
|
|
# ? Mar 6, 2017 04:02 |
|
Pryor on Fire posted:Oh ok sounds like Ian is on the up and up sorry my bad I'm sure that interview came with no strings attached my bad bro Apparently one other factor is he may be redoing all gaming tests on Win10 since they haven't built up much of a dataset (i.e. nothing) on that yet. Even the 7700K tests were run on Win7
|
# ? Mar 6, 2017 04:05 |
|
K8.0 posted:The steam hardware survey is meaningless at this point when it comes to westerners because it's probably >80% Russian/Chinese/Brazilian DOTA 2 players with computers from 2005. That doesn't make any sense if you look at the actual results they get. If we look at a system built out of the most common attribute for each parameter, we get the most common Steam device as: nVidia graphics (59.8%) DirectX 12 supported (74.89%) Intel CPU (78.15%) 2.3 Ghz to 2.69 Ghz CPU (20.41%) 4 real cores (47.74%) Windows 10 64 bit OS (47.71%) 8 GB RAM (34.22%) 1 GB VRAM (33.35%) 1920x1080 primary display (43.23%) Hard drive space above 1 TB (30.44%) OS Language English (43.67%) SSE4.1 support on CPU (88.21%) (SSE4.1 is only available in systems from 2007 or later in Intel-land (starting with Penryn Core 2 Duo) and Bulldozer or later in AMD-land so 2011 and later) That last part basically means that less than 12% of systems surveyed by Steam are older than 10 years, and since there were still computers getting sold or built new with new CPUs without SSE4.1 up til like 2009 or so in Intel and until a few years ago in AMD (because the Phenom chips were often still better than Bulldozer), it'd be even less of the non-supporting set that are really old.
|
# ? Mar 6, 2017 04:43 |
|
I don't know if I've been tumbling into the rabbit whole or not, but based on a lot of rumors, some future path is coalescing for AMD New WSA agreement gives AMD (supposedly) more flexibility in choosing foundry Raven Ridge has two separate dies, one targeting mobile and one targeting desktop/server AMD is delaying the launch of R5 until Q2 and R3 until H2 of 2017 (usually Q3) 7nmFF isn't planned for CPU's until late 2018, long time with zero new product Discussion on Anandtech with the Stilt leads to the possibility 14nmLPP is utterly frequency tapped out, and as long as they remain on that process they'll only ever get better efficiency but never better performance - this is okay for servers, maybe even mobile, but for desktop it's not good. This leaves 16nmFF+ and apparently 14nmLPU for AMD before 7nm and thus late 2018 (lets be real, mid 2019). Again, Stilt is of the opinion that since 14nmLPU is unproven, AMD will move to 16nmFF+ and that the cost of paying GloFo and porting the design is worth it. Calling back to before, this means GloFo will continue to manufacture Zen in 14nmLPP for Server and mobile as Zen Gen1 and Raven Ridge mobile. The second Raven Ridge chip for desktop and Zen+ will be made at TSMC on 16nmFF+ or the rumored optimized 12nmFF+. It's conspiracy theory level, but the the early release with highest margins possible is designed to get the cash flow to move designs (CPU, potentially GPU) to TSMC. The idea is to eat the early "bad" reviews for the better product in late 2017, early 2018. I have no idea why I am trying to make sense of this launch, it just feels so Phenom I.
|
# ? Mar 6, 2017 05:07 |
|
This isn't as bad as Phenom I because Ryzen is actually a functional chip The errata as of current are just slightly worse than the level of Conroe's teething pains, ie why Intel now has chipsets on lockdown
|
# ? Mar 6, 2017 05:59 |
|
Pryor on Fire posted:I'm sorry but if you don't think there was a reason behind the premier cpu review site on the internet deciding to exclude all gaming benchmarks at launch for the first time ever then you're just wrong. This is insane. It's loving Anandtech. "Deep dive" indeed.
|
# ? Mar 6, 2017 08:17 |
|
....Wait, wasn't that the _entire point_ of Anand selling Anandtech off?
|
# ? Mar 6, 2017 08:38 |
|
Also on Anandtech forums, something interesting was pointed out by The Stiltquote:Regarding the ST: 1800X at default, with turbo & XFR enabled scores 162 in Cinebench 15. With the TDP (PPT) limited to 30W the score is 155. Ryzen is not a high frequency design and basically gains nothing with it's clocks pushed way up. Yes it does eek out more performance, but from what we know, 3.0Ghz across all cores is 65W, something like ~2.8Ghz is, based on the TheStilts charts, the ~30W range and 3.3Ghz is the last step before it needs before voltage requirements shoot to the moon. My impression is that Ryzen can completely trounce Intel in performance per watt in mobile and server, but it really does need a high performance cousin for workstation, HPC and desktop. I know I was a bit woe betide about Zens server chances but the Stilt makes it sound like Ryzen can beat even a massive price drop for Intel's current offerings. Over/under AMD will be selling 16/24/32C Naples for 1200/1800/2400$.
|
# ? Mar 6, 2017 09:08 |
|
Interesting Ryzen talk to run in the background with a guy that actually knows what he's talking about. I linked to the ECC discussion which has some new info — the systems support ECC RAM and correct single bit errors but the correction reporting and two bit system halt features seem to be missing. https://www.youtube.com/watch?v=bIbwuLdHbMg&t=1896s
|
# ? Mar 6, 2017 09:29 |
|
Actually speaking of Wendell https://twitter.com/tekwendell/status/838617988920184832 http://valid.x86.fr/u5xhc7
|
# ? Mar 6, 2017 09:40 |
|
eames posted:Interesting Ryzen talk to run in the background with a guy that actually knows what he's talking about. A novel solution proposed by Wendell: Have the motherboard manufacturers present the two core complexes as two separate chips to the OS, would avoid the thread switching and cache issues that are presently being seen *without* the need for a scheduler patch down the line.
|
# ? Mar 6, 2017 10:06 |
|
SwissArmyDruid posted:A novel solution proposed by Wendell: Have the motherboard manufacturers present the two core complexes as two separate chips to the OS, would avoid the thread switching and cache issues that are presently being seen *without* the need for a scheduler patch down the line. Isn't there a disadvantage to doing this though?
|
# ? Mar 6, 2017 10:21 |
|
Does Linux have these thread scheduling problems? I want to see some Linux gaming benchmarks. I want someone to establish the high-end processor of choice for Linux gaming.
|
# ? Mar 6, 2017 10:22 |
|
|
# ? May 22, 2024 04:49 |
|
FaustianQ posted:My impression is that Ryzen can completely trounce Intel in performance per watt in mobile and server FaustianQ posted:Isn't there a disadvantage to doing this though? On the chip design, Intel has separated cache banks that are transparently unified by the time whatever reaches the OS. There's no reason why AMD couldn't do this, at least for Z+. Anime Schoolgirl fucked around with this message at 11:05 on Mar 6, 2017 |
# ? Mar 6, 2017 11:00 |