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 money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Locked thread
priznat
Jul 7, 2009

Let's get drunk and kiss each other all night.

SwissArmyDruid posted:

....But they _do_ have enormous negative value. You can't sound dampen that panel without obviating the window.

I've heard the tempered glass sides are pretty good at sound deadening.

They can look pretty slick too (especially Phanteks).

Adbot
ADBOT LOVES YOU

AVeryLargeRadish
Aug 19, 2011

I LITERALLY DON'T KNOW HOW TO NOT BE A WEIRD SEXUAL CREEP ABOUT PREPUBESCENT ANIME GIRLS, READ ALL ABOUT IT HERE!!!

SwissArmyDruid posted:

....But they _do_ have enormous negative value. You can't sound dampen that panel without obviating the window.

Incorrect, plastic windows have roughly the same sound dampening effects as various foams used for case sound dampening, at least according to the Fractal Design engineering and testing departments.

Deuce
Jun 18, 2004
Mile High Club

SwissArmyDruid posted:

....But they _do_ have enormous negative value. You can't sound dampen that panel without obviating the window.

Soundproof windows are totally a thing!

Paul MaudDib
May 3, 2006

TEAM NVIDIA:
FORUM POLICE

MaxxBot posted:

You're referring to the i7-5820k? What evidence do you have that a i7-5820k would "utterly destroy" the 1800X in single-threaded performance? Granted it was only one game but the AMD demo they had today showed the 1700 getting higher FPS than an 6800k at identical clocks. Your take on Zen is way more pessimistic than I have seen from anyone else.



It's still at least 25% higher IPC.

The 5820K is just as fast as the 6800K - just uses a little more power to do so. Actually since it overclocks so high it usually outperforms Broadwell-E despite slightly poorer IPC. There's a reason Intel decided to forget about Broadwell and press on with Skylake ASAP instead. In fact, IIRC they weren't even going to make socketed desktop processors until some enthusiasts got pissy, and they never did sell well.

What's the game? AMD undoubtedly picked something like AOTS or For Honor that scales super well with cores, there's no reason they wouldn't put their best foot forward. So that's almost certainly not looking at single-threaded performance, i.e. most games.

I'm just calling them like I see them here. If it ends up hitting 5 GHz then so much the better, but I think it's a valid interpretation here that Ryzen is pretty much clocked out from the factory, and this isn't my first time at the AMD launch rodeo, nor seeing the pre-order-before-the-NDA-lifts gimmick be played.

It's good that AMD is going to be competitive, but they are experts at doing this whole hype-train thing, and their fan-boys just want to believe so badly, and just eat that poo poo up. I firmly believe that a bunch of this stuff is going to get walked back a bit after launch - like when we start seeing how it plays in games that aren't cherry-picked thread-friendly games.

I'm sure we can all remember the pre-launch demos AMD gave of Fiji, where it was beating a 980 Ti or Titan X? And how when the reviews came in after launch it turned out not to be technically a lie, just they'd picked combinations of games and settings that nobody ever used that disadvantaged Maxwell and favored GCN? And how even 2 years later and even with AMD's vaunted FineWine™ technology the Fury X still underperforms the 980 Ti in a lot of games?

(Seriously though, the FineWine™ meme says everything you need to know about the AMD crowd. That's what happens when you don't upgrade your architecture for 4 years - you get driver improvements!)

Anime Schoolgirl posted:

he's exaggerating, but the i7-5820k is intel's best kept secret because people can't stop jerking each other off over mediocre kaby lake overclocks

it clocks 0.4-0.5ghz higher than the i7-6800k on average outside of obvious review samples and costs $50 less, which is why intel conveniently stopped making more of it

Actually I'm pretty sure it's still in production? Newegg and other stores have it, and it still shows in the Ark as "launched", rather than "end-of-life" or whatever.

Paul MaudDib fucked around with this message at 04:43 on Feb 23, 2017

SwissArmyDruid
Feb 14, 2014

by sebmojo

AVeryLargeRadish posted:

Incorrect, plastic windows have roughly the same sound dampening effects as various foams used for case sound dampening, at least according to the Fractal Design engineering and testing departments.

Would like to request unanimous consent to amend and extend previous comments.

Cannot _further_ soundproof the panel without obviating the intended purpose of the window.

Words are hard.

kirtar
Sep 11, 2011

Strum in a harmonizing quartet
I want to cause a revolution

What can I do? My savage
nature is beyond wild

Paul MaudDib posted:



It's still at least 25% higher IPC.

The 5820K is just as fast as the 6800K - just uses a little more power to do so. Actually since it overclocks so high it usually outperforms Broadwell-E despite slightly poorer IPC. There's a reason Intel decided to forget about Broadwell and press on with Skylake ASAP instead. In fact, IIRC they weren't even going to make socketed desktop processors until some enthusiasts got pissy, and they never did sell well.

What's the game? AMD undoubtedly picked something like AOTS or For Honor that scales super well with cores, there's no reason they wouldn't put their best foot forward. So that's almost certainly not looking at single-threaded performance, i.e. most games.

I'm just calling them like I see them here. If it ends up hitting 5 GHz then so much the better, but I think it's a valid interpretation here that Ryzen is pretty much clocked out from the factory, and this isn't my first time at the AMD launch rodeo, nor seeing the pre-order-before-the-NDA-lifts gimmick be played.

It's good that AMD is going to be competitive, but they are experts at doing this whole hype-train thing, and their fan-boys just want to believe so badly, and just eat that poo poo up. I firmly believe that a bunch of this stuff is going to get walked back a bit after launch - like when we start seeing how it plays in games that aren't cherry-picked thread-friendly games.

I'm sure we can all remember the pre-launch demos AMD gave of Fiji, where it was beating a 980 Ti or Titan X? And how when the reviews came in after launch it turned out not to be technically a lie, just they'd picked combinations of games and settings that nobody ever used that disadvantaged Maxwell and favored GCN? And how even 2 years later and even with AMD's vaunted FineWine™ technology the Fury X still underperforms the 980 Ti in a lot of games?

(Seriously though, the FineWine™ meme says everything you need to know about the AMD crowd. That's what happens when you don't upgrade your architecture for 4 years - you get driver improvements!)


Actually I'm pretty sure it's still in production? Newegg and other stores have it, and it still shows in the Ark as "launched", rather than "end-of-life" or whatever.

To me all that says is that literally dividing a 3DMark score by the number of cores tells you jack poo poo about individual core performance unless you 're comparing the same number of cores. If you take those numbers at face value, they tell you that the 6 core and 4 core variants have higher individual core performance than the 8-core despite being the same architecture and running at markedly lower clock speeds. The only straight comparison with similar clocks that doesn't involve handwaving the fact that multithreading isn't 100% efficient is to use the 1600X in this graph which indicates a ~10%4% difference in per core score between the 6800K and 1600X. There is also the quad core ryzen at 3.2 compared to kaby lake at over 4 GHz telling roughly similar story. That said, I've seen different values for the score of the 6800K floating around (e.g. the one from wccftech suggests lower score for the 6800K, but it's wccftech). In either case we'll find out when more rigorous reviews release.

kirtar fucked around with this message at 05:39 on Feb 23, 2017

Deuce
Jun 18, 2004
Mile High Club

SwissArmyDruid posted:

Would like to request unanimous consent to amend and extend previous comments.

Cannot _further_ soundproof the panel without obviating the intended purpose of the window.

Words are hard.

Would like to request unanimous consent to repeat previous comments that soundproofing on glass is totally a thing.

Oh god don't give Linus any ideas.

PerrineClostermann
Dec 15, 2012

by FactsAreUseless

Deuce
Jun 18, 2004
Mile High Club

People who don't care enough to upgrade from the stock cooler also probably don't care about RGB lighting.

Kazinsal
Dec 13, 2011


The last few days of rumours and leaks are making me want to buy a goddamn i7-7700K all of a sudden.

Fucks sakes AMD get it together.

Deuce
Jun 18, 2004
Mile High Club

Kazinsal posted:

The last few days of rumours and leaks are making me want to buy a goddamn i7-7700K all of a sudden.

Fucks sakes AMD get it together.

If you're just playing games it's still the top performer out there and it's also not $500. Ryzen isn't going to wipe out every Intel chip overnight, anyone who thought it was going to do that is a dumbshit fanboy.

But if you're recording/streaming while playing then Ryzen just dramatically dropped your cost of entry. Intel has been charging "lol gently caress you there's nobody at this level" prices on their >4 core chips for a while now.

Now, what might still be useful is seeing how Ryzen... 5....? chips pan out. If they're like Intel and achieve better clockspeeds on the 4-core chips, we'll probably see a 4c/8t chip that is at least in the ballpark of the 7700k for substantially less money.

yet another edit: although the leaked/rumor prices for the lower end Ryzens weren't as dramatic as the Ryzen 7 set, and the leak prices turned out pretty accurate for those.

Deuce fucked around with this message at 05:17 on Feb 23, 2017

MaxxBot
Oct 6, 2003

you could have clapped

you should have clapped!!

Paul MaudDib posted:



It's still at least 25% higher IPC.

I pointed out earlier in the thread that is an awful, awful chart. Those numbers are just the numbers from the multithreaded physics benchmark divided by the number of CPU cores. It makes absolutely no sense to present the data like that because the benchmark does not scale linearly with additional cores.


kirtar
Sep 11, 2011

Strum in a harmonizing quartet
I want to cause a revolution

What can I do? My savage
nature is beyond wild

MaxxBot posted:

I pointed out earlier in the thread that is an awful, awful chart. Those numbers are just the numbers from the multithreaded physics benchmark divided by the number of CPU cores. It makes absolutely no sense to present the data like that because the benchmark does not scale linearly with additional cores.




It also appears to not scale perfectly with frequency either.

BurritoJustice
Oct 9, 2012

It seems from checking launch motherboards that Ryzen is capping out at 3200MHz DDR4 (at least for the displayed max speed on motherboards), which is somewhat disappointing.

Risky Bisquick
Jan 18, 2008

PLEASE LET ME WRITE YOUR VICTIM IMPACT STATEMENT SO I CAN FURTHER DEMONSTRATE THE CALAMITY THAT IS OUR JUSTICE SYSTEM.



Buglord
Speculation

kirtar
Sep 11, 2011

Strum in a harmonizing quartet
I want to cause a revolution

What can I do? My savage
nature is beyond wild

Deuce posted:

If you're just playing games it's still the top performer out there and it's also not $500. Ryzen isn't going to wipe out every Intel chip overnight, anyone who thought it was going to do that is a dumbshit fanboy.

But if you're recording/streaming while playing then Ryzen just dramatically dropped your cost of entry. Intel has been charging "lol gently caress you there's nobody at this level" prices on their >4 core chips for a while now.

Now, what might still be useful is seeing how Ryzen... 5....? chips pan out. If they're like Intel and achieve better clockspeeds on the 4-core chips, we'll probably see a 4c/8t chip that is at least in the ballpark of the 7700k for substantially less money.

yet another edit: although the leaked/rumor prices for the lower end Ryzens weren't as dramatic as the Ryzen 7 set, and the leak prices turned out pretty accurate for those.

The 1600X (6/12) is supposedly ~$80 less than the 7700K. Even the 1700 is going to be listed for about the same (~$10 less) what the 7700K is selling for on amazon and newegg right now.
https://www.techpowerup.com/230916/pricing-of-entire-amd-ryzen-lineup-revealed (note prices of the top 3 SKUs were confirmed in the AMD presentation)

Canned Sunshine
Nov 20, 2005

CAUTION: POST QUALITY UNDER CONSTRUCTION



kirtar posted:

The 1600X (6/12) is supposedly ~$80 less than the 7700K. Even the 1700 is going to be listed for about the same (~$10 less) what the 7700K is selling for on amazon and newegg right now.
https://www.techpowerup.com/230916/pricing-of-entire-amd-ryzen-lineup-revealed (note prices of the top 3 SKUs were confirmed in the AMD presentation)
Yeah, I'd rather get a 1600X than a 7700K from what it looks so far, just because the extra cost doesn't seem worth it for the performance gain.

We'll see how the actual reviews go when the NDA lifts.

PC LOAD LETTER
May 23, 2005
WTF?!
I think spending over $200 on any consumer mobo is hiiiiighly questionable. Unless it has some must need function built in that you really want and can't get any other way (it doesn't) I'd look at another mobo.

BurritoJustice posted:

It seems from checking launch motherboards that Ryzen is capping out at 3200MHz DDR4 (at least for the displayed max speed on motherboards), which is somewhat disappointing.
Apparently there are some AM4 (edit) Biostar mobos that support up to 3600Mhz DDR4 so it might just be a board limitation.

IMO if you're really serious about pushing the memory speeds to the limit I'd wait for high clocked memory modules that have AMP support (its like Intel's XMP profiles) for those high clocks to be sure you get what you want.

PC LOAD LETTER fucked around with this message at 07:27 on Feb 23, 2017

Anime Schoolgirl
Nov 28, 2002

Unless they changed something about the memory controller completely for AM4, AMP is XMP though. It's just branded differently and a lot of "AMP" supporting boards still name it XMP on the bios

PC LOAD LETTER
May 23, 2005
WTF?!
AFAIK its different in that they have to do some sort of tuning for AMD's memory controller but AMP memory modules can have XMP profiles too if the manufacturer put them in.

apropos man
Sep 5, 2016

You get a hundred and forty one thousand years and you're out in eight!

SourKraut posted:

We'll see how the actual reviews go when the NDA lifts.

Is it common knowledge yet when the NDA is lifted? Will we see independent reviews done by testers in their own time on their own rigs over the weekend?

eames
May 9, 2009

Judging from the various leaked benchmarks they've exceeded my expectations but all in all a 7600K around 5 Ghz would still be the best value desktop CPU for me (mostly using it for simple casual :pcgaming:).
That's assuming the smaller, lower TDP 4/6 core parts don't turn out to be overclocking beasts which is rather unlikely. Oh and no showstopping bugs and erratas! :pray:

AMD needs process improvements from GloFo and more games like Watch Dogs 2 which scale from threads instead of Mhz. There are very good chances that future game engines will do that but by the time a 8-core CPU is better across the board than a 30% faster 4-core CPU, Intel will probably have caught up.

Fun fact: an analyst report stated that "over 99%" of Intel's desktop CPUs sold in 2016 included an integrated GPU, so 6-10 core high end desktop CPUs sales accounted for less than 1% of their sales. Not surprising but it puts the whole "6900K killer" hype into perspective.

Anime Schoolgirl
Nov 28, 2002

Intel seems to be actually way more worried about the Zen core's potential in the server market though, the 1700 and 1800x non-XFR efficiency per core is better than Broadwell-EP and they're using a marginally bigger process. Desktop sales are rear end pennies compared to that market.

priznat
Jul 7, 2009

Let's get drunk and kiss each other all night.

Anime Schoolgirl posted:

Intel seems to be actually way more worried about the Zen core's potential in the server market though, the 1700 and 1800x non-XFR efficiency per core is better than Broadwell-EP and they're using a marginally bigger process. Desktop sales are rear end pennies compared to that market.

Has there been any hints on when Naples will show up? This year perhaps?

64 PCIe lanes on the CPU is pretty exciting actually.

PC LOAD LETTER
May 23, 2005
WTF?!
I can't find any real good information on total system power usage with Ryzen systems under load but there is this from Charlie of all people:

https://semiaccurate.com/2017/02/22/amds-ryzen-7-1800x-beats-intels-i7-6900k-half-price/

quote:

For “Recorded Idle Wall Power”, the AMD 8C systems had an ~50% power advantage over Intel’s 8C systems while the 7700K 4C had a 25%, 10W, power advantage over AMD’s 8C 1700.

For HPC Intel will still be the safe buy but for anything else in the server market AMD is looking very competitive and maybe a sure thing to buy assuming they also price their Opterons far below Intel's Xeons much as they have with consumer Ryzen vs Intel's HEDT. We don't have much information at all on Zen Opterons though. Which is is kind've odd since supposedly AMD was going to launch the server Zen first then do desktops but that was the plan years ago and things change.

incoherent
Apr 24, 2004

01010100011010000111001
00110100101101100011011
000110010101110010

Anime Schoolgirl posted:

Intel seems to be actually way more worried about the Zen core's potential in the server market though, the 1700 and 1800x non-XFR efficiency per core is better than Broadwell-EP and they're using a marginally bigger process. Desktop sales are rear end pennies compared to that market.

I'd be loving worried as much as intel if they're going to hit compute times cheaper than intel. 4c AWS instances for the price of 2?

eames
May 9, 2009

Yeah they should be, particularly if AMD gets those "Zeppelin" 32 core Zen/Vega/HBM APUs out the door. I don't think Intel even has response to that unless they license it from AMD.

edit: typo

eames fucked around with this message at 08:28 on Feb 23, 2017

incoherent
Apr 24, 2004

01010100011010000111001
00110100101101100011011
000110010101110010
Free gpu with your compute time to run that nearly-compatible CUDA code on? Everyone should be worried.

Truga
May 4, 2014
Lipstick Apathy

eames posted:

oooh juicy


source with links

TLDR is that L2 and L3 cache latency seems higher than it should be but the AIDA dev doesn't have a sample so he cant tell if it's a software bug, cpu bug, cpu bottleneck or architectural issue. AMD complained about low scores but wouldn't send him a sample. Trying to bench AIDA64 with Ryzen will pop up a warning for the reviewer to ignore the score because "the benchmarks are not properly optimized for the current CPU."

Smells like the sort of thing you'd want an NDA for if this is all true.

the whole translated post history by an AIDA64 dev can be read here:

http://translate.google.com/transla...n&langpair=auto|en&tbb=1&ie=UTF-8

poo poo like this is why I probably won't buy just yet, there's bound to be a bugfixed zen+ coming down the pipeline in a year or 2, and until then my haswell i7 should hold just fine.

Anime Schoolgirl
Nov 28, 2002

PC LOAD LETTER posted:

Which is is kind've odd since supposedly AMD was going to launch the server Zen first then do desktops but that was the plan years ago and things change.
They probably worked on those first, but since they're server chips they have to validate them for longer through engineering samples before making it retail-ready, unlike consumer chips where those kinds of bugs are worked around in microcode or at least tolerated.

Truga posted:

poo poo like this is why I probably won't buy just yet, there's bound to be a bugfixed zen+ coming down the pipeline in a year or 2, and until then my haswell i7 should hold just fine.
This is a revamp of an architecture like Sandy Bridge was for Intel, and like SB it'll have some teething issues that may or may not be fixed in microcode. I think it took Intel about 2-3 months for the noticeable ones to be ironed out.

eames
May 9, 2009

It looks like AMD ran all official benchmarks on the the 6900K with dual-channel memory (2x8GB) instead of quad-channel. :crossarms:

http://www.anandtech.com/Gallery/Album/5485#32

Anime Schoolgirl
Nov 28, 2002

Ryzen has only has 2 channels vs. Broadwell-E/EP's 4, so to their credit, it actually equalizes that particular factor.

Nobody except crazies using that one mini-ITX board for that socket will actually use only two channels though :v:

some dillweed
Mar 31, 2007

Anandtech's numbers have the 6900K with a (DDR4-2400) quad-channel setup at 1547 multi and 153 single in Cinebench, compared to the AMD benchmarks at 1474 and 160. The AMD numbers are also slightly lower for the 7700K and 6800K multi-threaded (around 20 points). Would the performance lead the 1800X has there shrink even more if the Broadwell-E setup used faster memory? Does it matter if it's still half of the price?

I'm still put-putting along with my overclocked 2500K, but I'll be waiting anxiously like everybody else to see how Ryzen performs, even if just for (hopefully) some competition and change for once.

champagne posting
Apr 5, 2006

YOU ARE A BRAIN
IN A BUNKER

It matters in the sense that you may want to trade speed for dollars.

Sophy Wackles
Dec 17, 2000

> access main security grid
access: PERMISSION DENIED.





I'm tempted to pull the trigger to update my OCd 2700k, which I've owned longer than any CPU before. The thing is still a beast though!

I'm likely going to wait a few months though to see how things shake out and see if Intel lowers prices to compete.

some dillweed
Mar 31, 2007

Yeah, I guess if you're in the market segment that only cares about whatever the fastest option is regardless of cost then Ryzen still isn't going to be your choice at this point. Again, hopefully this will at least shake things up a bit. It would be nice to see some actual competition again.

SlayVus
Jul 10, 2009
Grimey Drawer
Looking forward to seeing if Ryzen will support IOMMU and AMD-V. The 1700 would actually make a good unRAID CPU for gaming hardware pass through. Though it looks like you might have to wait for some workstation boards from Asus or ASRock since the boards that have three 16x slots run the third slot at 4x. If they can run ECC at launch that would be awesome.

The only reason why I want anything like that is because I actually enjoy inviting my friends over to play games, but they don't have the best computers. I could also use it to run my Plex server and possibly a dedicated server for a game or two that I play.

SlayVus fucked around with this message at 10:45 on Feb 23, 2017

eames
May 9, 2009



mid-range R5 Q2 2017

low-end R3 2H 2017

http://imgur.com/a/U6ysr

edit: more bad news!

servethehome posted:

We did ask about a potential single socket Ryzen/ Zen part with ECC memory support and were told that AMD was not announcing such a product at this time alongside the Ryzen/ Zen launch.

https://www.servethehome.com/amd-ryzen-7-parts-available-for-pre-order-now/

eames fucked around with this message at 11:47 on Feb 23, 2017

Nam Taf
Jun 25, 2005

I am Fat Man, hear me roar!

On the overclocking wall front, if I understand correctly that's probably not a terrible place to be in insofar as hopnig that GloFo will gradually unfuck their process which will shift that wall forwards. So B and later steppings of the Ryzen should see more headroom through natural process improvements, yeh?

I mean I'm assuming it's not an architectural issue but Zen doesn't seem to be an architecture that would run dry at 4.2GHz so I'm assuming it's more the GloFo factor. Am I right in that understanding?

Adbot
ADBOT LOVES YOU

Anime Schoolgirl
Nov 28, 2002

ECC was never officially supported by AMD for a long while for consumer boards, but if everything since Socket 939 were anything to go by, just about every mainline (non-FM1/FM2) desktop socket can support ECC RAM if the mobo manufacturer provides traces for it.

  • Locked thread