What was Intel thinking releasing Skyshit-X? Even worse, marketing

File: mpv-shot0054.jpg (163KB, 1280x720px) Image search: [Google]

163KB, 1280x720px

Breakdown.

Some DX12 games are really acting funky, while their DX11 renderpath is perfectly fine on Ryzen

Unboxed blames Nvidia for this but I'm not even sure anymore

>>61485962
They though that more L2 cache would help it not to be starved for bandwidth.

That is pretty much it.

>>61485962
The absolute state of JUSTel

>hot as fuck
>expensive
>sucks for gaymen
>mediocre workstation performance
It's over, Threadripper won before it's even released.

>Mafia 3

>every new patch seems to lower Ryzen performance while boost Intel performance, I don't know what's up with that.
>at launch Ryzen was beating 7700k in this game

Lmao

>>61485962
>Even worse, marketing it for gamers?
Because gamers tend to buy overpriced shit and Intel likes money.

>>61486387
Who the fuck still plays (or ever played) mafia 3?

>>61486429
Apparently Intel considers it important if they have to pay developers to gimp

>>61485962
by 1% you mean the a=.99 , 1-a quantile? for what purpose?

File: 1490104602434.jpg (25KB, 641x530px) Image search: [Google]

25KB, 641x530px

4.0ghz vs 4.7ghz only 1 fps better

zen2 would kill

>>61485962
different cache strategy means that programs optimized for the old cache will run slower with the new CPUs.

Cooling setup for that Skylake costs more than the entire am4 platform

10% worse than deliddis that costs 50% more.

1600 is a steal especially with b350 boards

>>61485962
They though copying Bulldozer caches was a good idea (hint: it was not, large chucks of L3 made Core good).

>>61486880
You can't optimize EVERYTHING to use private L2.
Most programs live and die on shared L3.

>>61485962
>0.7 Ghz higher
>only beats Ryzen by 8%

Really make you think

>>61486954
No nigguhurz will help with L3 deficit.

>>61486958
Why did they cheap out on L3 cache though ? Is it because there were hardware limitations or because they were too greedy?

>>61486984
It's because they went for "larger private L2" route,, while forgetting Core and most software lives and dies on large chunks of shared L3.
Fuck Intel themselves pushed large chunks of very fast L3 in Nehalem as main source of performance boost and now they are going fucking backwards.

File: 1464754668558.jpg (23KB, 300x400px) Image search: [Google]

23KB, 300x400px

>>61485962
>mfw R5 1600 masterrace

File: DFOcZtJXsAImmwd.jpg large.jpg (78KB, 525x525px) Image search: [Google]

78KB, 525x525px

holy moley

File: just fine.jpg (15KB, 300x168px) Image search: [Google]

15KB, 300x168px

>>61487278

>>61487278
What?

File: wtf.png (146KB, 356x256px) Image search: [Google]

146KB, 356x256px

>>61487278

>>61487295
https://twitter.com/HardwareUnboxed/status/888225519141093376

this was basically unheard of outside pushing 5.5+ghz on LN attempts

>>61487336
Holy moley
what the fuck
It's been YEARS since i've seen burned pins on CPUs.

>>61487336
What the FUCK is Intel doing?

>>61485962
Welp, if Zen2 can really hit north of 4.5GHz clocks, I don't know if Intel will even have the gaming advantage anymore.

>>61486903
>>61486954
You jest? They perform 100% on par.

>>61487429
Try 5ghz.
https://www.globalfoundries.com/sites/default/files/product-briefs/7lp-product-brief.pdf

>>61487429
It's 5ghz.
It's a node intended for POWER first which means absolutely nuclear over-powering clockspeeds.

>>61485962
Where's the 1700? Should be slightly faster than a 1600 in gaming.

>>61487512
Quite the contrary. Not a single game can efficiently utilize more than 12 threads right now.

>>61487516
Not due to that, due to the need to do less IF > CCX hops on 4 core CCX chips like R7, should improve perf in gaming by a few %

>>61487527
Windows 10 hops threads around CCXs like it's nothing regardless if it's 1700 or 1600.
This guy has a quick rundown on the problem
https://www.youtube.com/watch?v=40h4skxDkh4

>>61487527
Pretty sure the 1700 performs within .1% of the 1600 in all games. Only time it makes a difference would be cinebench and other not-gaming properly multithreaded applications that can supply enough of a workload for the extra cores to matter.

File: Screenshot_20170721_125606.png (33KB, 643x288px) Image search: [Google]

33KB, 643x288px

>>61487543
>>61487553
1600X and 1800X have the same turbo and all core turbo clocks, 1800X is some 2% faster in gaming.

The effect should be the same with a 1600 and 1700 at 4.0
So yeah, there's obviously some SLIGHT benefit from doing less communication through IF when 3+ cores are used.

>>61487590
https://www.computerbase.de/2017-07/core-i-ryzen-ddr4-ram-benchmark/2/#diagramm-performancerating-fps-1920-1080

>>61487590
Of course faster and more consistent L3 hits will bring more perf but the difference is marginal at best.

>>61487442
is the 5 ghz the performance target of the same tdp as the 3 ghz they advertise for the 14lpp

>>61487618
Yeah, but it's still nice that there's more perf on the table, 8+ cores should really be bought if you have a need for them.

The 6 core Ryzens are by far the best perf for the money, for gaming and really decent multithreaded perf

>>61487633
>The 6 core Ryzens are by far the best perf for the money, for gaming and really decent multithreaded perf
Yes, hence why they are selling like hotcakes.

>>61487628
Performance, Power, Cost Advantage
from 7nm Area Scaling
>40% performance improvement at iso power (vs. 14nm)
>60% power reduction at iso frequency (vs. 14nm)
>Up to 30% lower die cost (vs. 14nm)

From the PDF.
These are just ballpark figures though. you're likely gonna be looking at Zen2 8 core with close to 5.0GHz single turbo at 95W, and all the cores around 4.4

>>61486013
I recently saw Adored blame Nvidia aswell
https://www.youtube.com/watch?v=0tfTZjugDeg
It's real indepth look at ryzen vs intel vs nvidia vs dx11 vs dx12

File: jewtel.jpg (83KB, 681x522px) Image search: [Google]

83KB, 681x522px

>>61485962
MUH SINGLE CORE IPC

AMD LIES. DON'T BUY AMD!

INTEL IS GOOD TO YOU. AMD IS BANKRUPT!!!!!

You don't need it unless you write AVX-512 asm yourself

>>61485962
Skylake-X is an abortion, what else is new.

>>61487657
> you're likely gonna be looking at Zen2 8 core with close to 5.0GHz single turbo at 95W, and all the cores around 4.4

i assume the 40% figure is including the enhanced euv lithography in which case you wont see it until zen3? still i dont see how intel can possibly keep pace with that kind of process advancement given that it's coming along with architectural improvements if i understood mark papermaster correctly

>>61487722
No, its for DUV.

>>61487722
EUV is just increased density and lower cost, electrostatic characteristics remain the same as DUV

>>61487762
wew lad

>>61487657
>>Up to 30% lower die cost (vs. 14nm)

What the fuck
This process is fucking whack

>>61487798
What do you mean?

>>61487809
Usually new processes cost significantly more than the previous ones when starting out until yields stabilize, ask Intel.
This implies that 7nm LP from Glofo/IBM is expected to have really good yields from the start.

>>61485962
>>61486013
The elephant in the room is that this was ran with a 1080ti.

On a 1080p monitor.
Go to 1440p or 4k and the 7700k's 10% advantage vanishes.

>>61487829
They are working closely with AMD and doing risk production runs starting H1 2018.
Also their yields on 14nm LPP are stellar.
It took time but GloFo unJUSTed itself.

>>61487840
>4.7ghz vs 4.0ghz
>B-but 1080p

kys

>>61487840
Of course the advantage vanishes, it's because the game would then be GPU limited and not testing the CPUs at all you fucking moron

>>61487829
Has 7nm LP completely fucking dehumanized Intel?


DUV has
>similar density
>higher performance
>lower Vmin
>lower mask count

Meanwhile EUV just a few quarters latter will have
>higher density
>even lower cost

Intel's 10nm seems like an abortion on the level of a Skylake-X

>>61487896
DUV has higher density.
7nm LP offers MUCH denser SRAM.

>>61487895
So if I game with a high res monitor why would I care about 1080p performance?

These gaming benchmarks are nothing more than "WHAT IF" performance, not actual real life performance.

At 1080p you use Polaris, or 1070/1060.
Not a fucking 1080ti

>>61487909
It is indicative of FUTURE performance differences when games become much more CPU intensive you 89iq simpleton.

>>61487922
Oh no, it's retarded.

>>61487933
I'm afraid retards have no means to appraise other's intellectual capabilities.

File: 1497391371679.jpg (69KB, 350x350px) Image search: [Google]

69KB, 350x350px

>>61487657

No fucking way they will release this Q2 of 2018.

I am so fucking hype, even if they have the same performance as Zen1 but at better temps and lower power consumption, it's still gonna be a day 1 buy for me.

>>61488071
It's probably early Q4 for Zen2.
Also it's definitely higher performance.
AMD will go for the throat.

>>61488117

But then does that mean that they will skip the 14+ process from their roadmap and go right for the kill with 7nm?

A Zen1 refresh would be nice, esp. since AM4 is going to be viable for a few more years.

>>61488235
Both Zen and Vega on 14nm+ are APU components.

File: uOEY7eUThSxsrZEnzx2_OKBxfOW0TLik8-P5TmQu-V0.jpg (105KB, 640x757px) Image search: [Google]

105KB, 640x757px

>>61487295

>>61488330
But it's just 6core model.
Why.
What the fuck.

File: 1487297768743.jpg (125KB, 960x878px) Image search: [Google]

125KB, 960x878px

>>61485962
SHUUUUUUUUUUUUT UUUUUUUUUUUUUP

>>61488330
HAHAHAHAHAHA
*breathes*
HAHAHAHAHAHA

>>61488330
ITS wORKING AS INTENDED

>>61488117

Nice, did they spill any more beans on when we can expect new APU devices? I wonder if my friend should put off buying an AIO.

File: KonyeJ East.jpg (14KB, 397x437px) Image search: [Google]

14KB, 397x437px

>>61488330

>>61488491
>Nice, did they spill any more beans on when we can expect new APU devices?
H2 2017 for APUs, probably early Q4.
They need to finish Vega drivers first.

I wonder if Coffee Lake will have the same gimped cache arrangment and thus be a downgrade from Kaby Lake for gamers.

>>61487922

Yes this future where games won't become any more multi-thread aware, even though they already are doing that. i5 is the perfect gaming CPU guys!

>>61488503
No, it's 12mb of L3 for 6core model.
Clocks though...

File: Alg_pkYA0avOZ3ldlB5diH1ENHFkDMOq4I5IohZu3y4.jpg (141KB, 1024x741px) Image search: [Google]

141KB, 1024x741px

>>61488330

>>61485962
They had to do SOMETHING for the sake of appearances.

>>61487657
What is it made of? Magic and dreams?
>tfw the era of AMD is actually starting

>>61488994
Pure condensed Nazi science designed to extreminate Jew menace in HPC.

If 7nm is going to make Zen2 so insanely overpowered, what will it do to Navi? I'm so excited

File: 1500585814492.jpg (510KB, 1315x794px) Image search: [Google]

510KB, 1315x794px

>>61488330
Please be removing this

>>61489037
make it passable

>>61489037
>what will it do to Navi?
Very very dense and peformant base dies for MCM.

>>61485962
whats even better is coffee lake is going to have the same problems going from ring bus to mesh architecture, you think it will just a 7700k with 2 extra cores? hahahaha

>>61489162
>you think it will just a 7700k with 2 extra cores?
It is, also moar cachez and lower clocks.

>>61485962
>4ghz r5 1600 as good as 4.7ghz i7 7800x in gayming
Wow, just wow. They completely fucked up IPC with their bingbusmesh

>>61489162
>>61489175
aren't they going to shrink it to that 10nm meme? I don't think they withdrawed from following their tick-tock way

>>61489162
Coffee Lake will still be bingbus.
Mesh is only for Xeon and SKL-X (aka badly binned beta Xeons)

>>61489175
its more then that, dont you find it odd a 7800k @ 4.7 is 11 fps behind the 7700k @ 4.9?..... .2 GHz doesn't add that much performance its a change its the change to mesh architecture trying to copy ryzen

File: Screenshot from 06-07-2017 19_12_35.png (36KB, 571x556px) Image search: [Google]

36KB, 571x556px

>>61489224
No, it's Skylake refresh.
Also Intel's 10nm node is poo.
>>61489233
It's not mesh, it's caches. 7800x is L3 starved.

>>61489227
hope your right if not its going to be garbage

>>61489242
It's already shit due to clocks.

>>61489253
i5s are getting HT right? they will actually be better for gaming with coffee lake the meme will finally become true

>>61487694
In terms of actual IPC, the ryzen beats or ties with the intel parts.
Intel just stays ahead because they're running their chips at toasty speeds.

>>61489285
No, i5s are Thuban reborn, i.e. 6/6.

>>61488330
That is pretty bullshit some defective board. 7800 don't draw much power not nearly much more than 7700k. If this was an issue a lot of 7900 would be burn

File: JUST_BRIAN.jpg (119KB, 1000x1000px) Image search: [Google]

119KB, 1000x1000px

>>61488330
Holy shite

File: 87080.png (47KB, 650x350px) Image search: [Google]

47KB, 650x350px

>>61489337
Intel uses more than 2x the power.

>>61488117
>Q4
Would be nice if true but why would they release it in less than a year and mess up their schedule instead of keeping it yearly?

>>61489465
Q4 2018, anon.

>>61489465
Glofo 7nm is ready. Christmas came early. Before they were planning on a stop-gap "Zen+".

>>61485962
Skalyake is not a gaming arch, it's a server arch, it's not the best in gaming.
Meanwhile Ryzen is a GAMING arch and it barely competes with Skylake and gets absolutely killed and destroyed to infinity in servers

File: glofo1.png (129KB, 1562x332px) Image search: [Google]

129KB, 1562x332px

>>61489482
>Glofo 7nm is ready
They are talking about hitting 5GHz clocks and lower power usage too. IBM infusion benefited Glofo massively

>>61489566
What?
Zen is a server core in everything including AES-128 engines attached to memory controllers.
This is some 2/10 tier bait.

>>61489572
No shit, a healthy IP and staff infusion can unJUST any company.

>>61488330
>1
>.
>2
>5
>v

>>61489566
>>61489580

And infinity fabric, CCX and multi-chip design, excellent power savings at conservative server level clocks. Everything in Zen platform geared towards servers. Intel must be desperate to call Zen a gaming platform, they never talked about competing AMD products before

>>61485962
those bar looks fucking stupid wtf I read the number and ryzen seems almost par skylake.

>>61489749
How old are you?

>>61489233
Mesh isn't copying Ryzen, it's trying to reduce inter-cache latencies and bottlenecks for high core count CPUs in a different way.

Zen is about giving small clusters a fast shared L3, Skylake mesh is about still keeping chip-wide L3 sharing but not completely falling over at 12+ cores when doing so.

Imagine AMD blowing all expectations and Zen2 actually has 5.5GHz clocks like POWER9

>>61485962
>What was Intel thinking releasing Skyshit-X?
"there's a market for moar coars, we need a moar coars product ASAP"

problem is that their uarch, die schematics and platform aren't built from the ground up to this do this specific thing well

meaning they get hot chips with terrible yields that only run on ~$300 mobos and everybody makes fun of them even though they still technically have the tech lead

they should take a page from AMD's book and realize that releasing late is always better than releasing unfinished

File: wtfspin.gif (1MB, 210x144px) Image search: [Google]

1MB, 210x144px

>>61489566
>Meanwhile Ryzen is a GAMING arch
Been drinking the Intel marketing slide Kool-aid a little too hard there, buddy.

File: 1399402966929.jpg (184KB, 1320x666px) Image search: [Google]

184KB, 1320x666px

>not still using an OC sandy bridge

File: 96vIsCsyN2OasUHT.jpg (358KB, 1500x844px) Image search: [Google]

358KB, 1500x844px

>>61489884
They are throwing shit at AMD's multi-chip while they are trying to rush their own
https://venturebeat.com/2017/03/28/intel-moves-tech-forward-by-putting-two-chips-in-a-single-package/
Totally caught pants down

File: Intel Merchant.jpg (9KB, 250x250px) Image search: [Google]

9KB, 250x250px

>>61487278
>>61488330

D E L E T
E
L
E
T

>>61489566
skylake-x isnt server LMAO its for MEGATASKING

File: extreme-memetasking.jpg (55KB, 740x416px) Image search: [Google]

55KB, 740x416px

>>61489972
Memetasking was Broadwell-E, dead and buried. It's EXTREME MEMETASKING now

>>61490018
I hope they're keeping ULTIMATE MEGATASKING for the next arch

>>61490018
I guess the "Ultimate platform" is binned Purley Xeon housefires. Who would have guessed.

>>61489566
Servers rarely have threads talking to each other.
It's either a gazillion of threads with each doing its own thing with the sock/SQL etc or a trillion of virtual machines.
And well, AMD just delivers more threads per dollar.

>>61490045
ULTIMATE SHADOWFLARE EXTREMETASKINGxxSEPHIROThXXx

>>61489802
They said 5.0, I wouldn't be surprised if it OCs to 5.5.

>>61490132
You're not funny.

>>61490208
That's what they said when they saw the gas chambers.

File: poster,220x200,ffffff-pad,220x200,ffffff.u2[1].jpg (7KB, 220x200px) Image search: [Google]

7KB, 220x200px

>>61485962
>7700k still shits all over Ryzen
>AMD IS STILL BETTER

>>61490369
>>shits all over Ryzen
>10fps over a fucking 1600

>>61490369
The picture looks a bit bad to AMD until you look the pricetag of those things.

>>61490390
>>61490427
>nigger reviewer is using $200 ram for ryzen
>Just check the price tag

>>61490460
standard 3200 mhz ddr4 sticks are 150 bucks though

>>61490491
Ryzen needs special expensive RAM

>>61490460
...and how much was the RAM that he used for the 7700K?

>>61490521
B-die and Hynix work fine for Ryzen, stop dreaming of 4 months ago.

In fact hynix double rank memory of the same frequency is slightly faster than Samsung B-die on Ryzen, B-die is really good for overclocking though

In summary

cheapo Hynix 3200Mhz > 3200Mhz b-die at stock
3200mhz b-die > 3200Mhz hynix overclocked

>>61490521
No it doesn't. It's may be kind of a pain in the ass to find RAM that's on the QVL list for the MOBO and runs at 3200MHz, but it's not any more expensive than normal.

>>61485962
Not actually good tests, both intel and amd aren't at their powerfull

Intel can hit 5.3GHz and 4500MHz memory normally

>>61487378
intel had a perfectly shaped plan around mild oc and milking the customers
when ryzen came even with all the problems people started to see how bad intel really is and how bad they were manipulating and still are the market
amd didnt really had to do much on the pr level simply because the cpu's were speaking for them selfs eventually they gained enough traction

>>61490460
Intel doesn't need RAM now? Sign me up

>AMD up $2.20 just today, most active stock on Robinhood
>Intel on track to match their 1 year low, with a (relatively speaking) massive drop earlier this month

Feelsgudman

File: Ryzen7_7700k_1080ti.png (2MB, 1920x1440px) Image search: [Google]

2MB, 1920x1440px

>>61490369
LMAO, a 7700k at 5ghz no faster than a 1700X at 4.0ghz in games so long as you use 3600mhz ram

>M-muh GUPU BOTTLENAKZ!!!
That's with a 1080 Ti at 1080p and 1440p, idiot.

File: glue inside.jpg (32KB, 680x510px) Image search: [Google]

32KB, 680x510px

>>61490369

>>61490560
OCing 7700K to extreme levels doesn't actually increase FPS much if at all, and you're talking golden silicon that few 7700K can even hit.

File: 1499799036743.jpg (281KB, 1599x899px) Image search: [Google]

281KB, 1599x899px

>>61490018
truly memetic megatasking

File: 1497495735797.jpg (59KB, 493x493px) Image search: [Google]

59KB, 493x493px

>>61490018
>mfw

I know reviewers like the non-X CPU's better than the X ones because the difference between them disappears when overclocked, but if he's going to do so many stock benchmarks he really should have used the 1600X instead due to the higher base clock/turbo and 50MHz highher XFR boost. It makes Ryzen 5 look worse at stock when in reality he's just using a lower clocked CPU, while the 7700K represents Intel's best. Youtubers can shout "but muh overclock" but the reality is most people don't OC their CPU's, so they'll judge the 1600 for it's low base clock performance when the stock 1600X would likely close the gap between the stock 7800X.

>>61490838
I've also seen plenty of clueless reviews test on B350 boards that have half rate XFR or might be even disabled

Anyway, you need x370 if you want XFR and aren't overclocking

>>61486429
I bought it a couple days ago, runs like trash on my 980Ti/1700X rig at 1440p. :^(
The game is dogshit from what I can tell though, so no big loss. Shame, the city atmosphere was nice and kinda reminded me of mafia 2 (the only way it does) in a comfy way.

>>61488330

Luckily, that contact was a placeholder since Intel just upped the pin count to artificially create a new socket for increased motherboard sales

File: 1497165045195.png (508KB, 566x482px) Image search: [Google]

508KB, 566x482px

>>61490202

No fucking way. That would 100% place AMD on top in single-core performance (and therefore in gaming) unless Intel starts bribing devs HARD. If 5.0 is not an exaggeration, then it would be literally another shoah.

>>61491047

My fucking sides

File: 1500517074521.png (124KB, 358x272px) Image search: [Google]

124KB, 358x272px

>>61490369
>14% more fps on 31% more base clock speed
>7% more fps with 5% more base clock speed
>Barely scales on 4.9GHz with a fucking 1080Ti
>100$ more expensive
>Shits all over

>>61490460
>>61490521

>Shelling out 20$ more on RAM to perform within 10% of a 100$ more expensive Intel chip

>>61485962

>700MHz frequency difference
>0% delta between a R5 1600 and an i7-7800X

Jesus christ. The fuckup could not be more massive.

>>61486995

Intel has what? Hundreds, maybe thousands of engineers?

They must have been screaming at management not to push these chips to market. No way they didn't notice the heat and power draw or the performance regressions.

Fucking marketing.

>>61486995
It was supposed to help with AVX, wasn't it? Or server workloads? Doesn't seem to have done much.

>>61492845
106K employees.

>>61486013
Nvidia DX12 drivers are pretty bad

File: 78ad8145114fea6083da125577ca97fa502f3285bfb565c9bce058b606bc416b-leftypol.jpg (130KB, 692x763px) Image search: [Google]

130KB, 692x763px

>>61490521
Hynix M-die is fine

>>61490521
My Hynix shit is working fine @ 3200 on my R7 1700 @ 4ghz.

File: 1470646625890.png (463KB, 1070x601px) Image search: [Google]

463KB, 1070x601px

>>61490623
>tfw didn't fall for the "buy AMD stock" meme when it was ~$2

>>61491426
16gb of 3600 is about $70-$80 more US. Also you can barely even hit that speed with a couple of mobos.

>>61490751
(YOU) deserve this anon

>>61494825
Point of diminishing returns is 3200 though.

>>61494825
But we're talking about 3200MHz, not 3600.

Cause gamers are faggots who will buy this stupid shit as a way to feel special.

>>61494986
are you sure about that?

File: 1435933790881.webm (2MB, 1920x816px) Image search: [Google]

2MB, 1920x816px

And just think, the real battle starts with future products and 7nm.

>>61487922
If you think that is an indication of future performance, YOU are the simpleton. Because games change too, you won't be playing the same games in 2022. Considering how bad i7 7700K multi-core performance is, even though it's winning today I doubt that will remain for long if APIs like Vulkan and dx12 along with multi core optimization get any traction at all. Which there is reason to believe will happen, considering it helps multi-platform releases.

>>61486013
how could you not be sure anymore when there have been others who have noticed similar as well?

hell i've even noticed it with my 980 ti. well the dx12 regression part. in mankind i get about 4 fps less in dx12 vs dx11 and more stutter. while my buddy with a rx 480 gets the same performance in dx12 as dx11 but doesn't get stutter. i know this is more of a gpu vs gpu thing but i don't doubt there are issues with their drivers with multi-core processors.

>>61495395
lol.
>>61489238

>>61487922
thats kinda horseshit. you would think the 1060's 10-15% faster lead in dx11 vs the rx 480 would translate over to dx12. since you know, INDICATIVE performance. it performs so well it HAS to perform equally as well in dx12!

well when dx12 comes around that lead actually goes for the rx 480. that 10-15% lead drops down to a +1-5% for the rx 480.

all of that with release performances of course. over the past year the rx 480 and 1060 dx11 gap narrowed pretty close. the 1060 now being ~5% faster average. dx12 is still the same.

switch over to the 580 and the 580 is 5% faster in dx11 and maintains a strong lead in dx12.

yeah it might HELP give a POSSIBLE idea but its not set in stone. look at the entire maxwell series. its pretty bad in dx12.

so when it comes to a gpu's 1080p performance the same thing applies. it doesn't necessary means it will translate into future performance for feature games. just because it performsn well in 1080p doesn't mean that two years from now at 1440p you will get the same level of performance. a lot can change. a lot can effect it.

like the 780 ti. same gpu performance as a 970, but severely held back by its 3gb of vram in newer games. you're not getting the same level of performance as you did at release since a lot of newer games break the 4gb vram barrier at 1080p. like battlegrounds. shit even space engineers at 1080p will use 5gb of vram.

File: Untitled.png (4MB, 2560x1440px) Image search: [Google]

4MB, 2560x1440px

>>61495584
>7.7gb of vram usage
i'm running space engineers in a 1920x1200 window since i have a 1440p freesync display. yeah, 3gb, even 4gb is NOT enough for even 1080p these days. to many games will GLADLY use more than 4gb at 1920x1080/1200.

>>61495833

Allocate =/= need. DX memory management is pretty meh (only because opengl memory management is something best left in the Eye of Terror).

>>61491396
5.0 is the number given by Global Foundries as the target clockspeed for 7nm

>>61491396
It's supposed to get 5 niggahurtz at overdrive voltage (that is also lower than 14nm LPP's).
The shoah is inevitable.

>>61486849
This. Zen1 is doing this well as a buggy AF first gen design on a shit-tier 14nm Samsung finFET node.

What happens when the first major revision comes out with ~10% IPC gains and on a new 7nm FinFET node with clocks up to 4.7-4.9ghz?

Intel is fucked. The single thread delta between their cores and Ryzen will be almost non-existent and due to the smaller more efficient design AMD can pack more zen cores into smaller die areas which means more cores at lower prices. Also their motherboards are much cheaper too.

>>61496414
Best number.
Doesn't mean a certain project like Zen will reap 40% extra perf or frequency. These are throrical top.

>>61498739
Well, the same slide said the current process was "3GHz+", and Zen can run 4GHz.

>>61487895
It's also not a realistic test to do with a 1080 ti.