Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us $3,400 per month for bandwidth bills alone, and since we don't believe in shoving popup ads to our registered users, we try to make the money back through forum registrations.
«15 »
  • Post
  • Reply
SwissArmyDruid
Feb 14, 2014



Your thread counts are wrong for the R7 parts, Sinestro.

e:fb

Adbot
ADBOT LOVES YOU

SwissArmyDruid
Feb 14, 2014



Let's christen the new thread's first page with some ultimately pointless news:

Remember Der8auer? The German delid guy? He did a 1600X to 5.9 GHz on LN2, DDR4-3000 CL12. All cores enabled.

https://www.youtube.com/watch?v=CZ0SxpGzbw0

All hexacore Intel records beaten: https://ocaholic.ch/modules/news/ar...p?storyid=16401

SwissArmyDruid fucked around with this message at Apr 15, 2017 around 03:26

SwissArmyDruid
Feb 14, 2014




Former founder and editor-in-chief of The Tech Report. He's the one that helped get AMD and Nvidia to start thinking about their products in terms of frame time as opposed to raw FPS numbers in the first place.

SwissArmyDruid
Feb 14, 2014



Not using the X300 chipset, huh? Huh. I guess X300 is closer to embedded than ITX.

Hm. Supports "AMD A-series APU / Ryzen CPU / NPU for Socket AM4", huh? I guess NPU would be.... Neutered Processing Unit... so, APUs with the graphics disabled?

SwissArmyDruid fucked around with this message at Apr 20, 2017 around 04:00

SwissArmyDruid
Feb 14, 2014



Oh, speaking of paste, remember that you may need to adjust your paste technique a bit, with how the die is.



Rice grain isn't going to cover the entire thing. One dot over each die edit: CCU instead, maybe?

SwissArmyDruid fucked around with this message at Apr 21, 2017 around 21:12

SwissArmyDruid
Feb 14, 2014



kirtar posted:

Do we know the orientation of that? I'm assuming it's aligned with where we see gaps in the pins on the underside

Corner alignment triangle, dude.

gwrtheyrn posted:

I actually did that with better results as far as getting thin lines for the gpu, and might just do that for the cpu when i redo the paste because gently caress it

GPUs are bare dies, they need it because they don't have their own heat spreader like a CPU.

SwissArmyDruid fucked around with this message at Apr 21, 2017 around 11:11

SwissArmyDruid
Feb 14, 2014



....You mean the hard bit of compound that's the first thing to come out of the tube when you open it up after having not used it in a long-rear end time? Yeah, that's probably causing half your problems alone, if it was particularly dry and crusty.

SwissArmyDruid
Feb 14, 2014



VealCutlet posted:

Correct, but I'm maxing at 63.3 degrees at 1.35v

My bad, I got you mixed up with gwrtheyrn.

SwissArmyDruid
Feb 14, 2014



Watermelon Daiquiri posted:

Wait, ryzen's 4/4 structure is not internal to the chip, but rather from 2 4 core dies? No loving wonder

No, that was a poor use of words on my choice. I meant to convey that a dot over each of the two smaller rectangles might be better than one larger dot in the center. I absolutely should not have used to word "die" in my post and apologize, and have fixed the post.

SwissArmyDruid
Feb 14, 2014



Passmark says AMD took 2% of market share away from Intel last quarter.

More significant when you consider that Ryzen was only on sale for the last month of Q1, which was further compounded by motherboard shortages.

https://www.cpubenchmark.net/market_share.html

Perf stepping for Ryzen when, AMD?

SwissArmyDruid fucked around with this message at May 1, 2017 around 22:13

SwissArmyDruid
Feb 14, 2014



Rhetorical: ECC is the most natural thing in the world: It's the removal of a variable that can't be controlled. In the context of a home server, why *wouldn't* you want ECC?

SwissArmyDruid
Feb 14, 2014



Obsurveyor posted:

ECC required for ZFS smells like bullshit and cargo cult behavior to me.

The key phrase here is "any more so than any other filesystem", meaning "ECC should not be a factor in deciding between a filesystem and ZFS". Because if it meant something to you, you'd be using ECC, whether you were on ZFS, Btrfs, EXT4, or whatever.

e;fb

SwissArmyDruid
Feb 14, 2014



Why *not* ECC in every device? It *should* be a security thing, but alas, speed is king, and cutting costs is the advisor manipulating the king like a puppet.

http://www.techspot.com/article/845...-vs-ecc-memory/

(I don't know of any similar tests with DDR4.)

SwissArmyDruid
Feb 14, 2014



Padding the dividends of shareholders and the board of directors! </bitter>

SwissArmyDruid
Feb 14, 2014



Mmm-mmm-mmm-mmm-mmm. Christmas cheapo desktops/laptops are going to be _interesting_~

SwissArmyDruid
Feb 14, 2014



My vote's for AMD. They are like a dumb puppy. Overeager and even sometimes occasionally useful.

Nvidia's wants to push SaaS onto you, and SaaS is a legitimate goddamn cancer.

SwissArmyDruid fucked around with this message at May 11, 2017 around 07:24

SwissArmyDruid
Feb 14, 2014



Ehhhh, I *guess* so. Either way, the Club GeForce Elite thing they announced at the end of last year was absolutely no doubt some diseased marketing drone's baby.

SwissArmyDruid
Feb 14, 2014



Funny you should mention that. We finally have a Threadripper, and it's a codename for Ryzen 9 parts that correspond to Intel -E/-X parts.

Usual WCCFT caveats apply: Salt now so you're not salty later.

http://wccftech.com/amd-ryzen-9-lineup-threadripper/

Of note:

quote:

The Ryzen 9 lineup, code named Threadripper, will be compatible with a modified version of the company’s SP3 socket, code named SP3r2, which was originally designed for AMD’s 32 core Naples server parts. The new high-end desktop platform is code named “Whitehaven” and brings support for quad channel DDR4 memory support and 44 PCIe lanes to hardware enthusiasts, content creators and developers.

quote:

The Ryzen 9 1998X is allegedly the flagship of the new lineup, featuring 16 cores, 32 threads, a base clock speed of 3.5GHz and a boost clock speed of 3.9GHz with XFR ( Extended Frequency Range ). Amazingly, this CPU is only rated at 155W, 5 watts less than Intel’s upcoming Core i9-7920X 12 core flagship Skylake X CPU.
The Ryzen 9 1998 is the 1998X’s little brother, featuring 16 cores with slightly lower clock speeds of 3.2GHz base and 3.6GHz boost and the same 155W TDP.

I can't be the only one who immediately thought "K6!" as soon as I saw that 1998 number, right?

quote:

AMD’s new Ryzen 9 lineup and Whitehaven platform will reportedly be showcased at Computex (May 30th – June 2nd). According to sources in the upstream supply chain, the new high-end desktop offerings will launch alongside AMD’s 32 core Naples processors in June.

Did I say I was waiting for the May AGESA memory compatibility update before I bought in? Never mind, I have much better silicon to be looking at.

SwissArmyDruid fucked around with this message at May 15, 2017 around 11:18

SwissArmyDruid
Feb 14, 2014



wargames posted:

But the most important question goons have what about ecc?

In all seriousness, given that these are clearly going to be server parts that failed server binning, probably yes.

SwissArmyDruid
Feb 14, 2014



Munkeymon posted:

Hell yeah party like it's nineteen ninety, uh, eight again because I'm actually thinking about making an AMD system

E: too bad 3DFX is dead :\


What's this? If I'm going to go through with a Ryzen rebuild it sounds like getting good memory is gonna be Very Important.

This month's AGESA updates which they are pushing out to motherboard partners is adding up to 20 memory registers, in order to aid compatibility with XMP to allow kits with higher clocks to be used.

At last check, as the Infinity Fabric interconnect between cores appears to improve significantly with increased memory bandwidth, the higher frequencies of DDR4 are desirable. (Diminishing returns kicks in past 3000, but this should help more kits hit that number.)

SwissArmyDruid
Feb 14, 2014



No, although more kits should surely be making their way onto motherboard vendor's QVLs in the near future.

SwissArmyDruid
Feb 14, 2014



Since Poe's Law clearly states that without a clear indicator of the author's intent, it is impossible to create a parody of extreme views so obviously exaggerated that it cannot be mistaken by some readers or viewers as a sincere expression of the parodied views, I must therefore assume the worst:

AMD are systematically using quantum tunnelling to siphon off a small percentage of performance from every chip they have created and will create to power their time machine, wherein they have gone back and hired my pimply-faced fifteen-year-old self, or some other individual in that massive sea of teenaged gamers during the 90s and 00s that thought things named "EPYC" and "RYZEN" *were* cool because of the name, instead of the actual underlying silicon.

Or it could just be that those people grew up but never stopped thinking that poo poo was cool and oh my god please STOP.

I swear, if the next product has leet speak for a name, I am done, we're going to In-Vidia and never looking back.

SwissArmyDruid
Feb 14, 2014




That's one big fuckin' package. Quick, someone get one into der8auer's hands so he can delid one of those fuckers, see how big the actual die is.

SwissArmyDruid fucked around with this message at May 16, 2017 around 22:21

SwissArmyDruid
Feb 14, 2014



repiv posted:

My condolences to the Asrock engineer who gets tasked with designing a mini-ITX board around this socket.

Condolences? Fifteen years of Asrock boards, and you still haven't realized that Asrock engineers are all a little touched in the head, and they LIVE for this kind of poo poo?


Zen 2 and Zen 3 are in development, and it will be on 7nm.

http://fudzilla.com/news/processors...announced-zen-2

SwissArmyDruid fucked around with this message at May 16, 2017 around 22:21

SwissArmyDruid
Feb 14, 2014



WARNING: VENDOR-SUPPLIED BENCHMARK AHEAD.



Apparently that's a linux kernel compile bench, head-to-head, no word on the setups.

SwissArmyDruid
Feb 14, 2014



I felt a single, sharp pain in my groin when I saw the $4938 MSRP on that Xeon.

I have no intention of ever buying one of those, and I think Intel *still* just exacted the price of my hypothetical future firstborn from me just to look at the spec sheet.

SwissArmyDruid fucked around with this message at May 17, 2017 around 00:41

SwissArmyDruid
Feb 14, 2014



No, because that would be a *logical* price for it. Probably somewhere around $800-$900, because AMD is absolutely hell-bent on leaving money on the table.

SwissArmyDruid
Feb 14, 2014



Same ballpark? That's maybe a bit generous. Same league, different conferences, and it looks like both teams are going to be seeing eachother in the finals.

SwissArmyDruid
Feb 14, 2014



AFAIK, Samsung B-die is still the thing to get. which means G.Skill and some Corsair.

(Not all G.Skill are Samsung, though. Venn diagrams and all that, but there's a lot less of it in Corsair's product stack.)

But the new AGESA update might change that.

More details on that Linux kernel compile test, c/o TR:

quote:

To illustrate the point, AMD offered a demonstration of a 2S Epyc system going against two high-end Xeon E5-2699A v4 CPUs, each with 22 cores and 55 MB of cache. AMD's machine had 256 GB of RAM thanks to its eight channels, while the Intel-powered server was rolling with 128 GB. The Epyc box finished a Linux kernel compilation in 15.7 seconds, around 43% faster than the 22.5 seconds it took the Intel system.

So, it's the memory bandwidth that was giving it the extra oomph needed, right?

quote:



Those figures may raise a couple eyebrows on their own, but AMD proceeded to point out that not a whole lot of systems are shipped with the mighty Xeon E5-2699A v4 onboard. According to the company, servers powered by Intel's Xeon offerings between the ES-262x and ES-265x series make up the bulk of shipments, and among those, the ES-264x series is the best seller among them. AMD proceeded to offer another demonstration, then: a single-socket Epyc system with 128 GB of RAM going up against a 2S box with Xeon ES-2650 v4 CPUs (12 cores each). For a Linux kernel compile, the Epyc system took 33.7 seconds, while the Xeon box did its work in 37.2 seconds. While this is but a single benchmark, it's nonetheless impressive and lends some credence to AMD's plan of going after two-socket systems with single-socket Epyc offerings.

Not the same top-flight Xeon as before. Even if they DID have a logical justification for running the test, they probably did their benchmark and just kept downgrading the Intel chip until they got one where the one Epyc chip beat two Xeons, and then back-justified it. =P

Either way, compelling, and cause for optimism. http://techreport.com/news/31927/am...powered-servers

SwissArmyDruid fucked around with this message at May 17, 2017 around 18:30

SwissArmyDruid
Feb 14, 2014



Sorry, there are actually legitimate mainstream reasons for building an AMD rig now.

SwissArmyDruid fucked around with this message at May 20, 2017 around 21:12

SwissArmyDruid
Feb 14, 2014



Fauxtool posted:

My understanding it that the card that most closely matches the average performance of both companies current line up is the mid tier.

Just about everything from AMD is low to mid and nvidia is generally speaking low to high.

I would call the rx580 4gb and 1060 3gb mid tier

I would call that "bare minimum for a dGPU" tier, and strongly encourage the buyer to do whatever is necessary for a few extra bucks to get a 480/580 8GB or 1060 6 GB.

SwissArmyDruid
Feb 14, 2014



I actually believe that 1080p60 should be the absolute minimum, and that's where I base my scaling from. Consoles have been trying to promise 1080p for god knows how long now, and they're still continuing to fail at it. They need to get with the times, they're holding everyone back.

SwissArmyDruid
Feb 14, 2014



Fauxtool posted:

its probably not going to happen, instead they will fail at 4k30fps while being able to do 1080p60fps but not offering it anymore.

The sad part is, you're probably right. The only people who are going to do 1080p60 are going to be the character action devs, i.e, Platinum.

SwissArmyDruid
Feb 14, 2014



I don't even understand this half-step generation bullshit.

There was literally no reason to do any of it, the stupid console makers heard "VR" and their goddamn genitals overrode their good sense and decided they needed to half-step so they could force VR where it wasn't a good fit.

The REAL thing to do would be to clean sheet both consoles and start with a Raven Ridge part, assuming that Raja hasn't been talking out his rear end for the past year+, and whatever GPU silicon they're working on now can actually hold up their end of the bargain.

But I'm sure I'm preaching to the choir.

SwissArmyDruid fucked around with this message at May 21, 2017 around 09:09

SwissArmyDruid
Feb 14, 2014



Combat Pretzel posted:

Interesting. If the frequencies shown there is what the CPU ran at, instead of nominal while they were actually turboing, bumping the Whitehaven to 4GHz would mean near perfect linear scaling. Which seems doubtful, unless they've improved the interconnect speeds. Which means that the 8C's handicap makes the difference for the scaling to appear linear.

So anyway, are there any significant teething issues with Ryzen, that haven't been fixed yet with BIOS updates?

One mainstream problem, and one fringe problem:

The mainstream: Memory timings continue to be an issue. Because of how the chip architecture is, Ryzen responds very well to higher-clocked RAM. However, as time as gone on, AMD has released firmware updates to the motherboard manufacturers that have improved them over time. There is a major AGESA update expected by the end of this month that will greatly improve memory compatibility. For best results, though, you are generally still looking at memory that uses Samsung B-die, and only two of the largest sticks that you can find.

The fringe: IOMMU groupings are not set up correctly for GPU passthrough to VMs. This is correctable with an ACS patch for Linux. Here's hoping it's correctable by software or firmware and doesn't require a hardware fix. If none of that made any sense to you, don't worry about it. =P

But this may be how AMD does their market differentiation between the mainstream and HEDT parts, we just don't know yet. But having that 1800X 8C/16T part got people excited about having a machine well-suited to this application, even if the actual number of available PCIe lanes was just barely enough to allow this. I, for one, am hoping to move to Linux full-time, with a Windows VM running on top of it to service those last few things that need Windows and I don't have a choice for.

SwissArmyDruid
Feb 14, 2014



Epic suck.

Well, AMD are under super-hyper-mega-ultra-crunch time to get ready for Computex.

On a scale of increasing amounts of suckitude, here's hoping:

* Their priorities are being consumed by having to get the new poo poo ready for when Computex begins on the 30th of May
* they can figure out what's wrong
* they don't have to hardware rev to fix things
* that these issues stem from how AMD intends to segment between mainstream and HEDT.

SwissArmyDruid
Feb 14, 2014



If it ain't Raven Ridge, I literally do not loving care.

Furthermore, Dell seems to have hit the price/performance apex with the Dell 7000 Gaming this year. For $1000 USD, I get a Kaby Lake i5 + 1050Ti + 1080p IPS display and that will last until the next console hardware generation.

Whatever Asus is showing, it needs to be at least that performance level on a APU, and it needs to be interesting, or cheaper, or whatever.

SwissArmyDruid fucked around with this message at May 22, 2017 around 22:49

SwissArmyDruid
Feb 14, 2014



Currently down on Asrock. I'm still loving waiting on that loving Z270M-STX board with the MXM slot they showed off in Taipei at the beginning of this year.

loving let me use this loving M2000M I've got, you jackholes!

SwissArmyDruid
Feb 14, 2014



Been thinking about what I said, and I think that in the end, i5 + 1050 ti in a Raven Ridge APU is way too high a bar. Maybe with Zen2 + Navi, but right now, no.

I still maintain that the Dell 7000 Gaming series is _the_ place to be as far as a gaming laptop is concerned, with that perfect intersection of size, power, screen, and battery life, where if you bumped any one of the component parts any larger, the rest of the system would hold it back? And I do not think that AMD can hit that with a single chip.

The best that AMD can do with Raven Ridge, I think, is if you were to plot the above system on a radar chart, and then an AMD APU-based system plotted on top of it, the APU-based system should occupy the same profile, but proportionally smaller, and commensurately cheaper, down around the $750-$800 retail mark.

SwissArmyDruid fucked around with this message at May 24, 2017 around 01:16

Adbot
ADBOT LOVES YOU

SwissArmyDruid
Feb 14, 2014



Infinity Fabric != Interposer. Infinity Fabric is, in effect HyperTransport 4.0, and a technology that AMD acquired when they purchased Seamicro a few years back.

Interposers are a physical interface that connects stacked components together.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply
«15 »