AMD Myrtle 16-core Geekbench result

holy moly, intel _is_ dead.

My 9700k scores 6200 which is quite good compared to the price of the 3950X, I am really excited for thermals dough, 9700k runs quiet hot in Ghost S1 even with undervolt.

Jesus Christ Zen2 has leadership IPC

The caches are all wrong what the fuck?????

Well we already know Intel's hyperthread is weaker than Ryzen's SMT.

512KB x 16=8MB L2
16MB x 4 = 64MB L3
Seems correct to me

>4.29GHz
it's not even at its full power levels

Yeah and your 9700k clocks to 5GHz and 3950X is at 4.3GHz

Just an FYI both of these results are with CPUs clocked on all cores at 4.3 and 4.5 respectively so they don't really say anything about boosts on either one.

Less than half

3950x boost is 4.7ghz

9700k all core turbo is 4.6ghz and single core turbo is 4.9Ghz

The 9700k doesn't ever hit 5ghz without OC.

It says 4.3GHz there, so the SC score is for 4.3GHz

So I guess they will have nearly the same single core performance?

Will 3950x come out on 7/7 or at later date? Also, is 7/7 the official release date or it'll be already available in stores by then?

They said September for 3950x

They're just hiding their power level so when third party benchmarks come out it's going to absolutely rape the competition and surprise everyone with even higher results.
They're not even running it on a 570x mobo.

If the 3950x gets 5868 SC at 4.3 GHz, it should get around 6410 SC at 4.7 GHz.

Your 9700k has 8 cores. Do you really feel the need to add 8 more cores to your PC? What are you going to use them for?

It's correct. The new Zen 2's have an insane 72MB of Cache.. That's double even kikeripper. Intel has nothing that looks like this even in the server Xeon chips.

Intel ram is @ 3600Mhz
AMD ram is @ 4100Mhz

Yeah, I'm going to say this is definitely a big factor. That being said, this is an onpar processor in terms of performance. X299 is HEDT for intel but the platform is a joke as there are desktop equivalent amount of PCIE slots coming out of the CPU and most shit is routed through the chipset. Thus, this is a fair comparison. This is great news for am4 but i'm personally waiting for Kikeripper2 to drop so I can upgrade my X399 platforms. I felt it better to add another threadripper vs buy the new ryzen CPUs. I think it was a wise decision. I'll give it some time before I upgrade my ryzen setups. In any event, good launch by amd. Happy to see them detail the 16 core. I've already paid the early adopter tax on Ryzen one dropping a grand on a 16 core. I'll wait this time around until prices drop.

Look forward to the 7/7 release nonetheless, performance reviews (that 72MB of total cache is insane but I feel it is used in part to mask the added latency w/ the I/O chiplet), and the hotchips technical talk on how this is all wired up. kikeripper 2 is going to be a beast !!!

Yeah, the performance reviews are going to be quite interesting. I want to see what the new 72MB of cache, 7nm, and I/O chiplet (added latency) do in terms of real world performance.

Yeah thats my point, right now core count is not so importent to me, thermals and single core perf is what matters to me

(which zen 2 seems to provide, but 800$ is a lot, x570 mainboards will be expensive, in the next two years pci-e 5 is coming and still not sure if zen 2 wants high clocked ram which is also quiet expensive)

isn't 4100 mhz on the other side of the 2:1 ratio for IF? so it might actually be hurting it in some tests

>That single core at 4.3Ghz
>All core boost most likely still has ways to go since even the last gen TR reaches those clocks.
>Last gen motherboard
>Mfw they're not even showing their full potential.

HHNNNGGG I can't wait to upgrade.
It's going to be one hell of an uplift. September can't come fast enough.

Attached: he-man.png (251x192, 139K)

If single core is what you care about then don't buy the 16 core $700 part

And only pci e 4 cards are some mid range navi, lets see if nvidia super cards are pci 4

not sure what to think about all this

it has the highest turbo of the bunch.

The 8 core parts cap at 4.5Ghz boost

It is. The sweet spot is below 4ghz on the memory, although we don't know if there's any options to force a 1:1 ratio when Ocing and if there's IF voltage control to crank the beast higher.

Awwww man. I want the 16-core one already. My 5 year old i7 is not nearly enough for a gpu-passtrough setup. It's bottlenecking GPU as hell.

Even the Intel subreddit upvote this. The only thing they have going for it is 4channel memory/8dimm support for serious work. Once the equivalent threadripper 2 comes out Intel won't have anything at all.

Good catch ! You are correct. I'm still skeptical of how high that multi-core is. I'm thinking it is more likely due to that massive L3 cache than anything. Something that won't weigh that much on my compute (lots of random access) vs cache-able access.

techquila.co.in/amd-ryzen-9-3950x-vs-intels-18-core-i9/

This site claims it's not the final chip but rather an early sample and that's why the clocks are lower than they should be.
>The best part is that this is an early sample, slower than the final chips that will hit the market.
Wouldn't be surprised if this was the case.

Well, Myrtle is the name they have been using for their testing platform since Ryzen 1 so it makes sense

>September
How are you doing, future 3950X buyer?

Attached: chad reply.jpg (912x1024, 71K)

Well that's in line with what we knew so far.
AMD really got the IPC crown back at it.
Never mind 5Ghz, they can beat intel 5ghz shit at less clock.

The thing I'm more interested in is how far you'll be able to overclock them. We know from their promo video that it can go beyond 5.3 GHz with LN2 and hit at least 4.2 on all cores with liquid cooling. Combine that with next year's Zen 2+/Zen 3 refresh and we might actually have a 16 core CPU that's capable of hitting 5 GHz on all cores with a high end air cooler.

Geekbench is a fucking worthless benchmark btw, AMD's liquid nitrogen "world record" multi-core score is at 65,000. No way an engineering sample of the same chip scores 61,000 when running at BELOW STOCK frequencies.

Will still have less fps in gta5 at 1080p idc

Maybe XFR3 just overclock your shit to whatever it's capable of.
Reminder: AMD didn't say anything about how xfr operates with zen2.
Could be that they made it auto oc altogether, and all you have to do is apply cooling.
Was close enough to that with zen+, but now it's 7nm.

>GTA5
You can play this gayme on any toaster 4 cores.
Are you retarded?

>idc
Nice post, it completely encapsulates the average buyer.

I was joking. And you need some compromises with the graphics settings and rendering resolution if you want to run it with a 4 cores toaster even with a gtx 1070. Game is optimized like shit once you start getting the sliders all the way to the right.

Their marketing team is fucking braindead, so it's possible. That review embargo date can't come soon enough.

Here's a trick.
Get sliders all the way to the right minus 1 step.
You'll find out it's the same fucking image and uses half the GPU power.

See, I wouldn't be having these thoughts if it wasn't for the 3800x.
What is it it has that wouldn't have you buy a 3700x instead?
Well, that 105W vs 65W.
XFR is gonna keep clocks high as core useage goeas up.

Wonderful my fellow chad.
Mainly working on my art and physique while I wait for the CPU to drop.

Overclocking will be interesting, but I'm not expecting that much of a boost over a single core hitting 4.7GHz. Maybe it's going to reach 4.9 on a decent chip and some rare golden specimens are going to reach 5GHz.
I'm expecting something along the lines of 4.5GHz all core, considering that the current 16 core Threadripper is able to do 4.2Ghz all core.
But even if the OC capabilities were utter horseshit it doesn't really matter much. If it touches the advertised numbers it's already going to be hanging out there with top tier Intel in SC and MC is going to be out of this world.

Attached: Average Ryzen owner.jpg (452x981, 113K)

Reminds me of absolutely everyone who benches deus ex mankind divided. You max everything except msaa (which isn't even parto f the presets - its in a seperate menu) and at 4k a 280/VII can push 50fps~ in the open world parts and a 2080ti is capable of a 60fps lock at 4k. However that is sensible and everyone benches goes ful lretard and runs at 4x (or 8x, I forget which it goes to) msaa and suddenly the mighty 2080ti is barely reaching 20fps.

>play on a smaller resolution
>quadruple your framerate
5 WEIRD TRICKS NVIDIA DOESN'T WANT YOU TO KNOW ABOUT

>75ns at 4133
they're beating their own marketing slides with Engineering Samples

Attached: 8-1080.130f555c.png (1777x998, 304K)

Better to play it safe and not disappoint, than it is to lie and come off like a thieving kike.

S A N D B A G G I N G

It does feel like they are sandbagging hard with this release.

Not applying the recent Windows updates that finally unfucks the scheduler problem.
Not gimping Intel with the security patches.
Showing benchmarks with lower frequencies/engineering samples and using shitty RAM. This one had good RAM though.
Having worse results in their slides with these latencies.
Using some Intel favored benchmarks and games.

It's like they're putting themselves in the worst possible situation imaginable with these previews.
Can't wait to see the results of actual benchmarks. They're going toe be even better than what we have seen so far.

>They're just hiding their power level so when third party benchmarks come out it's going to absolutely rape the competition and surprise everyone with even higher results.

That seems to be a common theme lately. Lisa Su also is very modest about the expected earnings for each quarter. It wouldn't be surprising if she sandbagged the results so far, creating kind of a viral hype. This seems far more exciting than Intel's overpromising and Jewish trickery nobody is falling for anymore.