Hacker News Reader

Comment by auvi | Hacker News Reader

I really expected a 16C/32T Ryzen 9 though.

Potentially is a yield issue / segmentation issue.

Releasing the top SKU as a 12 core instead of 16 core part gives AMD:

1.) Still the highest core count "Mainstream" part. I would expect Intel to fire back with a higher IPC / thread performance part rather then try and ramp core counts. Outside of productivity, content creation and server workloads I really don't see the need for anything past 6-8 cores right now.

2.) Lets AMD keep the flawless chiplets for Ryzen 7 where there will be high demand. While providing a way to move more 6 core chiplets. They can change the product mix once yields improve. I would imagine that there is likely better margin on Ryzen 7 (1x chiplet) rather then the R9 (2x chiplet).

3.) Keeps more of the Threadripper line viable for those that need high cores (especially if you don't require the extra memory bandwidth).

4.) Sells a bunch of CPUs now, while keeping the ability to quickly respond to Intel if needed (or as a spoiler around the Intel 10nm release)

Thanks for this analysis, although I wouldn't agree that R9 profit margin is lower than Ryzen 7. It would depend on the yield though, so if yields are low for 8 core chiplets, the R7 would cost more to make than if yields are higher.

It's interesting though that Ryzen 9 is almost the highest clocked part of the bunch.

AMD did not want to release the expected 16 core yet, but had to release some "monster" to get at those hype-affected buyers. So they clocked/binned those 6-core chips accordingly, to make this core-lacking monster more attractive. It also makes sense to sell as many silicon chips in Ryzen 9 package as possible, instead of selling them in Ryzen 5 package - more luxurious, better margin. Of course, anybody who cares about value and does some research before buying expensive high-end parts will wait for the Intel-AMD exchange to play out, or for the 16-core to be announced, and only then buy the best option.

1) Intel won't be able to, at least not until they finally manage to get the new architecture out. Heat is already a huge issue for intel.

2) AMD might be saving perfect core chiplets, but probably mostly for use in EPYC as the margins are higher.

4) I expect 16 core am4 to arrive someday.

Good point on EPYC, I had assumed these would be using a different design like earlier generations, but the chiplet design makes this not only possible, but likely where the best chiplets are going.

I agree that 16 core will arrive this generation, but I could also see it being 6 months or so. I have also wondered if it is waiting for a second generation I/O die with an enhanced memory controller to feed the extra cores.

I wouldn't be surprised to see that come out as a mid-generation bump, personally. But I imagine they ran hard into thermal/TDP issues adding the extra four cores - especially if the dies used for the 3900X are highly binned.

Its going to come after intel has a chance to try and respond to the 12 core. Let intel think they have a chance and finally bury them

Wouldn't that be the domain of ThreadRipper?

Not if AMD's looking to remove ThreadRipper from their product stack, and segment their server chips a little more.

Are they actually planning to? I just recently looked into "modern" hardware (was out of the fold for some years) and it would be a shame to retire the relatively new TR4 platform so soon...

Unlikely though. TR is AMD's only HEDT presence (qual-channel, high core count). X570 chipset will likely s be dual-channel memory, limiting to 64GiB, seems still positioned for mainstream. Intel currently have very high price HEDT offering, hard to see how AMD is going to give up that market. Probably after Rome?

ThreadRipper is not going away:

https://www.tomshardware.com/news/amd-threadripper-roadmap-r...

Quote: You will definitely see more Threadrippers from us

Maybe ThreadRipper 3000 will have 24/32/48 cores.

the problem is new TR is not on their 2019 roadmap PPT

That's too bad, it's a really neat chip line. More PCI-E lanes than a Xeon E5 without the ridiculous premium.

There are rumors that AMD is not happy with TR2's memory performance as the 4 channel memory controller is the bottleneck, they are probably looking at 8 channel memory option for TR3. ;)

Ha well I wouldn’t complain if they did that. The higher core count TR2s have a bit of a weird performance profile because of the memory channel setup - two of the chiplets have to hop through another chiplet for memory accesses. Is that what you’re referring to?

> Is that what you’re referring to?

yes, they had to introduce that extra hop to workaround the fact that it has 4 less channel compared to the regular 8 channel EPYC.

Using PCPartPicker as a reference [0], it looks like ThreadRipper bottomed out at 8c/16t with the 1900X.

So I guess there's general overlap between Ryzen and TR in AMDs lineup there?

[0] https://pcpartpicker.com/products/cpu/#s=63

Yeah, I'm guessing the general overlap is because Ryzen and TR run on different sockets (AM4 and TR4, respetively). The 1900X is an economic solution to get your own TR4 platform running and then you can upgrade it to some monster 32C beast of a processor, when you really need the extra CPU power.

Good point. I'd forgotten about TR4 and its absolute behemoth of a socket.

Somewhat, though bear in mind that even the 1900X ThreadRipper has 4 memory channels and a ridiculous number of PCIe lanes whereas Ryzen is 2 channel and has much less PCIe.

My guess is that bandwidth and cooling limitations meant that 16C/32T Ryzen wouldn't have been much faster than 12C/24T. It sounds like the 12C/24T outperforms the 12C/24T Intel; if they had released the 16C/32T reviewers would compare it against an 18C/36T Intel chip, which it may not do as well against.

Probably saving golden 8 core dies for Epyc.

It'll come whenever the market needs it, but right now they're doing fine without it.

16C is probably coming later once Intel responds.

Intel are planning a 10 core desktop part, so the moment it pops its head up AMD will drop their 16 core chip on it.

Yeah, that's almost a given isn't it? If AMD has really closed the single-thread performance gap then Intel is in for a hard few years in this market. Looking forward to independent benchmarks.

Reply on news.ycombinator.com