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

[Origami Cactus]

14 minutes ago, VegetableStu said:

Quick question: what are the chances that X399 gets an update? going off the current situation of needing to up the quality and design of the PCB itself just for PCIe 4 on X570 o_o

X499 and threadripper 3 is confirmed, but when will it come out?

End of year maybe.

I hope that they will atleast show something at june 10.

[RejZoR]

23 minutes ago, Stefan Payne said:

What Linus said in his last Video about AMD/Zen2 also looked like Zen2 might have been a good overclocker...

I find that unlikely. With Boost features, they'd already achieve that if they could out of the box. Would be stupid not to if silicon is capable. Days when CPU's were sold with huge clock reserves are long over. Intel boosts them as high as they go anyway and AMD basically does the same because they are different design. AMD is clearly chasing the AthlonXP dream again. CPU's may not clock as high, but at the clocks they have, they do more work. As it was showcased in the keynote where they compared same core count CPU's from Intel and AMD. Sure it depends on workload and all that, but if they could make a same core count CPU with lower clock do work faster than Intel's, they certainly are up to something. Besides, 4.6GHz isn't that far away from 5GHz.

[Stefan Payne]

1 hour ago, VegetableStu said:

Quick question: what are the chances that X399 gets an update? going off the current situation of needing to up the quality and design of the PCB itself just for PCIe 4 on X570 o_o

Very high.

Lisa said that Threadripper 3000 series will come.

44 minutes ago, RejZoR said:

I find that unlikely. With Boost features, they'd already achieve that if they could out of the box.

Why else would they be talking about Current in terms of OC??

That only makes sense if OC makes sense. 
Maybe AMD can get the Clock record back from AMD?

 

44 minutes ago, RejZoR said:

Would be stupid not to if silicon is capable.

No, because:

a) if they don't need to, why should they? If the specs they have specified for their products is good enough to beat the Competition, what more can they want?

b) higher frequency means higher power consumption. 

So it might be possible that Zen2 goes to 5GHz but with a similar power consumption as Intel. Meaning 250W or so.

They probably don't want that...

 

44 minutes ago, RejZoR said:

Days when CPU's were sold with huge clock reserves are long over.

Yeah because 100MHz back in the day were relatively almost 50% more.

100MHz today are 0,025%.

 

Or with cars:
Its easy to double your travelling speed from 40km/h to 80km/h.

But if you're already at 100km/h, doubling that is pretty hard and uneconomically...

44 minutes ago, RejZoR said:

Intel boosts them as high as they go anyway and AMD basically does the same because they are different design. AMD is clearly chasing the AthlonXP dream again.

Well, the TDP is different on both.

While Intel does not include Turbo no more, AMD to a certain extend does. 

And also AMD doesn't have the support from people and the Media, so they can't really do that.

With Intel, if they do that, nobody really cares as you can see here. And there are also a ton of people defending the Intel TDP specification.

I wonder if they'd do the same on AMD...

 

44 minutes ago, RejZoR said:

CPU's may not clock as high, but at the clocks they have, they do more work. As it was showcased in the keynote where they compared same core count CPU's from Intel and AMD.

We can't say that now.

Have to wait ~6 Weeks to be sure how it really is and how well the CPU overclocks...

 

44 minutes ago, RejZoR said:

Sure it depends on workload and all that, but if they could make a same core count CPU with lower clock do work faster than Intel's, they certainly are up to something. Besides, 4.6GHz isn't that far away from 5GHz.

only ~9% or so...

But that's also the reason why I don't OC no more.

You can get 10% out of it for +50% higher power consumption...

[GoldenLag]

1 hour ago, Johnny Who said:

When can we expect reviewers to received preview kits?

they will probably have kits to review a week or 14 days before launch, depends on what AMD wants to do

[Connor Price]

20 hours ago, _Syn_ said:

Placebo.. or something else

 

Spectre and Meltdown were never a big hit on Coffee Lake CPU's, maybe a couple of percents at most, but now with Windows 10 May Update it's completely mitigated and performance is back to normal

 

And you're not impacted at all with Zombieload and all the MDS stuff because your CPU doesn't even have hyperthreading

 

Not saying you shouldn't upgrade, that's up to you, but don't fool yourself

It's a combination of a lot of things that all seemed to stem from the whole spectre and meltdown thing. The bios updates for my particular board with the patches (Prime Z370) caused constant stability issues and just overall weird behaviour. I could never get my system quite the way it was prior to those bios updates. Whether it be blue screens, lag, hiccups or whatever.

 

Now obviously that's not all down to Intel, a lot of that is down to Asus or perhaps just my particular board as I didn't see many other people with issues like mine after the patches.

 

In terms of performance loss I probably could've worded it better. As you said, performance hit was a couple of percent at most but it was the constant hitches and lock ups that I never had prior to the patches that really pissed me off. To be fair those issues have all sorted themselves out by this stage and I'm probably not going to touch a Bios update on this board for the forseeable future in case something like that happens again.

 

And yeah I'm aware that Zombieload has no effect on me at all but in terms of Intels marketing and PR it can't be good. That's more of the way I was taking that one.

[Doobeedoo]

Amazing improvement! Nice prices as well. I look forward to see benchmarks for sure. Planning to get R7 3800X myself, I really wonder how much clocks can be pushed further.

[YaBoiWill]

NGL was kind of hoping for a 5ghz chip, but that's taught me to be skeptical of "leaks" and to actually reality check the facts. 5ghz might be achievable on the best delivering board and with the luckiest chip underwater perhaps. 

[Senzelian]

15% IPC improvements? So finally on the level of Intels last gen? 

I guess at least the price is right.

[Drak3]

1 minute ago, Senzelian said:

15% IPC improvements? So finally on the level of Intels last gen? 

I guess at least the price is right.

Skylake+ (Every generation since has had the same IPC in existing instructions) was ~5% ahead of Ryzen (which was tied with Haswell). A 15% increase puts Ryzen 3xxx ahead of Intel chips clock for clock.

[Trixanity]

14 hours ago, Stefan Payne said:

2x4 Core on 2 Chiplets are 8 Cores as well.

The difference is:
For 1x8 you need one fine die - with no defects

For 2x4 you need two crap dies - half of the die might have defects but its still salvagable.

 

I'd say two dies are more probable on the 8 Cores than 1 Die because of cost and "salvaging" half dead Cores...

Though it might be possible that the 3700X is two die while the 3800X is one. We have to wait and se.

Pretty much confirmed that all of them are single chiplet except R9 which is obviously double.

 

The broken chiplets are what you'll get in R5 with 6 cores (and in R9 apparently or at least disabled cores) and probably R3 with 4 cores.

 

The evidence for it is twofold: we only saw single chiplets until R9 was introduced (which means it's different) and R9 has double the cache. Would have to be a lot of stuff disabled or broken for two dies to hit the specs required for R7. Yields would be pretty bad if two dies are necessary considering the product stack. Of course there's a chance I'm wrong but it seems unlikely all things considered. It does appear to be a waste of silicon to use two chiplets for R7 when you have an R3 stack you need to roll out.

 

Keep in mind these chiplets are small (80 mm^2) and the 7nm process should be pretty damn mature at this point. $329 should be quite fine if we consider the size. If it isn't fine then that 275 mm^2 GPU with GDDR6 should be one expensive product.

 

Edit: have yet to find a single source not saying 1 chiplet for R3-R7 and 2 for R9. 

[Senzelian]

4 minutes ago, Drak3 said:

Skylake+ (Every generation since has had the same IPC in existing instructions) was ~5% ahead of Ryzen (which was tied with Haswell). A 15% increase puts Ryzen 3xxx ahead of Intel chips clock for clock.

Sure, that's probably true, but I'm still disappointed because of the difference in clock speed.

It just isn't enough to be ahead of Intel, even tho Intel didn't even do anything yet. As you said, it's still the same old architecture.

[The Benjamins]

2 minutes ago, Senzelian said:

Sure, that's probably true, but I'm still disappointed because of the difference in clock speed.

It just isn't enough to be ahead of Intel, even tho Intel didn't even do anything yet. As you said, it's still the same old architecture.

So you are not happy that amd beat Intel in single core performance and multi core?

[Anomnomnomaly]

I'm not that bothered by the stated core and boost clocks... I'm more interested in if you can boost them on all cores ALL of the time. My 2600X is 3.6/4.25Ghz, but will only really sit at 4.1Ghz on all cores all the time.

 

My old FX8350 4ghz/4.2 boost would sit at 4.4Ghz happily all day long, and my other system which was running an FX6300 (3.5Ghz) would happily sit at 4.3Ghz all day long.

 

Now I know the FX line wasn't very good for OC... and the Zen+ still isn't great... So to truly compete with and beat intel... those boosts need to be achievable at all times on all cores.

 

If they can manage that... I'm sold.

[Dabombinable]

4 minutes ago, Anomnomnomaly said:

I'm not that bothered by the stated core and boost clocks... I'm more interested in if you can boost them on all cores ALL of the time. My 2600X is 3.6/4.25Ghz, but will only really sit at 4.1Ghz on all cores all the time.

 

My old FX8350 4ghz/4.2 boost would sit at 4.4Ghz happily all day long, and my other system which was running an FX6300 (3.5Ghz) would happily sit at 4.3Ghz all day long.

 

Now I know the FX line wasn't very good for OC... and the Zen+ still isn't great... So to truly compete with and beat intel... those boosts need to be achievable at all times on all cores.

 

If they can manage that... I'm sold.

The FX line was actually fantastic for overclocking....before the FX9000 series launched, getting very close to or over 5GHz wasn't unheard of.

[Senzelian]

Just now, The Benjamins said:

So you are not happy that and beat Intel in single core performance and multi core?

I'm happy that they did, but I wish they beat them by a larger margin.
1% better single core performance and 2% better multi-core performance compared to a now almost 1-year-old CPU, which is based on an aging architecture just isn't interesting, even tho this is AMD's own chart!

 

 

That's why I said, at least the price is right.
 

I think PCI-e 4 and the security issues on Intel chips are a much more compelling reason to buy a 3rd gen Ryzen CPU. 

[kiwibacon]

On 5/27/2019 at 3:33 PM, rockking1379 said:

Personally I think the 3800X is gonna be the low seller. The price increase versus what you get is marginal from the specs given so far over a 3700X. But maybe they will pull an intel and reduce pcie lanes or something. 

3800x is going to be the best gaming chip in the line up. Clearly 3900x is cut down 16 core. While 3800x is higher binned 3700x

But is that worth the money over the 3700x will be the question. 

[Ithanul]

22 hours ago, Trixanity said:

Barring any hidden latency improvements, the latency should be a mixed bag. There will be no NUMA. Near latency should be worse but far latency should be much better. To clarify: those terms should not exist on Zen 2 but we're comparing to the previous products which did have near and far latencies. The latencies should be uniform and consistent hence no NUMA or NUMA-like shenanigans but there should be a penalty for moving memory controllers off die but it'll be easier on software and overall the trade-off should be good. The bigger caches should also help nullify the performance deficit from any added latency.

 

Of course this is speculation because we haven't seen any hard data on this and there might be some curve balls in there.

Reason I am looking forward to seeing layout breakdowns and how the 3900X handles in third party testing.  Should give an idea what the next Threadrippers will do.

[The Benjamins]

20 minutes ago, Senzelian said:

I'm happy that they did, but I wish they beat them by a larger margin.
1% better single core performance and 2% better multi-core performance compared to a now almost 1-year-old CPU, which is based on an aging architecture just isn't interesting, even tho this is AMD's own chart!

 

 

That's why I said, at least the price is right.
 

I think PCI-e 4 and the security issues on Intel chips are a much more compelling reason to buy a 3rd gen Ryzen CPU. 

Let's just forget about the 3900x which is what competes against the 9900k

 

[TahoeDust]

29 minutes ago, The Benjamins said:

So you are not happy that amd beat Intel in single core performance and multi core?

When did they beat Intel in single core?  Have we seen any single core scores?

 

[Drak3]

1 minute ago, TahoeDust said:

When did they beat Intel in single core?  Have we seen any single core scores?

 

AMD is touting that they edge out Intel by a few percentage points.

[Origami Cactus]

10 minutes ago, TahoeDust said:

When did they beat Intel in single core?  Have we seen any single core scores?

 

The graphic literally above you.

[Senzelian]

18 minutes ago, The Benjamins said:

Let's just forget about the 3900x which is what competes against the 9900k

Yeah right, I'm sure that 0.1Ghz difference in the boost clock is going to change everything.

It won't take Intel a lot of effort to beat that again and I'm saying this as someone that wanted to upgrade to a Ryzen 12 core from my current Ryzen 2700.

[The Benjamins]

Just now, Senzelian said:

Yeah right, I'm sure that 0.1Ghz difference in the boost clock is going to change everything.

It won't take Intel a lot of effort to beat that again. 

Lol the 4 extra cores gives you 50% MC dude.

[Origami Cactus]

3 minutes ago, Senzelian said:

Yeah right, I'm sure that 0.1Ghz difference in the boost clock is going to change everything.

It won't take Intel a lot of effort to beat that again

When was the last time AMD beat intel in single core? 2006?

That is why it is such a big deal that AMD has beaten them at all.

[Anomnomnomaly]

37 minutes ago, Dabombinable said:

The FX line was actually fantastic for overclocking....before the FX9000 series launched, getting very close to or over 5GHz wasn't unheard of.

I never really had a strong enough MB to push them further... I've put my old 6300 into my mums PC as a mini upgrade on an old ASUS M5A mATX board and it's still happy running @ 4.3ghz on stock voltage. The 8350 runs at 4.4ghz on stock voltage too. If I'd had a 9xx chipset board, I'd have pushed further.