This kills the intel faggot

The 9900K was released three months ago, though.

>2 years later
what are you talking about? the 9900K was released a few months ago.

>2 years later
I'm sorry, what?
When did I black out that 2 years passed between 9900k release and now?

I don't see how that's possible when you're suggesting they double the core count without significantly neutering the clockspeed.

S

Attached: 16c ryzen.png (590x310, 28K)

S

> almost touching
kek, what about Intel's cores?

just use a cooler, faggot.

> 2950X
> 4.4GHz boost
> 3.5GHz base
> 12nm

S

Attached: 1506378560466.png (489x394, 228K)

What about them? If you're glad AMD is joining the housefire meme, then great I guess.

Attached: Screenshot_20190109-152248~2.png (696x598, 598K)

Are you fucking retarded?

Binning. The more chips you bin, the better the ones you find (these are rarer of course). AMD needs to produce a few millions of these before they know what's really feasible. This is why only one chiplet was shown today.

He's probably an intard so yeah.

Yeah the 2950x has a massive amount more space for the dies to dissipate heat.

Binning costs money, 7nm is still new. I think you're overestimating how this will play out for AMD.

So... will we have to draw a penis in thermal paste on the heat spreader to get optimal heat transfer?

what if we stacked them on top of each other instead?

3 chocolate pieces?

user asking the important questions.

73mm^2 CCD shipping in humongous volume to satisfy hyperscale will produce enough leaky speedy bind to saturate DIY.

AMD has money now. They are selling Zen and Zen+.

>AMD has money now
lol, no, they're paying off the R&D costs of Zen.

Gee Bill! How come your mom let's you have TWO chiplets?

you clearly don't understand how thermodynamics works.

>Jow Forums
>knowing anything at all

He's making fun of the people posting about Radeon VII and the 1080ti

Yeah, just double the core count in the same space, or to compare to TR/Epyc, just take half the die size and shove the same number of cores in there, and then clock it even higher because we still need consumer single core performance for gaming.

Oh wait, that's never gonna fucking happen which is why they were showing off the 8 core part today, which will likely be their highest performing CPU when it comes to gaming.

I like the way you think

>2 years later
>9900k is couple months old
?

If the dies are apart, there are two separate hotspots. Further more, heat does not spread a lot vertically through the heatspreader. Thermal conductivity increases when the temperature difference between the hot surface and the cold surface is big, so a chiplet design should be easier to cool.

then why doesn't the R3 1500X clock higher than the 1800X? it only uses a single CCX! The 1800X should overheat so much that it brings down the gaming performance right?

TLDR: Just use a bigger cooler.

I agree compared to a monolithic design, but you're out of your mind if you're expecting fast clocks on a 12c or 16c AM4 socket CPU.

There is no way in hell you'll see 5Ghz for the 12c, or 4.8Ghz for the 16c like the AdoredTV leaks were claiming.

You MIGHT see 4.5Ghz on the 12c and 4.3Ghz on the 16 core.

With maybe some single core turbos that reach higher.

You should see /v/

Attached: 1547059135598.png (511x467, 230K)

No clock guessing? Worst case scenario it was 3.9-4.0, but depending how bad the sample is it could be even 3.6

NO THIS CAN'T BE HAPPENING! THEY ONLY OFFICIALLY ANNOUNCED EIGHT CORES! THEY CAN'T POSSIBLE PUT ANY MORE THAN THAT ON THE CHIP!!!!!

Attached: cores.png (867x862, 110K)

>There is no way in hell you'll see 5Ghz for the 12c, or 4.8Ghz for the 16c like the AdoredTV leaks were claiming.
That's just SC turbos and they go down as more cores get loaded.
Nothing stops you from overclocking the SiPs, assuming suitable mobo and cooling.

are you nuts? that score was at least 4.3GHz all core.

Reminder that this was a low clocked non final silicon chip, and it outpeformed the i9 9900k while consuming 75w in CB R15.
It handedly outperformed the 2700X in the same bench at lower clocks.

Silicon is thermally conductive enough that it doesn't matter. Heat is transferred up through the die at a high enough rate. There are plenty of MCMs out there with a similar enough configuration.

Ryzen 2000 series advertised turbos are all-core turbos.

yes. by drawing a penis on the heat spreader you should get optimal thermals.

confirmed by Lisa.

a 2700x all core at 4.4Ghz is just slightly lower, so no. There is no way in hell it was 3.6Ghz. What a joke.

they had an underclocked r5 beating the shit out of a i9 dude. iid say that really impressive

Attached: d8bgwwf-11cfb91d-ffad-46f2-8a03-23b79cb03897.png (1264x1057, 2.06M)

S

Attached: zen2 extra die.png (678x2140, 1.19M)

There was an increase in IPC, retards

4.35 for 2700X isn't the all-core turbo by any stretch of imagination.

>underclocked r5
That is simply speculation my negritudinally-inclined comrade. For all we know it was the R7 3700x

Straight from AMD at Next Horizon they claimed up to a 29% IPC uplift in certain ops.
This chip wasn't anywhere near 4ghz or it would have scored a lot higher in Cinebench.

I doubt zen2 will manage to reach a 13% IPC improvement as you're suggesting.

he's probably talking about this:

XFR frequencies are only achieved on one core, and only for very brief periods.
AMD does not advertise sustained all core turbo frequency.

They doubled the datapath width of the FPU, user. The Zen2 core can natively execute 256bit ops without needing to double up to 128bit FMACs.
AMD themselves touted up to a 29% uplift in IPC.

That's an overclocked chip.

Few things we know
9900K was NOT TDP limited in anyway. It was shown out of the box and full stop performance without any sort of throttling.

This ryzen 3000 chip is faster than any air cooled 2700x today. Some water chips have hit 2,000~ CB score but thats with 4.4ghz and extreme memory setups above 1.7 and timings custom tuned.

This chip new ryzen chips was 4.5ghz+ base I would guess it ran at 4.65ghz with XFR 3.0 boosting all cores.

I am not impressed by the performance so much as the wattage.
60~ watts less power and better performance? Thats amazing.

Attached: RyzenSkylakeIPC.png (1276x713, 349K)

I agree, efficiency is impressive as hell, performance was meh.

Will be interesting to see what clocks they can achieve on the 12c and 16c parts. But i'm not holding my breath.

>BAM!!!
this kills the shintel shill

Attached: 1541630210309.jpg (3169x2109, 434K)

I don't think they did that much work for a mere 5% IPC increase

Attached: Screenshot_20190109_200437.png (704x620, 160K)

Given that we know for an absolute fact that Zen2 has a huge IPC uplift in FPU ops, and that would be directly reflected in a Cinebench rendering workload, the assumption that this engineering sample was clocked anywhere near 4ghz or higher is ridiculous.

Anandtech's own testing put their 2700X at 1754 in CB R15 with their paltry test config.
This engineering sample scored 2050.
If each chip had totally equal clocks, then this would extrapolate to a ballpart 15%~ increase in perf/clock. And you think this chip was actually running the bench at 4.5ghz or more?

yeah, but AMD is claiming a massive 29% IPC improvement.

in some workloads yes but it wont be 29% across the board
keep in mind the 9900K was running 4.7ghz or so all core turbo
AMD is not magically going to match that performance with much lower clocks. IPC gains are minimal unless new instruction sets are being used like AVX 512. This also means the program has to make use of that too.

I suspect this ryzen chip is the 3600X with near final clocks. All core boost is 4.6ghz and could hit a bit higher with XFR
Max single/dual core is probably 4.9ghz maybe hitting 5.0 with XFR
Also this is a 95w rated chip

2700x will run 4.15ghz all core boost
stock scores are around 1750

>AMD is not magically going to match that performance with much lower clocks.
They very well could depending on how the FPU is structured and how it handles the workload. Intel's larger FPUs aren't better at everything universally, even Zen1 can hold its own depending on the instructions used, and on the interger side of things there wasn't much difference between Zen and Skylake in perf/clock.
Thats precisely why Zen1 also performed well in Cinebench at launch.

If this engineering sample had substantially higher clocks, it would be scoring far higher in R15. Its patently absurd to think something that highly clocked would score so low.

Attached: RYZEN7-1800X-38.jpg (614x405, 107K)

If the 2700X stays at 4.15ghz all core during a R15 run, and this engineering sample is scoring 15%~ higher, do you really think its clocked that high?
It would be a fucking embarrassment for AMD if they only managed a 5% or less IPC uplift in a rendering workload specifically when they reworked and doubled FPU width. Cinebench is a bench that would reflect these arch changes the most, not the least.

Frankly to claim that this chip was running at 4.5ghz or higher makes one a complete shit eating retard.

>This kills the--
Now either move that I/O up or down, or turn it on the side.
Kek.

I think the clocks have been set to match the 9900k and show it's done at much less power. It doesn't look quite as twice as efficient as the 2700x unless you consider the 2700x can use up to 140w if pb2 kicks in. That could be down to so many things though, 2700x is clocked above the sweet spot of the frequency curve so the 2x efficiency gain isn't all there, zen1 is already good at cinebench so there's not as much ipc gain in these situations or the new fpu paths in zen2 chew up power or the IO die uses a lot of power as well or even 7nm is a dud at 4ghz.

In the end it's only given us a teaser and is deigned to let us speculate to the ends of the earth to talk about ryzen 2. Come closer to launch they can give a full reveal and the speculators have nothing left to talk about.

If clocks were matching the i9 9900k then it would show an IPC regression compared to Zen1/Zen+.
Theres no way this chip was clocked north of 4ghz.

But we don't know the clocks and IPC is application dependent. I'm sure someone will draw a graph showing it's potential score and how it matches against the 9900k/2700x in power or ipc at the suspected frequency range of 3.5ghz to 4.5ghz. All i'm saying is that they clocked it to match the 9900k, from that we can only make extrapolated guesses at ipc change and frequency.

smaller feature size, lower voltage, less powerloss

>Theres no way this chip was clocked north of 4ghz
Was running Cinebench test with a ES stock Turbo of 4.6, according to AMD. And that's not even finalized spec.

2700x gets just over 19k at 4.2ghz so thats disappointing if true

Are you a retard?
9900K got 240 on Single Core, while 3700X ES underclocked sample got 275. Buy yourself a fucking brain and learn to count 2+2.

The 2700x with decent DDR4 and a lucky 4.4ghz OC can hit 2000. It'd be utterly horrible if Zen2 performed like this at 4.6ghz.

2050 is far higher than 1750. Ryzen 1700 scores around 1400 for reference. Same 8 core 16 thread design.

2050 is 117% of the score of 1750
4.6ghz is 112% of 4.1ghz
Yielding around 5% gain IPC on this test.
Thats 2700x vs this ryzen 5 8 core chip.

Do not expect 15% IPC in a test like cinebench.
CB has almost universally been about clocks and threads. Its not big on memory has no care about intercore dependencies. Its pretty much strictly floating point and thats why it scales so well on ryzen and other many core CPUs

We will see memory latency reductions and some other performance improvements but actually IPC or CPI(cycles per instruction) gains will be minimal. Expect 5-7% at most for IPC but overall in some applications we could see that 25% figure clock for clock.

Performance gains will be good not great but power consumption reductions will be great.

Attached: RyzenSkylakeIPCGame.png (1276x719, 559K)

Then you're a literal brainlet.

Both were 8 core 16 thread parts. They just showed they were matching the 9900k using 12% less power with no boost clocks.

How is it horrible? In what world...
If this 4.6ghz 8 core is beating intels 4.7ghz 8 core while using 40% less power how in any world is that horrible?

The 2000~ CB scoring 2700x are massivly tuned chips running on the edge. I dont think there is any air cooled 2700x chips going over 2K in CB. They all have very tight tuned memory sticks too.
This is an early ES sample probably running with slow memory.

The fact this chip will cost ~250$ vs the $500 9900k means its actually amazing. AND its compatible on motherboards that have been out for two years

the 9900k just came out a few months ago.

>AMD shows an early engineering sample non-final clocks 8c/16t Ryzen 3000 that beats a 9900k in Cinebench (indicating at least equivalent per-core performance)
>AMD can fit 2 x that number of cores on one Ryzen 3000
>Intel cannot even maintain the 9900k's per-core performance on their upcoming 10-core meteor lake due to thermonuclear housefire power consumption that will throttle on liquid cooling
>Intel literally has no answer for 16-cores of 9900k per-core+ performance until they unfuck 10nm AND roll-out a new arch
>Intel not utterly, totally, completely, permanently BTFO
(lol

The performance uplift would be horrible. A rendering workload is exactly what should respond to the greatest degree to a massive increase in FPU perf/clock. CB scores barely moving per clock is objectively awful.
You have no idea what DDR4 they were running. It could have been 2933mhz DIMMS, it could have been 4266 DIMMs.

Fuck off, I'm talking about the overall score here. If this was running at 4.6 that would be disappointing. It was probably at 4ghz. Stop being a pratt.

It'll be little Navi as the iGPU. AMD isn't going to be cannibalizing their Threadrippers 1-2 yet.

AM4's successor will see the first customer-tier 12-16c SKUs.

what the fuck is that?

what are you expecting out of it?
Its still the same basic design of zen but on 7nm and segmented into smaller chips.
When is the last time intel had a 5% IPC gain?
Ryzen had a 52% IPC gain over FX
Ryzen 2 had a 5% IPC gain over Ryzen 1
Ryzen 3 will have a 5+% IPC gain over Ryzen 2

Higher clocks and better IO/memory control is where most of the gains for Zen 2 will be.

Intel cant compete on price its that simple.

Attached: 2600TunedTimings.png (1297x901, 591K)

What's the point?

2018 was already an awful year for Intel. Seemed like just an excellent year for AMD. Now this comes out. What's even the point of ever buying Intel? Is there one single fucking reason? Didn't they delay making new products that are completely immune to shit like spectre and meltdown and not just patched up?

The only place where it seems a bit lackluster is GPUs. They need to step their game a little bit up, the Radeon VII ain't looking all that amazing. AMD has been always excellent at "doing same shit for cheaper" and that card doesn't scream good price. Hope they release something like a 680 that will beat the 2060.

APUs will still be monolithic after the newly announced APUs

Why do you think Raja has them hyping up 10nm promises with no delivery?

Intel is finished until Whiskey Lake

>What's even the point of ever buying Intel?
Superior single-core performance and higher clock frequency. That ES demo was running at 4.4GHz all core, but production silicon won't be that much higher.
Pre-keynote Ryzen ES Cinebench demo showed a 10% 1T deficit versus the stock 9900K at 4.7GHz.

>That ES demo was running at 4.4GHz
source: my ass

No they wont...
They will have a IO die single zen 2 chip and a navi GPU chip
This will allow up to 8 core 16t and also allow better memory control with possibly support for 4 channel memory on 5xx series boards. Boards could be wired in both dual and quad channel configurations

We could see 2.5x performance vs current APUs

Source: see the leaked ES sample

>Superior single-core performance and higher clock frequency.
Are you actually retarded? Please go ahead and explain how an 8c/16t Ryzen 3000 CPU was faster than an 8c/16t 9900k without having at least equal per-core performance.

I'll wait.