What in the fuck

What in the fuck

How


techpowerup.com/reviews/AMD/Ryzen_7_1800X

Here's the TPU review straight off the press

Other urls found in this thread:

hardwarecanucks.com/forum/hardware-canucks-reviews/74814-amd-ryzen-7-1800x-performance-review.html
tomshardware.com/reviews/intel-kaby-lake-core-i7-7700k-i7-7700-i5-7600k-i5-7600,4870-8.html
techpowerup.com/230446/amds-ryzen-chips-10-smaller-than-comparable-intel-skylake-dies
youtube.com/watch?v=ylvdSnEbL50
youtube.com/watch?v=6laL-_hiAK0
twitter.com/SFWRedditVideos

Some more neat stuff

>CPU that gets the most work done also uses the most power.

Looks pretty normal to me.

remember to subtract 20c to get real temps.

imo the 1800x should be left at default. it cant oc for shit.

It's using 28% more power for 10% more performance at best.

wtf? 400 mhz more and zero performance gain?

this

the i7 is probably sitting at 99% load and the 1800X at ~70%

In gaming, XFR boosts a core to 4.11GHz and it's disabled when OCing.

actually this confirms it

>Corsair 16 GB Vengeance LPX DDR4
>@ 2133 MHz 15-16-16-35
>@ 2666 MHz 16-16-16-36

WOW A FUCKING I7 WAY BETTER THAN RYZEN NEED 20 MORE WATTS TO RUN

INTEL IS FINISHED

Yeah, way better.

>testing the power consumption of the least efficient AMD 14nm processor on the market instead of its way more efficient little brother that's 5% slower

How did the performance gap drop from 20%+ to 10% in 2 weeks?

Early reviews likely didn't disable stuff like core parking which seriously fucks with Ryzen's performance.

TPU wasn't using ultra buggy ASUS motherboards, also they disabled HPET and disabled balanced mode that parked cores nilly willy.

If only they managed to get higher clocked memory working, it would look even better.

>b-b-but total war...
>w-what about fallout 4...
>it's not fair

>only 11% faster in reslet resolutions with 10% higher single core clockspeed and 20% higher all core clockspeed on top of 10% higher IPC in applications that use 4 cores

Wow Intel, whatever the fuck are you planning next year?

Would have liked to see one of Intel's 8 cores in the review for comparison.

There's shitloads of 6900k comparisons out there.

hardwarecanucks.com/forum/hardware-canucks-reviews/74814-amd-ryzen-7-1800x-performance-review.html

>Wow Intel, whatever the fuck are you planning next year?

MOAR CORZ on skylake v3
And I guess dual core cannonlakes for laptops lmao

That's why anyone smart will get a 1700 and OC it.
Or a 1700X if you want to squeze another 200Mhz more from the OC

Dunno if that's worth $100

Yeah but it only makes sense to compare them with TPU's results if TPU is the one doing the tests. We don't know if TPU or if the other reviewer's methodology is flawed so the result may vary.

'This is epic. We're assuming you've sifted through our game-test results before seeing this page, and so you'll find that the gaming power draw of the 8-core Ryzen makes Intel's quad-core i7-7700K look bad. The power-draw is as much as 30W lesser!

Ryzen is hands down the most energy-efficient performance CPU AMD ever made, and easily outclasses Intel's 14 nm "leadership." Good show. '

Once memory speeds are sorted out Ryzen will match and sometimes beat the 7700K hands down (including games which will only make better use of Ryzens architecture) whilst using less power.

We told you it was Windows, motherboard BIOS's and development of current games and apps holding it back

Ryzen seems to heavily favor high speed memory. The memory speeds may be almost as important as CPU clock speed for Ryzen performance wise.

TPU's admin hates AMD so quite surprised seeing him give it a positive review.

Yeah that's a real shame since memory is really expensive right now as it is. But maybe there is more to be had from BIOS optimizations.

Which is why this is probably the best review so far and to be trusted. He notes how the ecosystem is shit though and how AMD should have fucking got their act together with mobo manu's.

Don't worry, you'll still see the same shills and retards post the day one benchmarks months from now.

Day1 benchmarks on crashy asus boards are all that matter.

Holy shit that's insane, or are the 7700k's just overclocked housefires?

every entertainment tech website rushed out their Ryzen "review" so they can cash in on the hype

still waiting on anandtech's ryzen gaming benches...

There have to be more differences, the 1700x TDP is 30 higher. I'm still uncertain which would be the best option.

>one cpu is being used 80-100%
>the other cpu is being used 40-50%
Seems to make sense that the cpu under less stress is using less power.
Ryzen's poor gaming performance is down to the current state of optimisation and will get better with time as more of the cpu is being used.

link to this video?

This is likely measured from the wall.
TPU testing methodology is really slipping.

The 1800X stock will pull 110w under a 100% AVX load.
The i7 7700K will pull about 95w.

In a gaming workload the 1800X will pull just under 60w.
The i7 7700k will pull about 75w.

Yeah they only tested with 2133 and 2666mhz RAM.
LegitReviews showed a nearly 8% increase in game performance going from 2400mhz to 3200mhz RAM.
AMD also has a similar result up on their own site.

>con
> Lacks integrated graphics

Oh come on W1zard

More like why the fuck is there so much load on the SMT threads instead of the physical cores? Scheduler working fine my ass.

Figures for the i7 7700k here:
tomshardware.com/reviews/intel-kaby-lake-core-i7-7700k-i7-7700-i5-7600k-i5-7600,4870-8.html

Really shows the efficiency of AMD's new chip.

>Ryzen 134.5 FPS
>I7 130.5 FPS
>Ryzen's poor gaming performance
What.

SenseMI. We haven't seen everything out of it yet.

So, AMD shills. Can PSP be disabled yet?

> AMD should have fucking got their act together with mobo manu's.
Nothing but minor teething issues of a brand new platform.
Intel had full on recall of some things when they first launched Sandy Bridge due to bad chipsets. AMD's only problem in regards to motherboards has been vendors not fully sorting out their BIOS. Thats such a tiny thing to fix its barely a road bump. Though it did hurt them somewhat in reviews, they'll fully recover from it.

Yeah my brother had to return his mobo for a new one to fix the SATA 3 issue. Shit happens. Intel cucks have conveniently vacant memories when it comes to that kinda thing (or were too underage to know at the time)

Maybe it's reading it as 16 cores and simply throws threads wherever it can.

Not yet. AMD is a big company there's probably a lot of people behind the scenes that they have to consult about the feasibility of it.

I don't expect them to allow it but they might ship chips without it altogether which is a good compromise I think.

woah, TPU also known as The ultimate Taiwan shilling site gave AMD decent review

Microsoft hasn't fixed their scheduler.

Same deal as with Bulldozer's early problems.

Its not all that great. The conclusion includes lack of integrated graphics as a negative. The whole write up has minor quips worked into it, and the power consumption test being metered at the wall is pretty shit.

they will open part of code that is theirs
they did it with older versions I think

scheduler is not main issue, HPET apparently gave the biggest performance gain when disabled meaning it's not that hard to fix

nobody reads those, all people will see is graphs where 1800x~=1700 is only 5% away from i7 and costs 20% less

>1800x=1600x(for gaming)=~1700

fixed, in this case 1600x is 70% cheaper
8c will stay niche, anyone sane will buy 1600x instead of i5

Integrated graphics might be a positive for some, I'd say the lack of it is more of a neutral point though. People that buy these chips don't give a fuck about integrated graphics so it doesn't even affect them.

Without 10k$ worth of testing hardware like Tom's has it's pretty difficult to test power consumption

>anyone sane
so not a lot of people

Proper high refresh osciliscopes are way more expensive than $10k. USD

honestly out of the wall is more useful way of testing, you want to know what system would work like with this CPU not just the CPU

You can set the 1800X to work in it's efficient frequency sweetspot, just like any R7

You don't need to be underage to forget X99 and first stepping Skylake problems

Or you can overclock a 1700 or keep it at stock(zen sweetpoint) and save $200

it seriously is. 8 cores at 100% load, only 40 watts more.

Ryzen isn't the shit wrecker we hoped it would be but jesus fuck it's efficient

why would a circuit draw any substantial amount of power if it's doing literally nothing at all

I'd say its a pretty outstanding.
Teething issues aside, its super efficient, and it can trade blows with Broadwell clock for clock, only being slightly behind Skylake/Kaby Lake.

People were expecting it to only offer Sandy Bridge performance on average.

Said boards also brick themselves. Classical ASUS launch, i love how they cooked my 5820k.

It's not? It's just wasteful silicon

>Ryzen seems to heavily favor high speed memory.
It's internal data fabric (interconnect between CCXs) works at half the effective clock of RAM.

It was only expected to have a 40% increase over faildozer but it managed 52% which is pretty outstanding when you think about it. Only time will tell how this will pan out. The release could have been better no doubt but at last we have some stiff competition in an otherwise stagnant tech market.

It's half the effective clock right? So it's the "true" clock speed of the memory then?

It uses ARM TrustZone so i highly doubt that.

ANUS has really turned to shit, everything up to Nehalem and ASUS is recommended, most shit after just got worse.

Holy shit I'm about a few more ASUS bugs away from denouncing them as ABIT-tier

Rumor has it that AMD will address this issue. Probably with better chips further down the line (early adopters always get screwed anyhow). I doubnt we'll see any microcode improvements to this but who knows.

Kinda, yes.

other rumor has it AMD will do huge microcode revision in May that will add better RAM support

My Prime B350 Plus is stable, everything working fine more or less, but holy shit overclocking on this board is terrible. Having to enter multiple different sub menus to effectively alter one thing. Setting memory clocks and timings on it was a chore, and it wasn't because of poor compatibility, it was just finding where the fuck everything was located. Their UEFI is really shit.

In retrospect I should have waited for something else to come back in stock.

Yeah. A 2400mhz DDR4 kit = 1200mhz data fabric.
The actual nuance of how DDR4 signals is so complicated its not worth going into though.

That isn't an issue, its just how the clock ratio works.
Using faster memory increases the speed of the data fabric. Its how everything maintains coherency.

did you sue random "not validated" RAM sticks?
I'm curious how ryzen treats any non validated RAM if you manually set up timings

It's pretty amazing what AMD is doing with such a high latency memory controller, if they can get DRAM calls down to 50-60ns in a year or two these things will be monsters allround

I used a 2933mhz HyperX kit. The DIMMs are good, worked in another system, booted fine in my Ryzen build, but they defaulted to the default 2400mhz profile.
The ASUS auto overclocking tool flat out doesn't work at all. Selecting 2933 or even 2666 from the drop down autoOC menu causes the system to do an endless boo tloop. It apparently doesn't try to adjust the voltage of the memory at all, but tries to increase the clocks while leaving it at 1.2v.

Unless you go through a handful of different menus and set everything by hand it won't boot. I don't think I'd ever buy another ASUS board after this one.

3200mhz RAM with C16 timings yields a latency reading in the latest AIDA64 of 76ns.
Faster memory with tighter timings gets you down to mid 60ns range. Any improvement to the IMC in Zen2 will effectively make system memory so fast that they're already at the point of diminishing returns. Very high effective utilization, extrapolating over 95% of theoretical bandwidth, and low latency at the same time.

What was AMD thinking putting the ASUS in the review kit instead of the Taichi or Aorus? At least these two don't crash the system or lose performance.

Some dudes got Aorus boards.

ASUS probably forked over some cash to be the main partner, but then totally dropped the ball.

>What was AMD thinking putting the ASUS in the review kit instead of the Taichi or Aorus? At least these two don't crash the system or lose performance.

There's memory bugs on the Aorus as well, but that seems to be its only glaring problem at the moment.

Remember asus with x99? Took them 2 months to fix it

And this is why the ccx exist, bigger, monolithic core complexes like Intel has are harder to control when power is in question.

>20 watts
150 -120 != 20
>intel fanbois btfo by simple maths
Its no wonder you guys can't pick the better value cpu.

They also suck some gigantic dicks to scale. I wonder how fucking BIG 32coarz Skylake-EX will be.

Lol yields

Since the other shill thread died before I could make my post to comment on the matter here.
Here.
>Games comparison: 33 for intel and 7 for ryzen.
>Workload comparison:15 for intel and 8 for ryzen
>Over all: Ryzen costs 200 more and only surpasses the 7700k in "H.265, h264, Wprime, blender, cinebench, veracrypt, 7zip" so if you use those a lot and the performance really matters to you get ryzen 1800x
>for everything else it's both cheaper and better to get the 7700k.

The intel chip has more shit on the chip, and requires higher voltages

...

...

Eh, no? 7700k is smaller.
It was nuked, Ramesh.

I'd say that's fair but the chip wasn't intended to compete with the i7 7700K. It's meant to be a lower cost alternative to the 8 core i7s but it also has a performance advantage on them in a few cases and lower power usage as well

More shit? The Ryzen is a fucking SoC, has a full southbridge integrated on top of that has GMI links for multi-socket support as well, which are useless to this market segment, there's a reason it has more transistors.

"That's not even my final form."

- C.S.W.

Naw dude, Ryzen is smaller, AMD made a big deal about it, it's about 10% smaller

techpowerup.com/230446/amds-ryzen-chips-10-smaller-than-comparable-intel-skylake-dies

Meant like less FPU and stuff

youtube.com/watch?v=ylvdSnEbL50

Jesus christ

Ryzen is fucking garbage

What the fuck was AMD thinking

Ryzen cores are smaller, the die itself is bigger than 7700k.

PCper already went over the scheduler issues, AMD has stated there is no issue

youtube.com/watch?v=6laL-_hiAK0

The CCX is smaller than 4 of intel's cores, that is not die size. Learn the difference.
The Zeppelin die is somewhere around 200mm2~ desktop Kaby Lake is 126mm2.

>what the fuck was AMD thinking
Selling server dies is totally okay, Ramesh. Now stop samefagging.