Tigerlake has ~40% IPC gain over coffeelake

>~40% IPC gain
Sure it does. Just like Intel 10nm launched in 2015, (and 2016, and 2017, and 2018) and is going to launch again near the end of 2019, right?
Also
>userbenchmark
Is this Intel Corporations' "secret sauce"?
L M A O
M
A
O

J-Just wait, Intbros... we'll be b-back on top soon!

Attached: Everything's fine.png (653x726, 84K)

performance = instructions * ipc^-1 * f^-1

it's scoring as well as a 9900k at that frequency, which is pretty impressive.

>reduce multithread in favor of single thread
>mid 2019
>not paid shills

>((((((((((((((userbenchmark))))))))))))))

you know that companies bribe websites to cook up the numbers in order to boost sales for new products, right?

They changed the weights on the tests in their 'gamer' index to favor single core. It doesn't change any of the tests themselves.

>let's make fun of the clock frequencies of engineering samples
That's pretty dumb, senpai-tachi.

Attached: chiyo-4.png (251x225, 77K)

Yeah they changed it to strongly favor single core in 2019 when almost everything is getting more and more multithreaded, with even games properly using 6 and even 8 cores. Hmmmmmmmmmmmmmm....................

Attached: Userbench-Old-and-New.png (1600x918, 186K)

AMD was thinking way into the future with bulldozer. More cores is not a meme anymore, but unfortunately this shift happened later than expected.

The point being that processors are getting "multithreaded" even faster than software is, which, they would argue, revealed that their weighting was wrong.

All of these benchmarks have biases it seems. Geekbench heavily favors Apple's ARM implementation for example.

>processors are getting "multithreaded" even faster than software is
That's because of retarded pajeet programmers.

So what retard, it still shows you the points each CPU scores

Why change it then? HMMMMMMMMMMMMM

No, it's because machine language is inherently sequential, and multicore parallelism is a relatively unnatural way to increase parallelism, given the large gap of granularity between ILP and TLP. Processors need to get more parallel at the machine-language level. Not saying that dataflow architectures are the future, but...

>you can only run one application at once

let me know when they can use this """architecture""" on something better than a 3w laptop chip at 0.8ghz kek

Said who?

Well, laugh's on you since that's where like 90% of sales are.

That's why they have a quad core test. They could have easily increased the weight of that or added an octa core test, but instead they went full single core just after Zen2 was released, really suspicious.

THIS KILLS THE AYYMD RYPOO HOUSEFIRES GARBAGE

My 7600K currently sits at 1.2 GHz, it only goes up to 4.2 when I'm gayming.

Just because your Poozen needs 4 GHz just to move your mouse cursor on the screen doesn't mean all CPUs work that way.

>Intel has to pay userbenchmark to change their methodology in order to keep Intel ahead because Zen 2 is so good
Keep seething, faggot.

Not really denying that that part is really suspicious. If I were to play the devil's advocate to defend them, I'd say that, core configurations being somewhat similar prior to the Zen 2 launch, the exact weighting might not have made a great difference anyway and that when Zen 2's much higher multicore scores caused them to reevaulate it, they honestly discovered that single-core scores might have mattered more than they previously though.

All reviewers have completely thrashed intelbenchmark

First, that's not Tigerlake, since that's not even close to 40% IPC gain over CoffeLake.
Second, this is "unofficial leak" at best and Intel has been using tactics like this for 20 years now, where they release fake leaks with far better results than the real product, while denying they had anything to do with it, it's free marketing and scummy as Hell.

I can't wait for Intel to innovate again though.

Maybe he just doesn't like stutters.

Enjoy unsmooth desktop performance intcel fag

Might be a 60Hzlet as well, who knows.

Keep seething Francois

That's 30 Hz more than any AMDjeet has ever achieved.

Quite a weak "comeback".

Attached: 1561563554391.png (898x882, 108K)

Attached: ACO_1080p.png (1405x1517, 82K)

Attached: Premiere.png (1405x1516, 75K)

AMD and Intel shills should all fuck off and die

>multicore parallelism is a relatively unnatural way to increase parallelism
Might as well say any syscalls are unnatural then.

Why would you say that? Syscalls fit quite naturally into the sequential model.

How is talking to the OS to do threading any more unnatural than talking to the OS to do anything else?

Anyways, that scoring system never made sense in the first place. Their gaming/desktop/workstation scores seem more reasonable, maybe they should just remove the global score and allow the user to select what's more important for him.

>Their gaming/desktop/workstation scores seem more reasonable
My understanding is that the scoring we're all talking about is the "gaming" score.

I don't fucking care
where is the non-skylake PHYSICAL desktop silicon?
I'll take anything at this point

Talking to the OS to do threading isn't necessarily unnatural in itself. What I'm saying is that multithreading in itself (whether or not you need to speak to the OS to do it) is a fairly unnatural way to achieve parallelism, because it's so very coarsely granular.

>U3E1, 1 CPU, 4 cores, 8 threads
>Base clock 1.2 GHz, turbo 3.6 GHz (avg)
No one cares faggot, come back when the same leak happens for the 10990k.

>is a fairly unnatural way to achieve parallelism, because it's so very coarsely granular
You literally just copy-paste all the machinery necessary to run a thread. What's more natural than copy-pasting?

not having to access L3 for everything all the time

Yes, it is simple and natural from the hardware perspective, but not from the software perspective.

Most likely won't arrive until 2021 when they discontinue all 14 nm production. Desktop cpus are their lowest priority.

This
The evil antisemitic amd nazis just can't compete with our TDP and security hole numbers.

Attached: 1537419972674.png (1037x311, 340K)

>I'll take anything at this point
AMD's got your back, bro.

There is speculation that Golden Cove will be coming on 14 nm.

Not really, they changed the "effective speed" score, that's why it caused so much drama.

I thought Intel was completely against backporting 10nm designs to 14nm.

I mean, that being said, it's not like strong single-core performance isn't highly relevant for "effective speed", especially if you like at stuff like UI responsiveness, which is virtually always completely single-threaded. Virtually every single JavaShit web page is also single-threaded by the nature of the language.

True, but there has been quotes from Intel execs lately about decoupling architecture from process.

((("""user"""benchmark)))

Not entirely true. JS in browsers has workers.

Javjeets manage to make their code run like shit on any CPU anyways.

This but unironically, I wouldn't be surprised if they gained so much by compromising security even further.

userbenchmark isn't bad but you have to compare the partial scores (1c, 2c, 4c, multi etc.) and not the intel paid final score.

niiice, my old 3570k is showing its age and I'm currently eyeing that 3700x for my next rig, gaming wise it's the same as 9700k but just cheaper.

>I'm currently eyeing that 3700x
Dude just wait for Zen 3

>zen 3
but when?