im sorry, WHAT

oy vey

Reminder: only poor people buy AMD. No one's gonna take you seriously unless you have an Intel Core i9™ cpu.

>>61698729
This just in, highly multi-threaded workloads scale really well with more threads.

File: trashman6.png (591KB, 778x1018px) Image search: [Google]

591KB, 778x1018px

>company marketing slides

File: tr amazon.jpg (219KB, 1536x2048px) Image search: [Google]

219KB, 1536x2048px

>>61698767
>Wealthy AMD noblemen buying superior Threadrippers while Intel peasants slum it up with overpriced quad cores

>>61698767
/thread
I also buy 2 extra and keep them on my shelf to make sure no one ever mistakes me for a poorfag.

>>61698767
but they are the same price in the picture. why would you spend 999 dollars on something inferior

>>61698818
Affluent intel buyer can also afford a delidding tool and a fire extinguisher for rare emergencies. The same cannot be said for AMD peasants.

>>61698729
Just wait for the drivers

File: skylake-x aircooler.jpg (66KB, 679x452px) Image search: [Google]

66KB, 679x452px

It's a myth that Skylake-X can't be aircooled.

>>61698869
>CPU
>Drivers
Did you waddle into the wrong thread?

>>61698851
Of course, they process of delidding is to actually to get the glue out so they can eat it. It's the new cocaine, at 999 usa an oz. you also get a free inch burner in-case you want to go the injection route, that just so happens to also play CS:GO.

>>61698767
60% more cores
Just avg less than 30% extra performance
Worse case +60% cores for just 15% extra performance

It's just the toy cores at work.
This is what happens when you build an abomination glued together with and arm like cores

>>61698921
bro just wait for navi

File: 4ghz.png (589KB, 1609x859px) Image search: [Google]

589KB, 1609x859px

>>61698767
>>61698869
SOPA DELICIA

File: Intel-CoreX-Specs.jpg (192KB, 1200x729px) Image search: [Google]

192KB, 1200x729px

>>61698932
The 16 and 18 cores will lose, too, despite costing $700-$1000 more. The clockspeeds just shrivel up.

>>61698932
Worth it. Same price after all.

If the Threadripper cost >30% more then I would be thinking twice about it but you're getting up to 55% better performance for the same price.

>>61698973
Stay mad cuck your toy rupee will just serve poor customer or dumb idiots.

>>61698729
>trusting manufacturer benchmarks

>>61699107
You are just breathtakingly stupid. I almost wish IMG_xxxx was hack here spamming CS:GO benchmarks, even he's more tolerable than you. Kill yourself, BR glueeater.

>>61699107
Stay mad at Intel completely failing to put out an HEDT platform worth half a shit? Ok.

File: TB2f4vIvrJmpuFjSZFwXXaE4VXa_2341477237.jpg (335KB, 1060x1047px) Image search: [Google]

335KB, 1060x1047px

>>61698932
Too bad the Xeon Gold the 18-core Dresdenlake-X is going to be a rebrand of can't even crack 3,000 in Cinebench R15.

Oh, also: INTEL REBRANDXEON HOUSEFIRES

>>61698946
UMA

>>61699186
Because cinebench and WinZip are actual work since the damn of zen

>>61699186
He just ignores all replies proving him wrong and shows up in the next thread spouting the exact same horseshit. You could replace his BR ass with a fucking bot.

File: top_gear.18x07.hdtv_x264-fov.mp4_snapshot_00.53.42_[2016.07.20_21.09.02].jpg (55KB, 720x402px) Image search: [Google]

55KB, 720x402px

>>61698729
INTEL ON SUICIDE WATCH

>>61699234
Don't be mad. We know you are also shitposting cause a neet basement dweller like you don't have use, work or money for an hedt platform and stay jelly with your poor ryzen 5 toy

>>61699264
Back to the favela.

>>61699234
I only replied because I wanted to crack the rebrandxeon line, to be quite honest famalam.

File: 1501600030116.jpg (208KB, 960x874px) Image search: [Google]

208KB, 960x874px

>>61699279

>>61699264
>BR calling other people poor
Lets see your high end desktop then.

>>61699227
You know TR on Cinebench only has 40% more over i9-7900X compared to the 30% average it has over it in other tasks.

Assuming i9-7960X loses to TR by 15% in Cinebench you'd still be looking at a 5% loss on average.

Yes, the 16 core i9 will still lose. It doesn't have the high clockspeed the lower core count CPUs have. And it will cost $700 more.

>>61698942
How
Dumb
Are
You?

File: intel shia.png (535KB, 1097x802px) Image search: [Google]

535KB, 1097x802px

>>61699314

>>61698767
>>61698816
>he doesn't run 4 Xeon Platinums
Poorfags as usual

>>61699545
Screen cap this: when toy rupee is released all your niggers will be crying with how much power it will use.

>>61699623
Nobody's going to care because unlike Intel CPUs, it's actually soldered and can keep itself cool, dipshit.

>>61699623
1x overclocked 1800x - 180-200W
1x overclocked Threadripper 1950x - 360-400W.
Not bad for a 16c processor.

>>61699648
Didn't know Zen was so shitty
Muh 95w tdp

>>61699670
>overclocked
>shit
Fuck off favela monkey

>>61699670
And when you OC X299, it turns into the fucking sun.

>>61699680
>Oc zen
How? You oc 100mhz? Now that is fucking insane.. not even considering the 1.5+ v to past 4.1 GHz

>But muh 1700 oc
This shit will reach 4.0 only on rare exceptions. The better are binned as 1800x that either can't go past 4.1

>>61698767
SOPA DE MACACO
UMA DELICIA

>>61699817
4 GHz on TR is possible
4 GHz on 16 core Intel is not, because 1. jizz TIM, 2. thermals out of control, 3. power usage causes VRMs to melt

>>61699855
Rebranded xeons will be soldered

>>61699481
kek, nice reference.

>>61699874
fake news

>>61699874
Sauce or it didn't happen.

>>61699874
These ARE the rebranded Xeons fuckwit. Intel wholly and steadfastly refuses to solder anything that isnt the >500mm^2 silicon anymore simply because it costs more money than to use the cheap TIM (the gold layers on both the IHS and the die for optimal bonding is probably $1-2, and the indium solder is another $.50-1.00. The shitty TIM costs pennies at most per application)

File: nvidia.jpg (157KB, 700x416px) Image search: [Google]

157KB, 700x416px

>>61698729
>Marketing slides with bar graphs that appear to start at 0%
Jesus fuck AMD what are you doing? That's just wrong and misleading, everyone knows that this is how you do it.

>>61700381
funny how now the 290x shits on kepler

>>61698767
>he doesn t have a double xeon

eh nothing personal hobbo

File: 1457920886570.jpg (72KB, 752x816px) Image search: [Google]

72KB, 752x816px

>>61698729
Anon, you do know that they're lying, right?
Not even defending Jewtel, just think about it
They're both large companies with just one goal - make a fuckton of money
Of course they're both going to come up with bullshit benchmarks and schematics to jew you into buying their product
Buying a good processor is literally playing lottery, because you never get what you're promised

>>61700508
Yes this is true.
What amazes me is that from ground up Zen was made to look good on benches, remember muh blender times?
When the shit hit the streets performance was dogshit as usual.

Some cucks say it's not fair they didn't rerereviewd but oc3d done a rereview some time ago and bingo, zen performance was same shitty from release.

>>61700640
UMA DELICA

File: 1500379820327.png (744KB, 750x1268px) Image search: [Google]

744KB, 750x1268px

>>61698729

>>61700640
>blender is a benchmark and not a real application
>povray, premiere, handbrake, 7zip, veracrypt are also not real applications
kys cuck.

>>61698729
Assuming these numbers are true despite coming straight from AMD, that 7900X is holding up pretty well. I mean the 1950X has 60% more cores and only performs 14-55% better. Still a very good deal, especially with the extra PCIe and the proper soldering instead of thermal jizz.

>>61700749
can anyone explain why the monolith mesh is bad?

>>61700878
Shitty memory latency, consistently bad and overclocking it does almost nothing. RAM speed also does nothing.

>>61700849
>not using open cl/cuda for blender/adobe
kys cuck

>>61700508
If AMD was cooking this shit you'd see +25% on all those. It seems to line up pretty well with what we know about the two CPUs and platforms already.

>>61700887
>Shitty memory latency
To be fair, I doubt memory latency on Threadripper is going to be amazing when it has to hit memory channels residing on the other die. It definitely won't be as quick as accessing the local channels.

>>61700733
>>61700640
SOPA DE MACACO

>>61700931
https://www.servethehome.com/amd-epyc-infinity-fabric-latency-ddr4-2400-v-2666-a-snapshot/

>>61700878
>>61700887
Mesh is much more superior because it's interdie.
Infinity fabric is a latency joke and even worse it's speed is tied to memory speed.
With Intel mesh it's independent and can be changed without memory.

Infinity fabric latency penalties between dies is so bad it's better to just treat each h die as a whole different CPU. L3 latency is s shitty it's even slower than direct access.

Amd can only sustain performance on absolutely parallelizable loads like encoding. Performance drops when thread interdependence increases.

There is no magic with ryzen; it's a compromised arch designed to maintain somewhat linear performance on independent threads. Every other kind f load just shows the cheap and bulldozer like roots of this shit.

>>61700887
Memory latency isn't even important on high core counts, Mesh's real problem is fuckslow core to core latency.

>>61698729
> 30% faster overall
> consumes the same power
> same price
> HUUUUR DUUUURR AMD IS DEAD
fuck off Intel shills, Threadripper literally destroyed Intel.
> b-but muh 10-core performing only 30% slower than 16 core
Do you think cores scale perfectly? Fucking faggots, wait for the 18-core Skylake X firehouse and let's see if it can maintain decent clockspeeds without exploding

>>61700994
UNO SAL

>>61701029
Wtf amdrone? Uno sal means nothing in ptbr

>>61698903
ass

>>61701063
SOPA

>>61701063
QUE¿

>>61700951
Yes, that's exactly what I'm saying. Latency is higher and bandwidth is lower when hitting RAM on a different die (or an entirely different CPU, though that won't happen with Threadripper). These sort of results are only natural though.

>>61700878
mostly because IT WAS THE YIELDS
much simplifyed version cs anons dont get triggerd:
say on 1 wafer there are 10 imperfections == 10 broken dies
now if you have big ass dies that is a problem but if you cpu is made up of lots of smaler dies you loose much less from this since you will only have to replace 1 die (out of 2 or 4 dependinf if you are looking at tr or epyc) instead of the whole cpu

>>61701092
Each 2 channel is tied to a die with up to 2 full ccx. It's not unified ram bus like Intel, you gonna incur in memory extra latency with the secondary channel related to the other die.

Every thread the same mentally ill fuckers post the same memes.

What the fuck do I even come to this board for any more?

>>61701122
recycled memes

File: Wafer_die's_yield_model_(10-20-40mm)_-_Version_2_-_EN.png (163KB, 3000x1000px) Image search: [Google]

163KB, 3000x1000px

>>61701115
You need this chart

>>61701154
yes i do, thx anon

>>61701117
Yes, that's what I'm saying.

File: 1498329867841.jpg (30KB, 369x349px) Image search: [Google]

30KB, 369x349px

>>61701063

>>61701353
Give it back Enriqo

>>61698767
The (tm) gives it away, but this was only true before ryzen.

File: F12587!22-4862086_d.jpg (30KB, 300x300px) Image search: [Google]

30KB, 300x300px

>>61698903
rumor has it intel spent billions trying to make a propeler into a fan
eventually they found out that a jewish hat with a propeler is the optimal design

>>61698903
whats funny is that that fan wouldnt be enough anyways.

>>61700404
290x shits even on maxwell 2.0

File: 1497499213771.png (76KB, 326x250px) Image search: [Google]

76KB, 326x250px

>>61698946
EXCLUA ISSO

File: 1500350670495.jpg (55KB, 1280x720px) Image search: [Google]

55KB, 1280x720px

>>61701122
Just delid yourself

>>61698729
>no scores
ONE JOB

File: 1487101469436.png (164KB, 2044x2024px) Image search: [Google]

164KB, 2044x2024px

>>61701122
Because CPU threads are /g/'s equivalent of /v/'s console wars, or /o/'s C7 vs GTR threads.

>>61698729
Ok for a glued together cpu, I guess

>>61699218
DELICIA

File: 1427510764700.png (3MB, 3840x2160px) Image search: [Google]

3MB, 3840x2160px

>>61700404
>>61702657
>this delusion
A GTX980 has better or on par performance with the FuryX

>>61700994
95% scaling tho.
Explain this black magick.

>>61700899

>Being so uninformed when it comes to 3D rendering that you actually think GPU rendering and CPU rendering are interchangeable.

Go ask a major Arch firm why they don't swap from VRay to Redshift and drop $1,000,000 worth of CPUs for $250,000 worth of GPUs. Ask Pixar why Renderman is CPU only with GPU only allowed for Denoising.

>>61698729
Gaymers BTFO
>>>/pcbg/

>>61699107
UMA

>>61698767
t. Bettynho Zirigudum

File: 1495900565210.jpg (78KB, 600x600px) Image search: [Google]

78KB, 600x600px

>>61704980
>this delusion
Fury X fucking crushes the 980.

>>61699855
Don't forget BINGMESH :-----DDDDDDDDD

>>61698928
Underrated post

>>61698932

So what. Intel has the stronger cores, ok, we get it. We're talking about multicore CPUs, beginning at 16 cores. The AMD offering may have more cores and only thus offers more performance, but it runs cooler, with less energy and is priced the same. Additionally, I assume that mainboards on the AMD side will be cheaper.

You don't need to be a fanboy to understand that what Intel offers is worse than what AMD offers.

>>61698767
Oh, nice, now intel is new
>how to be poor but look rich
thing.