|
Maybe that's the reason for the staggered releases, the oddly separated dates may mean new steppings and any actual bins currently would exacerbate the current problems.
|
# ¿ Feb 25, 2017 09:06 |
|
|
# ¿ May 21, 2024 23:28 |
|
memory latency is the important thing, and there's progressively less of it with higher speeds of RAM since CL isn't fully proportional to clocks.
|
# ¿ Feb 26, 2017 16:21 |
|
The scuttlebutt of low bandwidth RAM is that it's much harder to find low latency 2400 than it is to find average latency 3200
|
# ¿ Feb 26, 2017 17:10 |
|
this was a few days ago, but this is the important part http://semiaccurate.com/2017/02/22/amds-ryzen-7-1800x-beats-intels-i7-6900k-half-price/ quote:Charlie Demerjian the wccftech article is fairly sensational but I would not be surprised about intel being petty about AMD merely catching up to their last-full-gen poo poo
|
# ¿ Feb 27, 2017 01:02 |
|
Combat Pretzel posted:What's low latency in this case?
|
# ¿ Feb 27, 2017 04:14 |
|
the errata seems to be significant enough apparently
|
# ¿ Mar 1, 2017 01:08 |
|
Can they just fire AMD's marketing and press relations department like 10 years ago, holy poo poo this is an embarrassing fractal of stupidity. Errata wasn't such a big loving deal a decade ago for Core 2 and POWER8 took a giant poo poo for a few weeks upon its unveiling but finally got things working right after a couple months.
|
# ¿ Mar 2, 2017 02:37 |
|
PC LOAD LETTER posted:Yup. With Ryzen they've gotten close enough to Intel's latest core performance to not matter for most things while still selling for much less. That is the best you could realistically hope for with Ryzen.
|
# ¿ Mar 2, 2017 11:42 |
|
Don Lapre posted:b b b but i thought it was gonna make intel have to basically give the 7700k away because it was going to be such a gaming monster SMT not working properly at all is a surprise production chipsets may be behind and BIOS with working SMT awaiting validation if this is theoretically a fixable thing, otherwise we're going to have early A64/Conroe errata comedy for days if not weeks.
|
# ¿ Mar 2, 2017 15:58 |
|
eames posted:Pretty much. Zen predominately seems to be a high IPC/low frequency design and it seems like they relied on the process to get the frequency up. Because relying on GloFo has worked so well in the past.
|
# ¿ Mar 2, 2017 16:08 |
|
redeyes posted:I am most disappointed with their 'TDP' which is bullshit. Figured as much but AMD just plain lied about it from where I sit.
|
# ¿ Mar 2, 2017 16:24 |
|
Twerk from Home posted:It's fine, and "barely keep up with Haswell" is a decent result given that Broadwell was a teeny tiny improvement over Haswell. The truth is somewhere in the middle, it's better at some things and worse at others than Intel's chips. It's competitively priced for things that want 6+ slower cores and a decent product overall. Also, why the unholy gently caress is SMT broken on release?
|
# ¿ Mar 3, 2017 00:48 |
|
sauer kraut posted:I forgot to add for 149$ sry. you'll have to play with memory timings and frequency though, but that's always been a thing for BCLK
|
# ¿ Mar 3, 2017 01:05 |
|
PerrineClostermann posted:Didn't they patch out BCLK overclocking? You'd need an old board.
|
# ¿ Mar 3, 2017 01:26 |
|
eames posted:Nice launch.
|
# ¿ Mar 4, 2017 02:26 |
|
Alereon posted:So to clarify, you're trying to use software to compress the video for the game you're playing in real-time? Yeah that'll be hard on a quad-core CPU but it seems like there's better solutions, like hardware encoding. Also, QuickSync duplicates your display when you encode with that, which is nice if you're playing at 1080p60fps, not so much when you're playing anything with a higher refresh rate or resolution
|
# ¿ Mar 4, 2017 02:37 |
|
Intel cores have always been arranged as 1/H1/2/H2/3/H3.... and so on in Windows, as Ryzen is. I'm pretty sure that's just baseless theorycrafting when the real problem is that SMT just doesn't work properly in the first place. In unix-based systems, intel cores are arranged as 1/2/3/4/H1/H2/H3/H4. I can see how the problems arise if Ryzen doesn't follow this convention. Anime Schoolgirl fucked around with this message at 12:56 on Mar 4, 2017 |
# ¿ Mar 4, 2017 12:47 |
|
The AMD fanboy hasn't been the same since irony died a quick and uneventful death
|
# ¿ Mar 5, 2017 13:10 |
|
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 |
|
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 |
|
Ragingsheep posted:What's the difference between a mobile and desktop process? The mobile can't take as much power? Pascal (16nm TSMC) = Lower density design, better for achieving insane clocks and heat dissipation, low chip density and higher power consumption Sky/Kaby/Coffee Lake (14nm Intel) = Tries to be somewhere in between the above two Intel could actually make the high speed Lake CPU designs denser if they wanted to, but they want to sell 4Ghz Xeons someday and they have to have 3.2ghz-for-one-second "4.5w" processors as a bullet point for sales.
|
# ¿ Mar 8, 2017 02:11 |
|
GRINDCORE MEGGIDO posted:Wondering about the 4 core Ryzen, as they are cutting down each core complex, isn't this going to make the thread hopping problem significantly worse?
|
# ¿ Mar 17, 2017 01:38 |
|
They didn't decide to name it Z399 for trolling purposes?
|
# ¿ Mar 17, 2017 09:15 |
|
repiv posted:Update on the Ashes benchmark discrepancy: it looks like the game silently scales the detail level based on the number of cores available. And Oxide didn't think to tell anyone this or make it configurable
|
# ¿ Mar 31, 2017 21:01 |
|
SwissArmyDruid posted:In case you thought that we *weren't* headed towards our chips being assembled piecemeal going forward:
|
# ¿ Apr 4, 2017 16:26 |
|
FuturePastNow posted:Have we found out what they're using that THREADRIPPER trademark for yet?
|
# ¿ Apr 5, 2017 19:55 |
|
PC LOAD LETTER posted: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 17:01 |
|
Or maybe it's just a chip design that doesn't clock well
|
# ¿ Apr 9, 2017 17:09 |
|
EdEddnEddy posted:Really look forward to what the Mobile market might have when it gets to them. especially with the undervolted ryzen 1800x at 3ghz running at 30-35w stable
|
# ¿ Apr 11, 2017 23:13 |
|
|
# ¿ May 21, 2024 23:28 |
|
SwissArmyDruid posted:Toshiba tried to get into nuclear energy and failed. The Westinghouse subsidiary filed for bankruptcy protection, but it has severely injured the parent company itself. Their DRAM and NAND businesses are likely to be sold off to its competitors. Combined with the usual demand outstripping supply, you can expect it to get even worse in the fall.
|
# ¿ Apr 14, 2017 19:49 |