Ryzen2 rumoured to have up to 16 Cores

[mackerel]

You were also using an 8 core CPU. Even with Intel, the higher the core count, the more gaming performance drops. The difference between that 6700k and a 6900k would probably be similar. Games favor high IPC, and higher clockspeeds, so that you have more of those high performing clock cycles.

If the clocks the the same and thermals are taken care of, higher number of cores shouldn't negatively affect performance. Base clocks tend to drop with more cores to help reign in that power. In the case of 1st gen Ryzen, I think I was held back mostly by the low clock wall even with overclocking. That's in part why I wonder if I were to repeat the same today with a 2700X, if I would have the same experience.

 

[Vishera]

If the clocks the the same and thermals are taken care of, higher number of cores shouldn't negatively affect performance. Base clocks tend to drop with more cores to help reign in that power. In the case of 1st gen Ryzen, I think I was held back mostly by the low clock wall even with overclocking. That's in part why I wonder if I were to repeat the same today with a 2700X, if I would have the same experience.

That's why I said more cores = less gaming performance, because of the necessity to downclock.

I think the 2700X's performance increase also comes from AMD letting the TDP go up to 95W. If the 1000 series Ryzen chips had all been 95W, like most of Intel's 4c 4/8t and 6c 6/12t parts have been for a little while now, I think we would've seen performance much closer to Intel CPUs on offer. Admittedly, this would mean the refresh wouldn't have been as much of a leap, but is there anyone upgrading from 1000 to 2000 chips anyways? I think I'm one of very few exceptions there, and that's only because I'm probably going to grab a 2600 over a 1600 when I upgrade from my 1200.

 

[mackerel]

That's why I said more cores = less gaming performance, because of the necessity to downclock.

I think the 2700X's performance increase also comes from AMD letting the TDP go up to 95W. If the 1000 series Ryzen chips had all been 95W, like most of Intel's 4c 4/8t and 6c 6/12t parts have been for a little while now, I think we would've seen performance much closer to Intel CPUs on offer. Admittedly, this would mean the refresh wouldn't have been as much of a leap, but is there anyone upgrading from 1000 to 2000 chips anyways? I think I'm one of very few exceptions there, and that's only because I'm probably going to grab a 2600 over a 1600 when I upgrade from my 1200.

Within this audience, 8 cores isn't that bad to keep under control, outside of using abusive voltages to push to the limit anyway. On Intel side I'm having no problems with air cooling on 6 cores to beyond 5 GHz. My Ryzen 1700 and 1600 were never stellar samples, and realistically were only efficient to 3.7 or so, and things were getting rather uncomfortable around 3.9. If AMD had made it 95W TDP, it could have been closer to that but in my use case it wouldn't have helped since I OC'd anyway. Note when I mention clocks and OC, it is all cores at that clock. Never cared for single core turbos.

I still feel a need to get a Zen+ CPU for testing to understand, for assorted compute cases, how that differs from Zen. As mentioned elsewhere, I have now seen an example where AMD SMT gave a high boost to output, far higher than Intel HT, and I'd like to investigate that area further. I've seen it on Zen, but would like to include Zen+ in that test.

 

[Alaric]

They're on a roll. I'd expect more cores sooner rather than later, because it seems to be working for them.
https://www.marketwatch.com/story/i...vince-you-2018-07-25?siteid=yhoof2&yptr=yahoo

Pretty common "rumor" at this point. https://www.hardocp.com/news/2018/0...dly_offer_higher_core_counts_major_ipc_gains/

If the IPC part is on track, Intel is looking at a long, hard road.

 

[Vishera]

They're on a roll. I'd expect more cores sooner rather than later, because it seems to be working for them.
https://www.marketwatch.com/story/i...vince-you-2018-07-25?siteid=yhoof2&yptr=yahoo

Pretty common "rumor" at this point. https://www.hardocp.com/news/2018/0...dly_offer_higher_core_counts_major_ipc_gains/

If the IPC part is on track, Intel is looking at a long, hard road.

It's nice to see Intel getting their much deserved kick in the ***, but let's hope AMD doesn't hit too hard. If Intel has to completely halt new consumer CPU production like AMD did (the Athlon chips were all still based on Bulldozer/Steamroller/Excavator/Piledriver) and go back to the drawing board, who knows what AMD will do. Might be seeing $2,000 10 core chips from them as well.

 

[Woomack]

If you look a bit deeper in the stats then you will see that market shares are going up faster than the actual AMD sales. It's because investors expect to see improvements in couple of months/years and see the potential which wasn't visible before Ryzen. It doesn't mean that AMD took significant part of the market from Intel. I guess that Intel panic and tries to lower expected loses in next years ... not now as now AMD still has low % of the market. Look at AMD product line ... processors for home/entertainment and graphics cards which are losing with nvidia. Their workstation/server processors/chipsets are barely selling even though performance/price comparing to Intel is really good.

Re some earlier posts. You have to keep in mind that Ryzen wasn't designed as higher TDP/wattage chip and quite high voltage/bad scaling and low frequency wall are related to that. Also TDP means nothing in comparison. There are similar chips at 35, 65 and 95W. Some Intel lines share the same TDP like i3, i5 and i7 at the same TDP.
TDP is thermal design. We see 4, 6 or 8 core processors at the same TDP while wattage can be between 90-200W under load.
To keep TDP in specs, nearly all modern CPU auto-adjust voltage and frequency. Most chips could run faster but out of specs.

Now what I find funny is that we see a lot of noise about CPU performance, new technology and other things while we don't really need that. It's just desperation of marketing/sales departments as desktop computers are selling worse each year and users don't see the reason to replace their well-performing PC just because there is something new on the market.
Both Intel and AMD could slow down and release well-designed and actually tested products... not beta versions that we later test for them. No matter how good Ryzen looks, it was a year of beta testing on our side to make it look good. Intel is not much better.

 

[Kenrou]

Now what I find funny is that we see a lot of noise about CPU performance, new technology and other things while we don't really need that. It's just desperation of marketing/sales departments as desktop computers are selling worse each year and users don't see the reason to replace their well-performing PC just because there is something new on the market.

This. While it's fun for us to upgrade and tinker most average Joe's simply don't really need more/faster cores. As many here pointed out, 5yo-8yo cpu's are still up to most tasks nowadays including gaming. Hell, you can flat-out annihilate the top consoles in the market with some budget setups...

Why would most people upgrade *shrug*

 

[Dolk]

..snip..

Now what I find funny is that we see a lot of noise about CPU performance, new technology and other things while we don't really need that. It's just desperation of marketing/sales departments as desktop computers are selling worse each year and users don't see the reason to replace their well-performing PC just because there is something new on the market.
Both Intel and AMD could slow down and release well-designed and actually tested products... not beta versions that we later test for them. No matter how good Ryzen looks, it was a year of beta testing on our side to make it look good. Intel is not much better.

I wouldn't call it beta testing. Maybe for the new materials such as the case for Zen 1 going to 14nm and Zen 2 going to 7nm. But outside of that, the logic and architecture is tested extensively during production phases. However, as we have made these newer processor architecture advancements, our ability to do full test coverage has dramatically decreased. I'm not aware of the coverage percentage Intel and AMD target for testing purposes on their new architectures, but from personal experience in hardware it could range from 60 to 80%. But I wouldn't be surprised if it is lower. Intel's Skylynx still has bugs coming up and its now 2 gens old.

Point is, we aren't really beta testing anything more than the new materials. The architecture is usually ok, but like you said Woomack, if they slowed down just a bit than we would see higher grade CPUs come out. Give me a $1000 CPU every 5 years but at least test it for several years before releasing it.

 

[mackerel]

No one expects CPUs to be perfect, well, outside of some questionable lawyers anyway. Part of the pain of original Ryzen launch was that bios was at a minimal working level. The improvements, in particular to ram support over following months were like night and day. Imagine how much more hype there would have been in enthusiast circles if that had been available earlier? 2nd gen seems to have gone much smoother as a relatively minor update.

Security related issues aside, which are appearing all the time, I'm not aware of recent Intel or AMD bugs. The only Ryzen bug I half cared about was that early samples (of which both of mine are potentially affected) was the SMT stability bug. It was found, it was fixed in silicon. I didn't care about it enough to do anything about it, although apparently AMD would replace on request those affected.

What I've seen people do where they value stability is to buy older generation products. That way you know it has been time tested and likely any bugs have been worked out, not to say new ones can't be discovered. If you live on the cutting edge like many of us here, the chances of encountering undesired behaviour is increased.

 

[Dolk]

I wasn't alluring to just security bugs. There have been several architecture bugs in Skylake-X/Xeon. They still pop up as more and more are used in the data center world.

 

[mackerel]

I suppose unless it is potentially likely to affect average users, it doesn't hit the news, so I'm not up to speed on flaws outside of the ongoing spectre stuff.

 

[notarat]

So it hits 5.1GHz on LN2...

Maybe 4.2GHz 24/7 on good liquid? Guesses?

 

[EarthDog]

That many cores at 4.2 GHz on water? Maybe. 4-4.2GHz

 

[notarat]

That many cores at 4.2 GHz on water? Maybe. 4-4.2GHz

Boost Clock is 4.2 so I figure with good water one could keep it at boost 24/7

 

[Scu84St3v3420]

The 2950X coming stock @ 16C/32T 3.5 GHz Base/4.4 GHz Boost... I would say that paints a pretty good picture for 7nm Ryzen chips even if they are up to 16 cores.

 

[bigtallanddopey]

The 2950X coming stock @ 16C/32T 3.5 GHz Base/4.4 GHz Boost... I would say that paints a pretty good picture for 7nm Ryzen chips even if they are up to 16 cores.

That would mean a boost clock of around 4.7-4.8 for the 3700x (if that is what it will be called) if you look at the current chips speeds etc.

 

[Scu84St3v3420]

That would mean a boost clock of around 4.7-4.8 for the 3700x (if that is what it will be called) if you look at the current chips speeds etc.

The 2950X is on 12nm process, so it could be even higher than that plausibly. I am excited to see what comes next, seriously got me considering an upgrade to Zen 2 when it comes so far, but I am one of those people who like being on the cutting edge.

 

[Dolk]

I suppose unless it is potentially likely to affect average users, it doesn't hit the news, so I'm not up to speed on flaws outside of the ongoing spectre stuff.

Most won't be talked about because they are protected under NDA or similar.

 

[bigtallanddopey]

The 2950X is on 12nm process, so it could be even higher than that plausibly. I am excited to see what comes next, seriously got me considering an upgrade to Zen 2 when it comes so far, but I am one of those people who like being on the cutting edge.

I forgot it was still on 12nm. And yeh I will be upgrading my 1600 to the 3600. Although as this thread has said it probably won’t give me big performance gains

 

[Scu84St3v3420]

I forgot it was still on 12nm. And yeh I will be upgrading my 1600 to the 3600. Although as this thread has said it probably won’t give me big performance gains

Overclocking my 2600X to 4.2 GHz all-core has me 23% SMT and 33% single above a stock 1600... That is fairly significant to me, let alone adding a plausible +1000 MHz to the boost from the 1600, add in 15% IPC improvement over Zen+ and I would say those are big performance gains. But I'm still fresh to the game.