|
Raven Ridge Zotac mini pc: https://www.anandtech.com/show/12319/zotac-displays-zbox-ma551-an-amd-raven-ridgebased-sff-pc Could be an interesting media center box/home theater. Likely worth waiting to see how it handles GPU video transcode and stuff though (and how loud that fan is)
|
# ? Jan 17, 2018 00:31 |
|
|
# ? Apr 24, 2024 01:34 |
|
That could be lots of fun! The cooling system looks rather robust for the form factor, too. Won't be quiet if you get to overclocking, but I have hope.
|
# ? Jan 17, 2018 05:16 |
|
AMD Retpoline benchmarks. https://www.phoronix.com/scan.php?page=news_item&px=AMD-Retpoline-Linux-4.15-FX-Zen Pre-Ryzen poo poo is either minimally affected or *gains performance*, Ryzen either minimally affected or *gains performance*, Threadripper takes a serious hit on random reads in two of seven tests?! EPYC largely unaffected either way, looks like margin of error difference. Also, I need to offer a mea culpa to whichever goon it was that argued that J. Average Luser WOULD be impacted by Meltdown/Spectre. Engadget is reporting that the rebooting problems affect Skylake/Kaby, too. https://www.engadget.com/2018/01/18/intel-spectre-reboot-problem-affects-newer-cpu/ SwissArmyDruid fucked around with this message at 15:36 on Jan 18, 2018 |
# ? Jan 18, 2018 15:34 |
|
quote:J. Average Luser WOULD be impacted by Meltdown/Spectre NVMe Random 4k goes down around half from everything I've read. This is not trivial at all.
|
# ? Jan 18, 2018 15:39 |
|
Seems weird that TR is affected but not EPYC.
|
# ? Jan 18, 2018 15:53 |
|
R5 2600 spotted posting 3.6-3.8GHz vs the current 1600's 3.2-3.4. Hopefully production samples will be able to break 4GHz on boost. E: WCCFTech salt now, salty later, etc https://www.google.com/amp/s/wccfte...hero-x470-leak/
|
# ? Jan 18, 2018 15:55 |
|
NewFatMike posted:R5 2600 spotted posting 3.6-3.8GHz vs the current 1600's 3.2-3.4. Hopefully production samples will be able to break 4GHz on boost. 200MHz more at the same TDP isn't exactly getting me excited, but I guess it's a start
|
# ? Jan 18, 2018 16:00 |
|
redeyes posted:NVMe Random 4k goes down around half from everything I've read. This is not trivial at all. but but but stack crashes in rare situations on some ryzen gen1 chips
|
# ? Jan 18, 2018 16:12 |
|
Ars Technica article on Spectre/Meltdown - old AMD procs now patch properly in Windows, but pretty much every -Bridge/-Well/-Lake variant now suffers with sporadic reboot issues.
|
# ? Jan 18, 2018 17:36 |
|
I had to disable stuff like Dropbox and shut down other processes to keep my Lynnfield from rebooting.
|
# ? Jan 18, 2018 17:47 |
|
Any info about when the price drops will be live? Living in , we didn't really get awesome black Friday pricing and I'm currently sitting next to a still wrapped 1600x... I'm thinking return it, wait and pick up a 1700x, but that depends when the prices drop and when that trickles down to Europe.
|
# ? Jan 18, 2018 19:57 |
|
Is Zen+ a paper launch, or is there an date and event where we're expected to get more info?
|
# ? Jan 18, 2018 20:19 |
|
I don't think theres a HARD date but AMD has already stated they're sampling the process right now per the CES slide decks and the across the board discounts on current ryzen parts are a hint of clearing inventory.
|
# ? Jan 18, 2018 20:35 |
|
ufarn posted:Is Zen+ a paper launch, or is there an date and event where we're expected to get more info? March/April is all we've heard so far. Sounds like the conservative estimates were closer to the mark, so there may not be too much of a performance reason to wait really.
|
# ? Jan 18, 2018 20:41 |
|
HalloKitty posted:200MHz more at the same TDP isn't exactly getting me excited Interested to see the silicon lottery overclocking results. How long after launch do they usually release those? Wondering if the 2700 is going to be the best bet.
|
# ? Jan 18, 2018 21:55 |
|
GRINDCORE MEGGIDO posted:Interested to see the silicon lottery overclocking results. How long after launch do they usually release those? SiliconLottery dropped all their binned Ryzen SKUs once it came to light that AMD was skimming all the best dies for Threadripper, so they might not bother testing 2nd gen Ryzen at all
|
# ? Jan 18, 2018 21:59 |
|
AMD actually does frequency binning, which is why you have a pretty tight range of overclock expectations per SKU with relatively few outliers and out of those, they seem to be restricted to utter comedy like an 8 core R3 1300x
|
# ? Jan 18, 2018 22:10 |
|
Anime Schoolgirl posted:AMD actually does frequency binning, which is why you have a pretty tight range of overclock expectations per SKU with relatively few outliers and out of those, they seem to be restricted to utter comedy like an 8 core R3 1300x This. Non X will consistently hit 3.9Ghz, X SKUs seem to hit about 200Mhz higher for similar voltage. Not a lot, but it's there, and I'd also wager it's easier to hit 3200/3600Mhz on X SKUs compared to not. I'm starting to think Ryzen+ is going to be less about clocks and more about making the uncore less poo poo. They talk about memory latency as a focus, which sounds like it could be either IMC or IF related, I have no idea if they could legit make their L1/L2/L3 any faster since comparisons on launch seemed to be either similar or faster to Intel sans weird poo poo where CCX1 would try to read from CCX0's L3 because programs treat Ryzen as unified core design.
|
# ? Jan 18, 2018 23:18 |
|
Do Noctua still have the offer of the free parts needed to mount their fans on AM4 boards?
|
# ? Jan 19, 2018 01:24 |
|
HalloKitty posted:200MHz more at the same TDP isn't exactly getting me excited, but I guess it's a start Given how abruptly 14nm runs into a frequency wall, it's possible that on a performance-optimized process there may be more OC headroom than 14nm allowed. But yeah, reality check, it's a new stepping on a tweaked process, 10% would be good, 15% would be exceptional, and more than 20% would probably be unrealistic. 5% isn't impossibly low either. Uncore improvements and stability fixes are probably the focus here, and big IPC gains (other than from uncore/IMC) are probably going to wait until Zen2. On the other hand, AMD already has a leg up this time around, since they're less impacted by Spectre/Meltdown. Paul MaudDib fucked around with this message at 01:59 on Jan 19, 2018 |
# ? Jan 19, 2018 01:51 |
|
If its 10% then so be it. Even better with all the stability/errata/memory latency improvements you just mentioned. Small as that margin may be, it represents quite a bit of extra life and especially so if Zen2 is delayed by any significant amount. And speaking of Zen2, I'd rather drop it into a more up to date motherboard anyway.
|
# ? Jan 19, 2018 04:47 |
|
I was thinking of getting a i7 8700 non-k, but with the crazy ram prices and now this Meltdown crap I just might wait for Zen+. Hope it's worth it!
|
# ? Jan 19, 2018 11:39 |
ufarn posted:Do Noctua still have the offer of the free parts needed to mount their fans on AM4 boards? It wasn't a limited time thing: If you have a Noctua cooler and they make a mounting kit for it that you need and don't have, they'll give you one for free, always.
|
|
# ? Jan 19, 2018 16:08 |
|
Anime Schoolgirl posted:AMD actually does frequency binning, which is why you have a pretty tight range of overclock expectations per SKU with relatively few outliers and out of those, they seem to be restricted to utter comedy like an 8 core R3 1300x How does Intel do it if not by frequency?
|
# ? Jan 19, 2018 16:08 |
|
Aaaaaand there's the AMD Lawsuit over Spectre/Meltdown. https://www.theregister.co.uk/2018/01/17/amd_investors_sue_over_chip_flaw_silence/
|
# ? Jan 19, 2018 21:40 |
|
SwissArmyDruid posted:Aaaaaand there's the AMD Lawsuit over Spectre/Meltdown. https://www.theregister.co.uk/2018/01/17/amd_investors_sue_over_chip_flaw_silence/ That was a few days ago, and it's shareholders trying to sue AMD for inflated the price even though it's gone up during this the entire time. It didn't seem to do anything to the stock price when this was announced I doubt anything will come of it either. There's security reasons to not disclose early, they also will need to put a number on the losses which is going to be extremely hard when the price has gone up.
|
# ? Jan 19, 2018 21:43 |
|
Munkeymon posted:How does Intel do it if not by frequency? The K parts are the absolute worst failure state of their binning strategy and it's hilarious that it's at a markup. Anime Schoolgirl fucked around with this message at 02:20 on Jan 20, 2018 |
# ? Jan 20, 2018 02:18 |
|
I walked into my not-oft used computer lab to find 4 of the 11 AMD Athlon x2 4000 PCs hung at the windows 10 startup logo because of the spectre/meltdown Windows update. Thanks Intel/MS/Obama? I have yet to find a fix that doesn't require reinstalling, is anyone aware of any?
|
# ? Jan 20, 2018 04:37 |
|
OldPueblo posted:I walked into my not-oft used computer lab to find 4 of the 11 AMD Athlon x2 4000 PCs hung at the windows 10 startup logo because of the spectre/meltdown Windows update. Thanks Intel/MS/Obama? I have yet to find a fix that doesn't require reinstalling, is anyone aware of any? System restore.
|
# ? Jan 20, 2018 04:56 |
|
Anime Schoolgirl posted:Efficiency. Everything we get on desktop that's below LGA 2066 are leaky runt bins that can't be used for laptops. There's no frequency binning being done at all and the K parts are actually incredibly leaky (which turns out to be better OC the more powerful your cooling solution is), which is why when BCLK overclocking was a thing some people found that their i5-6400s clocked better than some 6700ks, which is exactly why Intel stopped it again. Huh! Thanks. That makes a lot of things about overclockability make sense that previously didn't.
|
# ? Jan 22, 2018 16:38 |
|
This is just a *little* bit too delicious. And again, a reason why a combined CPU thread would be bad. http://lkml.iu.edu/hypermail/linux/kernel/1801.2/04628.html
|
# ? Jan 22, 2018 17:45 |
|
SwissArmyDruid posted:This is just a *little* bit too delicious. And again, a reason why a combined CPU thread would be bad. code:
|
# ? Jan 22, 2018 17:48 |
|
One might say Linus is having a meltdown
|
# ? Jan 22, 2018 17:53 |
|
For the less-technically-inclined: https://news.ycombinator.com/item?id=16202882
|
# ? Jan 22, 2018 17:57 |
|
Thanks for shaping our industry Intel. Also how does one foster a populist uprising of hardware consumers????¿???¿??????????
|
# ? Jan 22, 2018 18:04 |
|
Find out if Windows accepted the patches without saying a word, and then blasting them on it, I suppose.
|
# ? Jan 22, 2018 18:07 |
|
Sidesaddle Cavalry posted:Thanks for shaping our industry Intel. Unfortunately, our predecessors were just fine with using proprietary bullshit, so there's no real competition right now. Start buying open source hardware and hope it catches up some day.
|
# ? Jan 22, 2018 18:10 |
|
I don't think there will be enough unrest to topple all this inertia until people actually start dying over the issue, goodness forbid. Alternatively, who wants to join me in starting a buttcoin based on the level of danger loosed on the public by developer neglect and/or collusion
|
# ? Jan 22, 2018 20:15 |
|
Sidesaddle Cavalry posted:I don't think there will be enough unrest to topple all this inertia until people actually start dying over the issue, goodness forbid. I'd rather see folding at home turned into some kind of buttcoin like currency so the mining would actually do something and people could trade them on the value that they did something good and there is a limited amount of research.
|
# ? Jan 22, 2018 22:24 |
|
|
# ? Apr 24, 2024 01:34 |
|
SwissArmyDruid posted:This is just a *little* bit too delicious. And again, a reason why a combined CPU thread would be bad. Not much of that is readable, but Linus has gotten just as mad at things like GNOME and SuSE before, so his being salty over something isn't exactly world shattering. The name of the exact file is lost in my hazy memory, but this reminds me of the version of MS-DOS that was so terrible that developers started writing applications that refused to run if the version of DOS was higher than the previous release, assuming that this was the new direction and MS-DOS would be flaming garbage forevermore. When Microsoft released the next major revision of DOS and addressed many of the complaints, they included a .SYS file or something that looked for programs that had this characteristic and lied about it's version. If you're trying to run enterprise Linux, something similar where modern era Intel CPUs are treated as untrustworthy will probably be in order. Most of the enhancements and efficiencies of the past decade would be lost if you disabled branch prediction on consumer CPUs, it's understandable why Intel wouldn't want it enabled by default in home/consumer environments. My only thing about the threads was that there's nothing stopping people from posting in other threads or both. The implication that there's any sort of a wall is purely an illusion.
|
# ? Jan 22, 2018 23:01 |