https://www.pcper.com/reviews/Proce ssors/AMD-Ryzen-and-Wind

File: W10_HighPerformance_3GHz_Ryzen_CR15_8T_SMT_ON_Affinity_0246.jpg (105KB, 672x622px) Image search: [Google]

105KB, 672x622px

pcper is full of shit

>>59346176
>>59346169
>Rajeeshsoft Winloo 10
Wew

>>even AMD does not believe the Windows 10 scheduler has anything at all to do with the problems they are investigating on gaming performance.

what actually we see on the article

In fact, though we are waiting for official comments we can attribute from AMD on the matter, I have been told from high knowledge individuals inside the company that even AMD does not believe the Windows 10 scheduler has anything at all to do with the problems they are investigating on gaming performance.

wew lad pcper has sources inside amd.. PCPER the site that said that 480 was 60% below 1060 and a 1050ti is a better choice than a 480

haha

File: 1488919452885.jpg (143KB, 960x686px) Image search: [Google]

143KB, 960x686px

>>59346092
>even AMD does not believe the Windows 10 scheduler has anything at all to do with the problems they are investigating on gaming performance

>>59346222
>PCPER the site that said that 480 was 60% below 1060

it was a real thing? how do people still trust them on anything?
I guess they have small audience, but still.

>>59346230
lmfao

>>59346176
So it's not even 8 "real cores"?

>>59346265
are you actually brain dead or pretending to be retarded?

>>59346250
its being known that they are literally a nvidia shill site

its like the legitreviews article from yesterday that it claims that amd rushed the ryzen launch and didnt tell the aibs that it will launch 2 weeks before...

we fucking knew it was going to launch on march 3 months ago and legit reviews claimed the motherboards manufacturers learned in mid feb...

seriously there is so much shilling around that i rather go and see any msm talking about trump

>>59346275
It says 4 cores right there. Maybe you can't read?

>>59346284
yes that is the problem amd treats every ccx as a cpu not as a core and windows is treating every ccx as a core thus having it as a 4c/8t


also not to self the ryzen patch is already on the 17xx built on the fast ring it will be available at 29-30 of march

>>59346284
Half of the cores are disabled on the pic. It has 8 cores and 16 threads normally.
With 4 cores enabled it has 8 threads.

>>59346250
PC hardware has just 2 out 3 contenders, it's the only industry with this small number of competitors, it's only logical their market tactics are so different from most of other industries with thousands of competitors.

I wouldn't be surprised if any of the companies would be engaging in malpractice, shilling and other illegal stuff, the reason being, if they by any chance manage to bankrupt their competitors, that's like 50% to 100% of their competition disappearing. That's fucking a lot, and that's also just how small this industry is

>>59346092

AMD needs to say something already. Microsoft themselves promoted win10 only adoption for Zen and Kaby Lake.

https://blogs.windows.com/business/2016/08/11/updates-to-silicon-support-policy-for-windows/

>As previously communicated earlier this year, future silicon platforms including Intel’s upcoming 7th Gen Intel Core (Kaby Lake) processor family and AMD’s 7th generation processors (e.g. Bristol Ridge) will only be supported on Windows 10, and all future silicon releases will require the latest release of Windows 10.

>>59346391
yeah, notice the timing of article? friday.
Nobody that actually in charge is working.

>>59346092
>Allyn Malventano
Is that the guy who works for the US Navy
They got to him. He's shilling Intel because muh government killswitch

>>59346432
it is also the guy that comes out of woodwork whenever AMD releases something major.

File: oh.jpg (127KB, 635x472px) Image search: [Google]

127KB, 635x472px

>>59346092
>AYYMD IS FINISHED & BANKRUPT
doesn't looks that bad

So ryzen is a bit of an early adopter situation right now, anyone who didn't expect this has pretty much brought it on themselves.

>>59346462
people just brain dead after same releases year after year
nobody remembers what nehalem was like or even sandy

>>59346492
I remember those motherboards with cancer.
It was horrible. You'd recommend a build to someone and they thought they'd save some shekels here and there and change the motherboard to one of those suspiciously cheap ones.
Then get angry with you when their SATA lanes slowly came to a halt.

>>59346462
yeah its like people forget how bad was the x99 platform launch and just shit on amd...

x99 chipset werent working

the cpus were overheating because of bios problems

the perfomance was even worst than ryzen now..

but no one really did care so much to point them SO MUCH

>>59346492
>people just brain dead
Current underage shitposting wasn't even in their diapers when Nehalem or SB launched, of course they don't know.
Not like BDW-E or Skylake didn't have a shitload of bugs at launch.

>>59346092
>>even AMD does not believe the Windows 10 scheduler has anything at all to do with the problems they are investigating on gaming performance.
Can't say I didn't tell you that the Windows 10 scheduler wasn't the problem.

>>59346556
god, AMD didn't say ANYTHING at all it's their ass pull "my dad works at nintendo"
MS said something though, which rarely ever happens at all meaning something is actually wrong

>>59346598
Is obvious that the scheduler in Windows is all fucked up when the same workload benches way worse there than on 7 and recent Linux kernel

Let's see. The focus on game benchmarks has been a little silly

I'd like more benchmarks of stuff like Adobe Premiere and Vegas to come out from several sources

Ryzen is showing good potential imo. But I would personally never adopt this early

>In this way at least, the CCX design of 8-core Ryzen CPUs appears to more closely emulate a 2-socket system. With that, it is possible for Windows to logically split the CCX modules via the Non-Uniform Memory Access (NUMA), but that would force everything not specifically coded to span NUMA nodes (all games, some media encoders, etc) to use only half of Ryzen.

Nope, not a windows fix at all...

This is why you shills fail.

>>59346726
and then he goes to explain that windows is clearly seeing the numa on amd

but later he cant understand why it doesnt load it properly

i mean this is GN level of shilling

Here's a list of software/firmware issues Anandtech managed to dig up only in the last week, there's probably other edge-case stuff we're missing.


Windows 10 RTC disliking 0.25x multipliers, causing timing issues,
Software not reading L3 properly (thinking each core has 8MB of L3, rather than 2MB/core),
Latency within a CCX being regular, but across CCX boundaries having limited bandwidth,
Static partitioning methods being used shows performance gains when SMT is disabled,
Ryzen showing performance gains with faster memory, more so than expected,
Gaming Performance, particularly towards 240 Hz gaming, is being questioned,
Microsoft’s scheduler not understanding the different CCX core-to-core latencies,
Windows not scheduling threads within a CCX before moving onto the next CCX,
Some motherboards having difficulty with DRAM compatibility,
Performance related EFIs being highly regularly near weekly since two weeks before launch.


http://www.anandtech.com/show/11170/the-amd-zen-and-ryzen-7-review-a-deep-dive-on-1800x-1700x-and-1700/23


Most of these are on Windows and motherboard vendor side, so unless MS implicitly states they won't fix them, AMD is fucked.

>>59346092
>vastly different performance when comparing windows 7 and windows 10
>has nothing to do with windows 10

>>59346726
you know that there is a ton of people that use dual xeons for everyday stuff, including gaming?

https://www.youtube.com/watch?v=vA0vAYCiDds&t

>>59346877
It's less than optimal.

>>59346889

It also doesn't shit the bed in one specific task.

>>59346877
Oh yes of course. Like there are people who play games on their laptop at sub 40 fps because that's the machine they have.

I think the point is that focusing as hard on gaming benchmarks as has been done suggests this is going to be the next big thing for gamers.

This doesn't seem to be the case at all. The architecture may well go on to do great things. But not necessarily for eSports

>>59346092
>Doesn't bother to include source for their "apps"
>Didn't test anything with process lasso
>Hardly realize what NUMA is and don't test any 2S intel configuration.

Yeah "gold" right there, buddy.

>>59346904
Compared to a 7700k? It does.

You're acting like Ryzen can't play games at all, which isn't true.

>>59346911
>But not necessarily for eSports
>gaming

you realize only 3% game on anything beyond i5 there? esports is toaster entertainment, wast majority of league/dota players didn't upgrade in the last 7 eyars

>>59346936
you make it out to be like it is the only truth
when in reality it's 9-11% worst case difference that's excluding clock for clock and cherrypicked benchmarks

>>59346936

No I'm acting like Ryzen shits the bed in one specific task because the software is still learning how to communicate with the hardware.

RAM speed is going to make Ryzen killer if they can get it to work.

>>59346533
To be fair, X99 is a premium platform so less people bought it, so less people to be vocal about bugs

Ryzen is a mainstream platform with more people buying it, so they are finding more bugs.

>>59346945
I just wanted to use a different word so as not to tire you out.
But yeah I realize there's no Witcher 3 or Ghost Recon in eSports.

>>59346969
2P boards are shit for gayming, this is a fact in both price/perf and absolute performance.

You're trying to tell me there's only 10% difference in applications that use 4 cores with a 50%+ clock difference? Get real, even you don't believe that shit.

>>59346343
Do you even know what you're talking about? Grocery has like 3 major competitors. There are like 3 telecoms companies that own everything. Even fast food is entirely operated as a set of subsidiaries with three major brands.

>>59346889
it runs at 2.6Ghz and 5 years old.
Now imagine it running at 3.7Ghz with modern RAM and pcie3.0.

>>59347010
>xeon at 3.7
Great dreams there, unless it's a 4 core Xeon.

>>59346990
amd OFFICIALLY state that the r7 is a HEDT and that is going after the broadwells

it was never a mainstream r5 and r3 will be

also

http://windowsreport.com/amd-ryzen-performance/

i guess pcper knows better than microsoft now

the update on which normally ryzen was going to get fixed as planned with the previous security update which it has be postponed for at least march 14.................

2/15/17: We will deliver updates as part of the planned March Update Tuesday, March 14, 2017.

>>59347017
ryzen = dual xeon software wise
remember what we started this talk first about?
it's latency is under 40ns when it's not messed with

>>59347010
>>59347017
Is this not the reason gaymers should be waiting for the R5 chips? Less cores better clocks or something?

>>59347007
Except you can buy groceries, watch TV and eat fast food not owned by those corps because all of those things are simple and have grassroot producers making and selling them.

CPUs, GPUs and other computer parts on the other hand are complex and cannot be just made by ANYONE, only the most dedicated corporations with billions of dollars of capital

That was a horrible comparison mate

>>59347030
Windows seeing it more like a dual CPU system than a single CPU system is ONE of the issues.
Honestly I'm more worried about Windows seeing each core having 8MB of cache and Windows not filling up a CCX first before moving to the other.

>>59347032

R5 is going to perform slightly worse than R7, it is going to vost significantly less.

They will be price to performance champs.

But this is about PCPer needing to be shut the fuck down.

>>59347056
>Windows seeing each core having 8MB of cache

Lol, doesn't that just mean the scheduler doesn't even know where to send threads and a lot of stuff ends up skipping the L3 and going straight into the DRAM because Windows can't do basic math.

>>59347032
they might offset clock difference with bigger core

nobody knows how they make them, they may just cut r7 in half and it would mean 8mb cache for 4c/8t cpu at 3.7Ghz for $150 it would be madness though
intel asks more than 1k for that kind of l3

>>59347104
>with bigger cache

>>59347096
that's what everyone is talking about

>>59347066
dude they wont shut up its a massive miss information campaign going on..
like this one
http://www.legitreviews.com/one-motherboard-maker-explains-why-amd-am4-boards-are-missing_192470

an ANONYMOUS MOTHERBOARD MANUFACTURERS says that amd in late december said they will launch in late Q2
but we knew since november that amd will launch on march...
>>59347096
it means that the date from the thread remains on the l3 and if windows throw the thread on another ccx then the cpu see the same data on l3 and pause the op for it to flush the previous data(assuming that the l3 isnt full)

>>59347135
There's 3 main Windows problems.

Misread caches.
Windows not populating a CCX(core complex) before moving to another CCX, inducing massive latencies(this is the dual xeon comparison)
.25x multiplier timing issues


There's other minor issues but that's what MS should be focusing on at the moment, the other problems are mostly BIOS or software specific(ie needs an update) related.

Thankfully these Microdick Wangblows issues seem to be consumer related, as soon as all cores are populated Ryzen's shine.

This spells well for Naples, but nobody will really be using Wangblows with Naples so it's a moot point.

>>59347151
Eh, its like when microsoft told everyone Vista's launch date.
They all had a laff and knew they had another 3-4 months past that date to actually get their shit together.

Then Vista actually came out on the date they said it would and drivers were unfinished and shit.

AMD said "it's coming out at the start of march"

"hahaha yeah sure m8" (we'll have till the end of Q2 lolol, don't rush yourself boys)

"Announcing Ryzen! It will be available first week of march!"

"ah, fuck"

>>59347192
More people run BSD with manycore CPUs than they do Windows, lmao

>>59347212
good thing it performs amazingly well on linux then

>>59346169

This graph shows Windows 10 using all cores to minimize SMT usage. What's wrong?

>>59346092
https://www.youtube.com/watch?v=40h4skxDkh4

Windows 10 does not respect the slow interconnect between 4 core clusters. This will be fixed I believe.

>>59347338
>What's wrong?
that's i72600
the graph lower is ryzen

>>59346092

I hope your mother becomes paralyzed in a car crash, /v/ermin.

>>59346534
People also tend to forget that first X99 mobos literally cooked the CPUs.

>>59346276
they got final silicon to bios for 3 weeks out from launch,

>>59347717
no we know that glo fo currently has a output of 62000 waffers and lisa told that they will launch with 1 million cpus stock

they didnt got it 3 weeks before the article is a total shit

>>59347943
62000 = 11160000 assuming 180 dies per wafer, even if half were fucked that's still over 5 million.

amd is under producing ryzen
amd is stockpiling ryzen and letting out a steady stream
or
amd didn't go into final production ramp till 0 hour and final silicon was delivered to make bioses 3 weeks before launch.

>>59348190

Now split those between 32, 16, 8, 6, 4, and hyperthreading.

And they are stockpiling. Server parts take precedence over desk top parts, while 6 and 4 core parts are being stockpiled for even larger number.

The motherboard makers undervalued Ryzen, and now are trying to cover their ass, and they have absolute shit designs compared to the Z270 chipset. Which is inexcusable.

>>59349454

Oh! And then we have whatever Xbox and Playstation are using and being allocated for.

>>59349454
>The motherboard makers undervalued Ryzen, and now are trying to cover their ass
Early reports suggested they were gun-ho about it.

Gigabyte and ASRock line looks decent, only really bad and overpriced ones are asus(per usual) and msi which the ones that seem got ought out in the open can't find a single MSI board on sale.
Biostar is usual jack of all trades.

The only good AMD CPUs are the AM1 ones if you need to build a low consumption computer for some reason.
For anything else, go Intel.

File: 1365289551051.jpg (126KB, 800x598px) Image search: [Google]

126KB, 800x598px

>>59349574

File: 1474945912873.jpg (396KB, 1920x1080px) Image search: [Google]

396KB, 1920x1080px

>>59349574
Have a comfy picture, calm your nerves.

>>59346230
Kekked

>>59346092
>Intel shills lie about AMD more
such news thanks for spreading it, good goyim

File: IT'S PIKATCHU!.jpg (54KB, 469x354px) Image search: [Google]

54KB, 469x354px

>>59349593

File: jew-mask.jpg (36KB, 688x456px) Image search: [Google]

36KB, 688x456px

>>59349868

File: 1310429820535.jpg (13KB, 300x300px) Image search: [Google]

13KB, 300x300px

>>59349868
FUCKING KEK

>>59346726
>Nope, not a windows fix at all...

At the very least it would make software use the core set which is intercommunicating quickly. Even 4C/8T is much for most software and <4 core processors won't be affected.

>>59346793
>AMD is NOT fucked.

Remember, they finally got the scheduler for Bulldozer right after years.

>>59347350

They disabled half of the cores for testing and ran a dual threaded program.

File: 43697.png (25KB, 550x420px) Image search: [Google]

25KB, 550x420px

>>59350051
And it barely made a difference.

http://www.anandtech.com/show/5448/the-bulldozer-scheduling-patch-tested/2

>>59347159
most workloads would do better with threads spread across CCXs for higher L3 availability.

very few workloads demand minimal IPC latencies outside of HFT and supercomputing simulations.

>>59350082

It barely made a difference for programs with many identical threads. If a program has n/2 threads and does a lot of FP math it will gain much.

>>59349517

>You mean we can sell more product? Fuck yeah let's do dis!

That's about as far as it went apparently. The designs are crap, the materials are crap, no attention to detail with slap-dash, glued on garbage. Don't even expect consistency when overclocking. And so far there were two motherboards on day zero that weren't absolute trash for RAM memory speed.

And now that Zen has met or exceeded every design goal in the HEDT space we have pure garbage to put an amazing piece of silicon on and it's somehow AMDs fault they weren't ready? That even their updated BIOS was fucked?

If the motherboards can't offer at least Z270 quality, whose fault is that?

They had zero faith in Zen.

>>59350247
Didn't AMD rush them to get it out on the market?

File: zen_bimodal.png (51KB, 1269x791px) Image search: [Google]

51KB, 1269x791px

>>59346092
lol, these fucking hacks.

Ryzen has clear bimodal performance distribution characteristics, which is 95%+ likely to be from the scheduler bouncing threads across CCXs and thrashing cache.

Zeppelin L2s are 2x the size of contemporary Intel cores and have only 1/2 the bandwidth to L3 since the don't have monster AVX units, so rebalancing/relocating even on the same CCX is more costly, and eager relocation back and forth between CCXs will absolutely murder it.

> t. HFT system architect

>>59346793
Those are just discussions Anandtech found, not things they themselves discovered. Anandtech is too incompetent to do proper CPU testing. Ian Cutress is a hack.

Hardware dot fr was one of the first sites to find some of the cache limitations so they get credit.

>>59348190
ofc they are stockpiling..

they have to launch r5 and r3 and naples and snowy owl

>>59349517
this is what i also think they didnt believe that amd will actually deliver and they thought it will be another bulldozer

they literally got by suprise

here is some irony for you

amd made athlons intel had to bribe everyone

amd made bulldozer fucked up despite having awesome boards..

amd make ryzen and gets fucked up by the boards and windows

>>59350247
don't they use same materials across board
z270=x370 with same naming scheme
some b350 even same quality as z270 of lower bracket
at least gigabyte ones look that way

there is also mention of intel in UEFI in a ton of them
not as fancy as x99 yet, but much much cheaper at the same time

>>59350376
there is also the true HEDT platform of amd the x399

>>59350392
>HEDT platform of amd the x399
was it mentioned somewhere? I missed it. rumored 16 core?

File: holy shit.png (32KB, 680x461px) Image search: [Google]

32KB, 680x461px

>>59346092
if anyone had any doubt this article was written by an overreaching imbecile...

>>59346275

are you? each CCX has a set of 4 cores, 8 threads. it's basically two 4/8 chips on 1 die connected by threading.

hence some limitations when CCX_1 needs to communicate to CCX_2 over that not-as-good-as-Intel threading.

>>59350403
16C32T
8CH DDR4
200W TDP
X399
and

32C64T
8CH DDR4 @ 4GMI LINK 100GB/S
240W tdp

>>59350404
That's glaring obvious. Intelfags are only clinging to it because of confirmation bias.

File: hillary-crying.jpg (54KB, 854x570px) Image search: [Google]

54KB, 854x570px

DELETEETEEE

MY AMD STOCK IS FALLING

HELP PLS DELETE

>>59350434
its almost at 15, again, already
try updating your memes

>>59350447
Spotted the AMD stock shill

Kill yourself you scamming fuck face

>>59350457
your intel shilling since march 2 has actually produce the exactly different outcome if you havent notice

get your penny and go back to a cave to fuck a squirrel or something

>>59350392

What is x399? Chipset for two sockets?

>>59350521
basicly the HEDT line of amd chiphell had an article on it
see >>59350425
they will be 1000-1200 for double the cores of a 6900k 8ch ram and 60w more

>>59346265
It's 8 real cores and 16 threads. The way they're connected is very different from Intel though.

File: AMD-Ryzen-Slide-1.jpg (234KB, 720x540px) Image search: [Google]

234KB, 720x540px

this is a ccx

>>59350280
No, that's a lie pushed by review sites
AMD had already announced the boards in January, most chipsets were already announced and out in October

>>59350668
seriously tho the amount of missinformation is getting at trump levels around ryzen..

amd kept everyone on dark and launched ryzen

ms didnt knew nothing about it
boards didnt knew ryzen was going to launch..

fucking seriously

>>59350144
The latency and bandwidth for cross-CCX communication is about 10x slower than for inside a single CCX.

So while yes, spreading threaded workloads across both CCXs would work well, jumping threads from one CCX to the other causes enormous performance hits, upwards of 30%.

In addition, bouncing threads around on the same CCX results in >>59350320

File: amdloo.png (316KB, 882x758px) Image search: [Google]

316KB, 882x758px

>>59346092

>>59350758
>The latency and bandwidth for cross-CCX communication is about 10x slower than for inside a single CCX.

sauce on this?
Zeppelin L2<-->L3 is 32B/clock (~115 GB/s at 3.6 GHz), and it's extraordinarily difficulty for me be believe that the cross-CCX link would be a mere 10-12 GB/s. That's literally not enough to support data striping across both on-die MCs.

I'm thinking about getting a 1600x to upgrade my i5 4440. Would that be a safe bet?

>>59350879
its 22gb/s the interconnection
but apparently the infinity wiring allows for 100gb/s

File: ryzen.jpg (951KB, 1920x845px) Image search: [Google]

951KB, 1920x845px

>>59346265
Nah, just like >>59346176 shows, it's two cores and 8 threads.

>>59350879
https://thetechaltar.com/amd-ryzen-clock-domains-detailed/

Cross-CCX bandwidth is (x/2)*32, where x is RAM speed. With 2666 MT/s RAM you have (2666/2)*32 = 42.6 GB/s. Since you can transfer data back and forth simultaneously, you have 42.6 * 2 = 85.3 GB/s.

I have a 1060 and it keeps crashing in 8.1 what the fuck I hate Microsoft now

>>59350546

Sounds like bullshit because x399 obviously is a name for chipset.

Will a 4 core Ryzen chip be a single CCX?

>>59351015
no shit sherlock platforms gets named after the chipset

>>59350758
it's about under 40ns inside and 140ns between if pcper soft to be believed, but they don't want to disclose source code so I wouldn't

>>59351034

It would be natural to suggest so because smaller die area yields better wafer usage.

>>59350989
it seemed really weird to me at first that AMD would scale cross-CCX bandwidth with memory and not core clocks until I considered that Naples will not likely be sold anywhere north of 3 GHz, so the local/remote bandwidth disparity will shrink where it matters most.

>>59351034
probably, but nobody knows what cache it will be
if they keep it at 8mb per ccx like on r7 it will destroy low end xeons
how much is it on cheap xeons? 10mb?

>>59351034
That's probable. You can probably expect the R3s to be a single CCX at least. They might be APUs with or without the graphics part disabled

>>59351034
no 1-CCX die has been announced, and it would be foolish to not use all 16 MB L3 if possible

>>59351147
I wonder if they could've made Zen transfer 64B/s instead. This thread (and AMD bashing) wouldn't have happened.

>>59351209
its 8mb in one ccx

they want to keep them cheap, don't know what is cheaper to make different mask or just split r7 in half

So a CCX is almost a self-contained CPU itself right? So it's kinda like MCM or multiprocessor setup but contained inside one die. Is it more space efficient this way or something?

>>59351245
64B/cycle*

>>59351265
that's where cheap 8 core price comes from
also it's easy and cheap to scale up

>>59350989
stupid question, would ryzen benefit more from low latency ram instead of ghz?

>>59351304
No

It will break the infinity fabric™

>>59351304
Latency doesn't seem to affect it, only clock.

>>59351319
why do people get worse results with higher clocked RAM then? or it's simply misreading aida64?
which didn't work at all day 1, probably still doesn't

>>59351339
mobo's are still fucked aren't they?

why didn't they beefup memory controller then if its so important?

File: amd-ryzen-ddr4-scaling-copy.jpg (77KB, 852x537px) Image search: [Google]

77KB, 852x537px

>>59351339
http://www.legitreviews.com/ddr4-memory-scaling-amd-am4-platform-best-memory-kit-amd-ryzen-cpus_192259/4

There are diminishing returns after 2933, but latency isn't that bad to be of influence. Ryzen isn't memory bandwidth starved, actually, increasing memory speed increases the IF clock and that's what improves performance. Only applications that test memory bandwidth will show the difference. Also this >>59351389

>>59347584
Your mom is dead and sucks cocks? Why anon, everyone already knows that

>>59351463
How does she suck cocks being dead? There will be just bones and maggots, do you want maggots in your dick?

>>59351470
perhaps he likes the maggots tingling his dick

>AYYMDPOORFAGS BUYERS REMORSE SPENDING MONEY ON UNPROVEN CPU WITH SEVERE ISSUES INSTEAD OF BUYING PROVEN, HIGHEST PERFORMANCE, LOWEST POWER CONSUMPTION INTEL CPUS

TOP KEK

File: 1466715625612.png (122KB, 973x540px) Image search: [Google]

122KB, 973x540px

based germans

>>59351584
It's time, anon

File: 1489257967719-2.png (121KB, 973x540px) Image search: [Google]

121KB, 973x540px

>>59351806
Oops, fixed

File: 1477085200020.gif (101KB, 500x513px) Image search: [Google]

101KB, 500x513px

>>59346092
Then why does Ryzen perform better on Windows 7?

And why is pcper.com the only one reporting this?

>>59351806
funny, but I mean that ryzen snapped at least 30% share of intel in germany
that's a lot

>>59351938
>but I mean that ryzen snapped at least 30% share of intel in germany
no, you're just retarded. its 30% of 5k people. Most people dont know what processor is in their computer and just buy a HP or Dell. And considering that they pretty much use Intel all the time, this poll means AMD has 1500-2k customers of some tiny blog in germany, many of which probably already had aging AMD chips.

>>59352072
germany is pc heaven
they know very well what is in their pc and who put it there, same way in eastern europe

>>59346092
Except the problem is not with SMT, but how Windows throws a thread from one CCX to another CCX

https://www.youtube.com/watch?v=40h4skxDkh4&feature=youtu.be&t=3m6s

>>59352148
>implying muslims know what is in their PC

>>59352188
No the problem is that AMD gimped their chips with a slow as shit link between the CCX units make these chips half the price of an Intel part. You got what you paid for and are now trying to blame Microsoft for your poor buying decisions.

File: 1483514253446.png (245KB, 800x612px) Image search: [Google]

245KB, 800x612px

>>59352268
>"What, this software can't be shit because it works great on my hardware!"
>*R7 literally on intels ass with better minimums and slightly worse averages*
>"NO, IT CAN'T GET ANY BETTER, DON'T OPTIMIZE, PLEASE!!!"

>>59351245
>>59351270
they easily could have, at a quantifiable cost to area, power, and possibly metal layer count.

32B/clk is honestly better if a processor doesn't have multiple 256b AVX units that can gulp down a cache line per cycle if not bottlenecked by the memory delivery subsystem.

https://www.youtube.com/watch?v=yyCXBfblYPQ
INTELGOYS ON SUICIDE WATCH

>>59352802
yes because i need to run 3 games at time with 720p with poor framerate. Seen to be the best amd can do

File: 1476908358045.gif (315KB, 234x159px) Image search: [Google]

315KB, 234x159px

>>59352961
>yfw Intel can run only ONE 720p game at the same framerates
KEK

File: 1408227430726.jpg (121KB, 768x720px) Image search: [Google]

121KB, 768x720px

>>59346230
>7zip 4 miles

>>59346391
>microshit is saying that you should be using windows 10
>p-please switch to win 10, we literally threw every tactic at you to get you to use it
What else is fucking new. You should really learn to block out the shills voices otherwise you will end up a good goy one day.

>>59346457
>prople do hype-price stock selling
>stock drops
>new investors buy in
>price rises

This is 100% normal and AMD is on track to hit at least $25 a share by the end of the year.

File: laugh.gif (2MB, 193x200px) Image search: [Google]

2MB, 193x200px

>>59346092

>pcper
>ryan shrout

File: uHnS0nn[1].png (9KB, 543x296px) Image search: [Google]

9KB, 543x296px

>>59346457
>Meanwhile

>nVidia dropped $30 in 3 months
NVIDIA IS FINISHED AND BANKRUPT

>>59352268
Except it's a software issue not a hardware one since W7 gets better performance than Win10 because it doesn't more threads like a retard (not saying W7 is using optimally too)
And Linux supposedly already got it fixed.

IT IS the fault of Microdicks the same way it was when Nehalem was relreased (and btw, Intel fucked that one even more than AMD did with Ryzen)

>>59351936
>Then why does Ryzen perform better on Windows 7?
Because everything that can run on windows 7 performs better than windows 10.
You know why?
BECAUSE WINDOWS 10 IS A PIECE OF SHIT THAT WAS MARKETED, SHILLED, AND PUSHED DOWN PEOPLES THROATS!
If intel didn't literally go mad with pushing it on to people it would be no more popular than win 8.

So... when is the R5 release date?

>>59353747
Q2

File: AMD-Ryzen-Die-Shot8C.jpg (867KB, 1000x787px) Image search: [Google]

867KB, 1000x787px

>>59350928
>>59346265

Thats 8 cores bro, if you don't know what the fuck you are looking at then don't post it. Attached picture shows you one CCX with 4 cores. The 1700/1700X/1800X has 2 CCX,

>>59354581
so I gather that the NE and SW blocks are the DDR4 controllers, but which die parts are likely to be the inter-die MCM links and which the PCIe/inter-socket links?

>>59354827
The MCM links are the two long ones above the left CCX.

>>59354941
Shouldn't there be three distinct MCM link controllers?
I can't fathom why they'd pay for an interposer then not bother supporting fully connected on-package meshes for Naples.

>>59354581
your picture clearly shows 4 cores, learn to read datasheets and get off my g

>>59355661
The actual Ryzen 7 core is below. Are you stupid?

>>59355580
I dont think anyone but AMD knows which are the PCI-E/Infinity Fabric Controllers and the Memory Controllers.

If you look at the blocks in the upper right and lower left, then along the bottom, you can see many similar structures (specifically near the center under core 4 left CCX) and under the L3 cache block for Core 8 Right CCX. The one near the bottom center could be dedicated lanes either for the IOH or talking to the M2 SSD.

The connections in the upper left also have a block of SRAM right next to it, whereas the blocks lower left upper right both dont have a fat chunk right next to it.

Oddly enough I think the upper left block is the IMC, and the upper right/lower left are PCI-E/InfinFabric controllers. If someone has a fully labeled diagram of the Ryzen die shot that would be greatly appreciated.

File: 0GSCodm.jpg (693KB, 2550x1400px) Image search: [Google]

693KB, 2550x1400px

>>59356449
The best I could find, still based on early leaks.

File: currentocandtemps.png (329KB, 738x496px) Image search: [Google]

329KB, 738x496px

I'm more a fan of the "underdog"

Recently bought a 470, satisfied with the results, I will be picking up an r5, likely the 4c/8t but I may spring for the 6c/12t.

Realisticly the r7 is currently just fine in games, is fantastic in other aspects of the compute world. Is cheaper then its competitor and may actually get faster with future "tweaks"

>>59356259
Are you blind? That's clearly 4 physical cores.

>>59357478
Let me help you.

>>59357445
Keep dreaming

File: 1489269866186.jpg (209KB, 1000x454px) Image search: [Google]

209KB, 1000x454px

>>59357523
Shit, forgot the pic.

>>59357445
I'm strongly thinking about the 6C, knowing AMD its going to be 6C till someone figures out 2 cores are locked and theirs a hack to unlock them.

>>59357585
I was initially planning on getting the 1600x but the more i hear about the way performance is jewed by cross ccx talk the more i'm worried that having less than the full complement of cores on each ccx will mean it'll never even come close to the relative performance of even an intel four core in gayming, unless the six cores have uneven numbers of cores in the ccx's.

File: 376_13.jpg (38KB, 511x319px) Image search: [Google]

38KB, 511x319px

>>59357531
Im sure I will keep dreaming, Im not aware of any benefit to stopping.

I have ofcorse read and watch multiple r7 reviews, and sense the r5 is the same thing with two less cores/4 less threads, Im sure ill be happy with it over my 8320, especially considering its about $150 less then its competitor.

>>59357585
That makes sense based on past experience, Im not sure if it can be expected with the lastest platform as with every newer gen they seem to have gotten away from being able to unlock cores.

The Lead pencil trick was my favorite

>>59357634
One less core per CCX will bottleneck less, and by then most issues will be fixed.

>>59357680
there will never be a 'fix' for cross-ccx latency, the bottleneck is the infinity fabric bandwidth. it's a hardware limitation

>>59357731
It's only a bottleneck in games, and even a handful of games already perform better with Ryzen. It's clearly a software issue.

>>59357634
Interchip bandwidth isn't an issue. Data fabric clock is half the DRAM clock, use a fast memory kit.
AMD had partners comes out with Ryzen branded 3200mhz-3600mhz kits for a reason.

>>59357731
Latency and bandwidth are not the same thing.
The only user facing issue is L3 access time in cycles when going from one CCX to another. If the OS properly addresses the hardware this will never be a real problem.
Faster clocks and signal rates decrease latency while increasing bandwidth.

>>59357800
It's clearly not be keep shilling and justifying, I don't care. I just feel bad for the kids that will come across your post and waste their money getting detective hardware.

>>59357634
Funny enough although ryzen benches a bit less in 1080p games, its consistently providing better frametime then intels comparison.

Leading to less/no Stutters/freezes vs gaming on intel.

Not really an opinion that everyone would rather prefer never dipping into the teen fps VS "only" hitting 200fps vs 230fps avg

>>59357875
>Funny enough although ryzen benches a bit less in 1080p games, its consistently providing better frametime then intels comparison.
Several reviews showed Ryzen having better frame times in Battlefield 1, but also showed it was worse in virtually all other titles tested.

Don't tell half truths.

File: intel shill.jpg (38KB, 688x456px) Image search: [Google]

38KB, 688x456px

>>59357870
I should have guessed.

>>59357523
its like numbered 1 to 8 and he still cant count

>>59357928
Either he can't see that the cores are mirrored or he's just a lame troll.

>>59357825
> If the OS properly addresses the hardware this will never be a real problem.

What you really mean is that if the software can manage to limit cross-ccx communication then the performance impact can be minimized, which may or may not be possible given that more and more games are managing to saturate cores with work, and given how game logic operates there may or may not be unfixable dependencies. you are right about latency and bandwidth. i'll wait for the benchmarks either way.

>>59357928
there are clearly 4 groups of 4 PHYSICAL processing units, clearly subdivided and reporting as 8 DISCRETE units to software but there is STILL on 4 PHYSICAL units there

File: ryzen-r7-1800x-aots.png (130KB, 806x843px) Image search: [Google]

130KB, 806x843px

>>59357907
>virtually all other titles tested.

AOTS

>>59357971
Games that showed big performance regressions from load balancing under Windows 10 don't have the same poor performance under Windows 7.

>>59346092

http://wccftech.com/amd-ryzen-4-core-benchmarks-intel-core-i7-7700k/

>simulated 4c/8t
>all other variables equal
>clock for clock performance
>on average a 4c/8t Zen @ 4GHz is 4.1% slower than a 7700K at the same clocks

That's pretty god damn impressive, considering how much of a turd bugler Bulldozer was.

>>59357995
Lame troll.

>>59346092

How could anyone believe this when we see w7 ryzen beating w10 ryzen?

File: ryzen-r7-1800x-mll.png (130KB, 806x925px) Image search: [Google]

130KB, 806x925px

>>59357997
Metro:LL

7600k 0.1% 25.7fps
1800x 0.1% 62fps

Far from half truths, Im not saying the ryzen offerings in their current config take the title for best gaming cpu in the world, ofcorse they arent.

But AMD has come a long way from fx series, their products are competing with intel now on all fronts. and in some aspects are cheaper/faster

>>59358000
Because windows 10 shifts threads across CCX's all the time. it's having a massive impact to be sure but it's not the whole story. there will still be cross ccx communication, it'll still be a performance bottleneck. maybe the 1600x will still be superior to say the 7700k despite that but again, i'll wait for benchmarks to confirm.

>>59358063
>comparing the i5-7600k to the 1800X

Too many retarded people here.

Show me where you can buy the 1800X for the price of a 7600K.
It's a deal then.

Otherwise kys

>>59358579
fuck off intel shill

>>59358639
fuck off amdtard

>>59358651
>>59358639
now kiss

File: 1489214401416.png (1MB, 2484x3052px) Image search: [Google]

1MB, 2484x3052px

>>59358651
Umad?

All this pointless arguing. You're not going to get any honest information out of these reviews.

The <tips> I were given in my dealings with Intel was to focus on the 7700K rather than the 6900, and on brand strength.
And in exchange Intel help put my kids through college, so why shouldn't I oblige?

>>59358940
>shill

>>59358940
oh my god what a scandal and not at all obvious bait

>>59346534
So accurate. Most of the people posting this "RYZEN DOA" shit are still on their first CPU (4460, 6400, 7400) in a pre-built their parents got them strictly for gaming.

File: 5b5a711e1cad26c75252a280c662f8f9.jpg (192KB, 1278x1678px) Image search: [Google]

192KB, 1278x1678px

Fuck this poojeet company and its retarded fans. Literally the poor nintendo of pc.

File: 1488855993719.png (557KB, 433x713px) Image search: [Google]

557KB, 433x713px

>>59359621
>poojeet
watch it.

>>59358063
>all the i5s getting beaten by an fx 8370 stock
Damn

>>59350376
bulldozer got fucked hard by windows wayyy worse than ryzen is currently.

>>59354827
>>59354941
>>59356449
The top right and bottom left blocks are actually almost certainly each the PCIe x16 controllers, which double as inter-socket Infinity Fabric links on 2S Naples.

I would actually guess that the two long rows on the top left are the DDR4 controllers. The 72 bidirectional data lanes (these are ECC controllers) match with the 9 large repeating blocks, and there are a smaller number of smaller transceiver lanes that probably match the send-only control lanes of each channel.

The MCM transceivers should be substantially smaller per bit lane than the DDR4 PHYs and especially the PCIe ones, but it's hard for me to guess which blocks they might reside in without knowing the bus widths. My preliminary guess is actually the really thin strips on the bottom left-center, bottom right-center, and maybe even top right-center.

File: battlefield-1.png (11KB, 800x401px) Image search: [Google]

11KB, 800x401px

if ryzen is such garbage and working as intended could someone please tell me why when you take a 1800x and turn it into a 4 core 8 thread cpu magically its in par with the 7700k in games?

>>59352961
heh, Wildlands seems nice

>>59350989
> (x/2)*32
So, just x*16 then.

>>59351248
Re-using dies with defective cores is cheaper than throwing them away.

>>59346793
>Ryzen showing performance gains with faster memory, more so than expected,

And then it's released with RAM-timing-issues and only supporting 2133MHz with all four slots populated.

pottery.

>>59361630
that's motherboard issues
with beta bios my times went up 600mhz already and the bios is still getting worked on

The worst part about Ryzen is that you can't even run memory at greater than 2933MHz. Other wise, your PCIe gets gimped down to PCIe 2.0. That's if you can even get your system to boot with greater than 2933MHz (only a handful of boards support it). How fucking lame is that???

>>59361691
nice meme

>>59361743

not a meme. go check it out for yourself.

nice try though, Lisa.

>>59361691
This is just pathetic shitposting.
You do not have to tweak base clock to get high speed memory working.

>>59361691
that's a lie
OC guy that talked about it is wrong, he doesn't know shit about how ryzen works
read what Stilt explains about whole RAM thing, it's all inside CPU => motherboards suck

>>59350710

Are you retarded. All of /g/ knew of march 2 for weeks. Are we somehow omniscient or do you have another explanation for how this board knew as the only group when Ryzen was going to drop.

>>59361798
show me any ryzen system with memory running at 3200mhz+ with out blck adjusted.

>>59361836
you do know that blck greater than 104.8mhz gimps pie down to 2.0, right? may not matter to most but the fact that you have to put up with such a compromise in 2017 is pathetic.

>>59362036
here
>forgot
this, read it again, it's sarcasm

>>59362041
107mhz is the cut off of PCI-E 3.0 mode, you lying little child.
You can't even get basic facts right with your infantile shitposting.

>>59362041
pie == pcie. damn autocorrect

>>59362041
>you do know that blck greater than 104.8mhz gimps pie down to 2.0
it doesn't

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

read whole thing, focus on stilt responses to make it faster, people made wrong conclusions

>>59362041
do you have the autism, son?

>>59362050

I'll call your bluff. show me proof.

>>59362055

go to sleep, Lisa.

>>59362064
>I'll just talk out of my ass like a disphit lying little kid
>try to argue with more informed people at the drop of a hat
I want you to prove you're actually 18 years of age or older. I don't believe for a second that you actually are.
You're a tech illiterate retarded little child.

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/
Overclocking the base clock (BCLK) on AM4 platform is possible, however generally not recommended. This is due to its frequency relations with other interfaces, such as the PCIe. Unlike with Intel's more recent CPUs, there is no asynchronous mode (straps / gears) available, which would allow stepping down the PCIe frequency at certain intervals. The PCIe frequency relation is fixed and therefore it increases at the same rate with the BCLK. Gen. 3 operation can generally be sustained up to ~107MHz frequency and higher speeds will usually require forcing the links to either Gen. 2 or to Gen. 1 modes.

>>59346092
>intel damage control

your tears are salty my child :^)

>>59362112

so base clock of 107mhz is the max limit for overclocking on ryzen platform unless we settle for PCIe 2.0?

>>59346793
Ryzen's problem isn't physical vs logical core handling, its bigger problem is windows scheduler is unaware that it should be treating each core complex as non-uniform memory, and thus should be allocating tasks using similar cache calls to a single core complex. What happens at the moment is that doesn't occur, causing cache misses. The addition of SMT just adds a higher chance of causing a cache miss.

>>59362112

https://www.reddit.com/r/Amd/comments/5yrihg/best_memory_right_now_for_ryzen_gskill_3200/deshpyc/

>>59362141
You're not going to overclock it past 4.1ghz on all 8 cores on an air cooler, and water doesn't give you much headroom either, so its a moot point.

File: 618x613x4000-cine-CPUZ-1024x1016.jpg.pagespeed.ic.r3m7HsB3EJ.jpg (79KB, 618x613px) Image search: [Google]

79KB, 618x613px

>>59362182
>this random reddit comment is totally more valid than a technical analysis from a well known source
As I said, I want you to prove you're actually age 18 or older. I would bet good money that you're not.
You're a stupid little child.

http://www.kitguru.net/components/cpu/luke-hill/amd-ryzen-7-1700x-cpu-review/2/

Default 100mhz base clock.
3200mhz CL14 DDR4 with T1 command rate.


Go slit your wrists, kid.

>Intel did nothing to improve the performance of their CPUs for ~7 years
>AMD releases a new "Intel" killer, it's barely manages to compete with it

This is really sad and I have no idea what AMD wants to proof here. Maybe they should stick to making CPU/GPUs for 1080p/(barely)30fps consoles.

>>59362214
>no idea what AMD wants to proof here
that your English is worse than mine

File: tough-luck.png (122KB, 485x365px) Image search: [Google]

122KB, 485x365px

>>59362248

>>59362211

Thanks for doing the homework for me. I was going to ask if 3200mhz+ memory would gimps pcie but questions usually don't get answered on /g/ so I started shit talking instead to get an fan boy to defend AMD thus providing me with the info I was wanting. thanks man!

>>59362302
>pathetic shitposting little child shill tries to save face after getting monumentally blown the fuck out
Kill yourself

>>59362312

sorry I used you. :-)

>>59362322
>>59362302
DEVILISH

>>59362302
Actually really clever.

>>59350879
Speed is not that relevant. Its the latency. In the most pessimistic case a shift of a thread between CCXes will require setting up a memory barrier, which stops all cores, and moving data through the physical memory. This will cost millions of cpu cycles.

In the best case, you merely need to copy contents of a cache between CCXes and reschedule threads. This alone costs thousands of cpu cycles.

File: Screenshot from 2017-03-05 11-10-11.png (63KB, 543x679px) Image search: [Google]

63KB, 543x679px

Good thing I don't use windows.

>>59350521
If that was the case, I would immedietly buy it, along with dual 1700s and a bucket of ram.

>>59363368
You don't need to compile your kernel on Windows.

>>59363373
You should know that the Chinks on Chiphell have been laughing at all the stupid foreigners who took their intentionally started rumor seriously.
The X399 "Eel" chipset is not real.

>>59358702

Lmao no replies

B-BUT FAKESITE SHILL

>>59363368
Ryzen doesn't have issues with multithreaded stuff even on WIndows, it's just more pronounced on Linux.

>>59360335
>Downclocked 7700k
but I get your drift.

>>59363377
or on linux...
it's a benchmark, you tard

>>59363552
S-SHUT UP STUPID GOYIM

Don't like Ryzen? Buy Intel! Do it!

Want to try something new? Buy Ryzen!

File: BabbyLake.png (157KB, 1110x375px) Image search: [Google]

157KB, 1110x375px

Today I will remind them.

>>59364343
What was there to develop? Wait for 14nm to mature so they can clock it higher?
Seems like most of the job was done by the fab folks,

>>59346492
I remember going through a gigabyte and asus Z67 before i settled on an asrock that actually worked well, seems like the ryzen launch is no different

I look forward to the 1600X launching with mature boards and software and 'magically' matching the 7700K in games and destroying it everywhere else

>>59364518
it will match 6900K that's for sure, and i'd take 6900k for 400 with motherboard for 100 that would last me until 2020

>>59364456
Unfortunally its already stated by intel that their primary concern isnt raw performance but better wattage. You end not seeing this on flagship desktop cpus, but rather on server and ultra low tdp markets like ultrabooks.

I really wish AMD to bit intel sales so they improve their desktop archs

>>59353578
>If intel didn't literally go mad with pushing it on to people
Ehh?
OEM and Microsoft pushed it.
What's Intel has to do with it?

>>59364911
>so they improve their desktop archs
I am sure if they could, they would.
The 7700K runs on 4.3 stock... just to have enough performance.

Don't get me wrong, it dominates the market, kills the 1800X, and so on. But still, that's the strongest Intel has on single thread.

It's not like they hold back, but this is the current state of the market.

>>59365080
I wonder what cofeelake will clock like. I don't think they can push 6core higher than 4Ghz on 14nm.
And if they can't it's going to be 6 core skylake+2%. I don't know what hope is left for high performance CPU in next three years, zen+ is only thing that can grow but I wouldn't hold my breath.

>>59346092
>the problems they are investigating on gaming performance
Corporate speak for "we gave a team of ten guys an impossible task and will blame everything on them when they can't speed up our shitty processors in games"

File: pepe blocky.png (2KB, 215x217px) Image search: [Google]

2KB, 215x217px

>>59346176
>>59350928
>>59354581
>>59355661
>>59356259
>>59357478
>>59357523
>>59357539
>it's another "8 cores but really just 2 cores" episode
you guys got played

>>59365146
Around 4ghz turbo on one or two cores at most.

The 7700k already draws more wattage and has a higher TDP than the 8 core Ryzen CPUs in most task.
They can't just have 50% higher power draw when they're already borderline housefires.

>>59365225
ironic how TDP meme backfired on them when they are the ones that pushed it hard last 7 years
even nvidia learned marketing form them on this topic

>>59346877
Impressive. I've been considering 2x E5-2670 for VM things.

>>59365260

Power draw/TDP ceases to matter the instant Intel or Nvidia lose the edge against AMD. You can see it with ryzen and we saw it with the GTX 480.

For low power deployments (30w and below) ryzen is a fucking monster even with a full 8c/16t.

File: image.jpg (18KB, 200x267px) Image search: [Google]

18KB, 200x267px

Could you possibly shill any harder?

>this thread

>>59350320
>>59350758
can you bottom-line the real-world implications of these graphs for me please?

>>59365385
whole thread is about it

>>59365385

Once windows stops spamming threads across the CCX performance will increase because the latency hit will be reduced. This is exactly the same shit MS dealt with for the Core 2 quad chips and has seemingly been ripped out of the win10 scheduler because win7 handles it fine.

>>59365404
Then why do we only see a 5% performance boost in games on Win 7 if it handles it fine?

>>59365427
more like 15%

>>59365260
What backfired on them is getting games optimized for 4 cores or 4c/8t at most. If they were better optimized for 6 or 8, AMD would have much better competed.

The 7700k is pretty much the limit of 4c/8t performance yet it already CPU bottlenecks a great number of games at under 144fps with 98-100% usage on all 8 threads. They can't magically make it any faster, they've hit a wall going all-in on this when they should have made 6 cores mainstream at least 2 or 3 years ago.

Now they're going to finally come out with a 6core that'll be worse in many games as their mainstream desktop gaming processor when it's almost too late.

>>59365427

Windows (all flavours) still doesn't like the very small increments ryzen clocks in - 25mhz is a very low amount when most chips clocks up and down 100mhz at a time.

File: thanks-amd.png (142KB, 390x360px) Image search: [Google]

142KB, 390x360px

Nothing better than AMDfag tears

AMDfags never learn, they just keep eating AMD shit over and over again

File: wasting the prime of your life Screen Shot 2017-03-11 at 9.14.28 AM.png (74KB, 650x310px) Image search: [Google]

74KB, 650x310px

>>59365500

>>59363704
This is how you compare IPC, you retarded Shintel cockmongler.

>>59365474
6800K is pretty good for games, nowhere near as stable as 6900K though
so AMD guning 8c for games was right path, raw pwoer is tehre, games don't like it compared to 6900K for reasons described in this thread
all kaby got for it is clocks and price which it's losing to 1700

really optimization happened already in 7 games out of 10
if ryzen doesn't get fixed though, that's another story

File: 1daamdang.png (72KB, 601x830px) Image search: [Google]

72KB, 601x830px

>>59365567
Umm no, just no

The 6800K and the 6900K both get outperformed by a fucking 7700K that only costs $340

>>59365534
who cares? we know the IPC is within 5%, all that matters is that Zen can't scale above 4 GHz—full stop.

>>59365602
>we know the IPC is within 5%

Before the benchmarks we didn't. Are you retarded?

>all that matters is that Zen can't scale above 4 GHz
>hurr durr why can't an 8 core CPU clock like a 4 core one

Yes, you're retarded.

>>59365567
Yeah the 6800k is better in some games, worse in others.

a 4ghz turbo 6 core is still going to be worse than a 4.3ghz 4 core with lower L3 latency in many games because of how Intel encouraged developers not to highly multithread for so long. Only recently at engines spreading out object transforms and physics and such over potentially dozens of threads to scale across more hardware more cleanly.

The power with Ryzen is there. Games that get 98-100% usage on all 8 threads with a 7700k never use more than like 60-65% on average across 8 threads of Ryzen. The problem is Ryzen isn't currently utilized in games as well as it could.

>>59365602

>IPC doesn't matter, only clockspeed

And so we check another one off the list. Current tests that no longer count (in no particular order)

>cinebench multi-thread
>cinebench single thread
>blender
>power consumption
>gaming at anything above 1080p

>>59365636
The reason it can't go much above 4-4.1ghz is the manufacturing process for the silicon, not because it's an 8core.
4core Ryzens are also not going to go above 4.1-4.2 at reasonable TDP.

Will have to wait for q1 2018, most likely, for the updated process that's more geared for desktop than servers which will probably clock at around or at least 10% higher.

But 4ghz on 8 cores with all that cache is fucking good. Its single core performance is great, it's just not quite as high as the 7700ks in many regards.

>>59365636
>Before the benchmarks we didn't. Are you retarded?
jesus christ dude just fuck off, you are not special or unique or even more intelligent than the average person

>>59365760
>this much projecting

Stay mad, retard. Go back to shilling for Intel

>>59365686

>at reasonable TDP.

Once motherboards sort their shit out I'd be interested to see some heavy overclocking results. Namely what sort of temperatures you are looking at under torture loads on big air and 240mm AIO and the required voltage to get there. After all the 9590 is considered a nuclear inferno and that fucker can be air cooled even when clocked to over 5ghz under torture loads (its not easy but it can be done).

I've got a 1700 on order and give zero fucks about TDP so i'm itching to see where my chip stops with a huge dual tower heatsink paired with some angry fans working to keep it under control. Hell for laughs I once bolted a 3900rpm 38mm thick delta fan to said heatsink and let it run 100% while torturing my 8320e at 4.7ghz and the sound was horrific but cpu temps dropped nearly 10c (alas my motherboard can't take anything over 4.7ghz at 1.428v without the vrms melting). I'm going to have a lot of fun overclocking.

>>59365937
You can't compare the old 28nm gloflo process to Samsung 14nm LPP.