(16core added)AMD 3000 specs! 4.7 GHZ, R9 3950x, R7 3700x, 3800x.

[melete]

3 minutes ago, VegetableStu said:

 

GN may mean the PUBG benchmark here. 

[nick name]

12 minutes ago, RotoCoreOne said:

Think they were comparing price points there, i was confused too and then they did the comparison between 3800x and 9900k which im sure the 3700x performs similarly to the 3800x

But the 9700K costs what the 3800X will so they weren't comparing similarly priced CPUs.  

 

Honestly, I'm just hoping that with that stupid low 65W TDP for the 3700X means you can actually expect some decent overclocking out of it.  It's 40W less than the 3800X with very similar speeds.  

 

 

[IAmAndre]

1 minute ago, SenioRR said:

They have lost their last advantage and that's the IPC. I bet we are going to see some rushed panic launches in the near future.

I really hope AMD is saving their 16c monster for when Intel tries to respond, only to further humiliate them.

 

AMD DESTROYS INTEL WITH CORES AND IPC

Last they panic launched a product, it was the i9 lineup after the first generation Ryzen. I don't know maybe we'll see something like that.

[Origami Cactus]

1 minute ago, VegetableStu said:

 

dafuq is going on in this slide, LOL

Graphic design 101. Don't do whatever that is.

[BryanY]

Given the extremely similar spec on 3700x and 3800x, i'd assume 3800x is just a better binned 3700x, which makes 3700x having very little headroom for overclocking.

Aside from that, i think 3700x is the best bet for regular users, while 3800x and 3900x should be better for overclock users.

[Stroal]

26 minutes ago, melete said:

Ryzen 5 also isn’t confirmed yet but we’ve got plenty of reasons to think that’s happening, too. 

Yeah it is. It's on their website.

 

https://www.amd.com/en/products/specifications/processors/11781,11756,11761,11766,11771

[melete]

7 minutes ago, Stroal said:

Yeah it is. It's on their website.

 

https://www.amd.com/en/products/specifications/processors/11781,11756,11761,11766,11771

Hadn’t seen the press release when I wrote that. Ryzen 5 wasn’t mentioned in the keynote. 

[YoloSwag]

Looks like they skipped a 10C/20T chip lol

[Drak3]

18 minutes ago, SenioRR said:

They have lost their last advantage and that's the IPC.

1. Intel has a clock advantage. The difference between Ryzen 1/2 IPC and Skylake+ IPC is not exactly huge.
2. No one should believe AMD's IPC claims until verified by MULTIPLE independent sources. Intel has claimed multiple times to have a 15% single core performance increases between generations since Haswell and not once has it actually been true.

[melete]

Just now, YoloSwag said:

Looks like they skipped a 10C/20T chip lol

There’s a good reason for that, actually! It’s down to the chiplet design. 12 core is two 6 core chiplets, and a potential 16 core is two 8 core chiplets. 

[Drak3]

Just now, YoloSwag said:

Looks like they skipped a 10C/20T chip lol

AMD does their core jumps in multiples of 2 for single die, 4 for twin die (basically one core per CCX is active on Zen/Zen+ and currently no reason to assume that changed). Assuming that it's still a 4C CCX, that means single dies can come in 2/4/6/8 variations and twin dies can come in 4/8/12/16.

[Origami Cactus]

Just now, Drak3 said:

AMD does their core jumps in multiples of 2 for single die, 4 for twin die (basically one core per CCX is active on Zen/Zen+ and currently no reason to assume that changed). Assuming that it's still a 4C CCX, that means single dies can come in 2/4/6/8 variations and twin dies can come in 4/8/12/16.

Ryzen 7's seem to be single chiplet, so that means that 8 core CCX, meaning that the 16 core will come out later.

[Drak3]

2 minutes ago, Origami Cactus said:

Ryzen 7's seem to be single chiplet, so that means that 8 core CCX, meaning that the 16 core will come out later.

CCX and chiplets are not the same thing.

[YoloSwag]

9 minutes ago, melete said:

There’s a good reason for that, actually! It’s down to the chiplet design. 12 core is two 6 core chiplets, and a potential 16 core is two 8 core chiplets. 

 

9 minutes ago, VegetableStu said:

a chiplet with 5 active cores is... strange? o_o

 

9 minutes ago, Drak3 said:

AMD does their core jumps in multiples of 2 for single die, 4 for twin die (basically one core per CCX is active on Zen/Zen+ and currently no reason to assume that changed). Assuming that it's still a 4C CCX, that means single dies can come in 2/4/6/8 variations and twin dies can come in 4/8/12/16.

 

It was meant as a joke... really.

[Drak3]

@Origami Cactus

Think of it this way:

You have the standard monolithic die. Everything in one piece of uniform silicon.

 

Stitch multiple dies together, you get an MCM.

 

Take the monolithic die, and make it take on some characteristics of MCM (separating cores and resources like cache, but still working together via interconnects), but keep it one physical die, one piece of silicon, and you have a non standard monolithic die. You have a multi CCX die.

 

Now, take certain resources, like cores and cache, and split them from other resources, like IMC and PCIe controllers, and have them all on the same substrate as separate dies, you get chiplets.

[Unixsystem]

Have they absolutely confirmed that the 3800x is a single chiplet part?

 

If it is 2x 4 core chiplets that would largely explain the significantly higher power draw, as well as make it actually make sense as a product; spreading out the heat would likely make OCing easier, and it's possible that it's easier to bin two chiplets for half quality cores than trying to get one chiplet that has all 8 quality cores.

 

I know the presentation sort of implied that the 12 core was the only one with 2 chiplets, but I'm pretty sure that it was never outright said that that was the case.

 

EDIT: I missed that link to the Anandtech article that claims it is only one chiplet. Seems like my theory may be wrong, but I still wouldn't be surprised if they got that from the way the presentation was shown and not necessarily from inside info. Hopefully we get an actual die shot at some point relatively soon.

[plsno]

I really do hope these will OC to like 5GHz. 

[D13H4RD]

Now I'm gonna ask exactly where those rumors about having more cores came from... 

 

That aside, that looks like a healthy IPC bump, but the big question is how much you can overclock them by. Any closer to 5GHz and Intel really will have to start coming up with a proper response. 

 

The long standing argument is that Ryzen is better for productivity applications (especially from a value perspective) whilst Intel is more suited for gaming, but even that is slowly being undermined. 

[williamcll]

Gonna wait for the reviews first. 

[CarlBar]

50 minutes ago, melete said:

Really looking forward to Adored explaining why he wasn’t totally wrong, just mostly wrong on 12 core Ryzen 7 etc.

 

He allready did months ago when he noted that pricing and branding can be changed anytime upto a few weeks from launch. both the low end 8c and low end 12c match the specs he gaves. AMD just chose to use a different branding and pricing structure.Which given they've apparently decided to hold back the 16c actually makes perfect sense. The 16 core is no longer their flagship model, the 12c is and everything else in the product stack has shifted in naming and pricing to reflect that.

 

It even makes good business sense, (the original leaks did too, but it ultimately comes down to what AMD's objective was/is, from this i'd say AMD are pursuing a different objective from what they had when the launched Zen1/Zen+). Personally whilst it's just educated guesswork i'd say AMD went for different pricing and the branding that goes with that because of how various recent factors, (more deals signed with big suppliers and Intel's shortage getting more major tech OEM's using AMD parts), have made AMD more prominent and competitive in the CPU space. That means undercutting intel on both price and performance may now present a degree of diminishing returns in marketing terms that would be detrimental to profits. AMD has been undercutting all these years because they've needed a really compelling argument to overcome Intel's mindshare and general entrenched position.I'd say the new pricing scheme and delayed 16c represents AMD deciding they've overcome that adequetly to adopt a more short term profit focused stance. They no longer feel they need the pricing advantage to get good sales, performance alone is enough.

[Quadriplegic]

57 minutes ago, SenioRR said:

They have lost their last advantage and that's the IPC. I bet we are going to see some rushed panic launches in the near future.

I really hope AMD is saving their 16c monster for when Intel tries to respond, only to further humiliate them.

 

AMD DESTROYS INTEL WITH CORES AND IPC

Intel still has their main advantage since 2017, clock speeds

[SenioRR]

4 minutes ago, Quadriplegic said:

Intel still has their main advantage since 2017, clock speeds

IPC is what matters here. A lower clocked Ryzen CPU is outperforming a higher clocked Intel CPU in single-threaded benchmarks.

Higher clocks don't do very much for Intel anymore as they have pretty much hit the IPC ceiling for their current process.

[Bananasplit_00]

That's one hell of an IPC increase. Benchmarks are going to be interesting for sure

[Quadriplegic]

2 minutes ago, SenioRR said:

IPC is what matters here. A lower clocked Ryzen CPU is outperforming a higher clocked Intel CPU in single-threaded benchmarks.

Higher clocks don't do very much for Intel anymore as they have pretty much hit the IPC ceiling for their current process.

Which is almost irrelevant as Intel CPUs have higher ceiling of clock speeds, entirely negating difference in IPC. Zen 2 chips have to overclock well to be successful

[RejZoR]

Not quite 5GHz, but this ain't bad I guess. 4.6GHz max turbo and 15% IPC uplift should give quite decent performance. I don't think these will be hitting 5GHz even on high end daily cooling, probably like 4.8GHz and 5GHz if really lucky if we are basing it on the existing OC capabilities.