Speculation: Spring refresh for Ryzen

[Vattila]

As I've posted elsewhere, I think AMD fell short of their frequency targets for the Ryzen 3000 series. A later launch than expected, statements relating to frequency challenges, Max Boost redefinition, boost issues, 3900X shortages, and now the 3950X delay from September to November, do all corroborate this. I guess they aimed for 5 GHz, but fell ~8% short.

However, TSMC is firing on all cylinders. Amongst a number of process roadmap announcements, they now offer N7P as a refinement of their N7 process on which the Zen 2 chiplet is built. N7P is compatible with N7 design rules, so it should provide a fairly simple and cost-effective opportunity to optimise existing designs.

With that in mind, is there a chance that AMD will do a spring refresh based on a faster stepping of the CPU chiplet?

 

[Thunder 57]

Read again. MANUAL overclocking.
Boostgate what? The clock rates that AMD have improved on with their most recent bios?
So performance would actually be **up** from the reviews. That boostgate?
Did you really expect large multinational companies to continue to give you a free lunch forever?

I think you should check up on Watergategate.

 

[Thunder 57]

So it is OK for AMD fans to brag about "winning" the "core wars", but intel should "shut up" about their frequency advantage? No double standard there, right?

You are such a bum with that stupid username. Who cares about frequency? The P4 clocked very high, but it was so bad it was tossed for a lower frequency, higher performing chip.

 

[VirtualLarry]

Where's that on AMD anymore? Factory overclocks should not be celebrated by enthusiasts!

I'm not so sure that I would agree with that. Plenty of people pay extra for "OC edition" video cards. Enthusiasts included.

I don't manually OC my GPUs. Thought I do my CPUs. And I did experiment with a static all-core OC on my Ryzen R5 3600, but I found that I got better results overall, just letting it do it's "Precision Boost" thing on it's own.

 

[Fanatical Meat]

So it is OK for AMD fans to brag about "winning" the "core wars", but intel should "shut up" about their frequency advantage? No double standard there, right?

No it’s fine and honestly I find it amazing intel has squeezed so much performance out of an old design. They are simply at their limit right now.
I like intels ability to put out a quality product with superior compatibility but I also don’t like how lazy intel has become regarding releasing new products.

 

[Thunder 57]

No it’s fine and honestly I find it amazing intel has squeezed so much performance out of an old design. They are simply at their limit right now.
I like intels ability to put out a quality product with superior compatibility but I also don’t like how lazy intel has become regarding releasing new products.

I don't get the "superior compatibility" when AMD has AM4 and plenty of CPU's for it when Intel makes you need a new chipset all the time.Did you mean comparatively? Intel has done great stuff with 14nm and is comparable to AMD in some aspects today despite being stuck on 14nm++.

 

[Fanatical Meat]

I don't get the "superior compatibility" when AMD has AM4 and plenty of CPU's for it when Intel makes you need a new chipset all the time.Did you mean comparatively? Intel has done great stuff with 14nm and is comparable to AMD in some aspects today despite being stuck on 14nm++.

I mean historically better memory compatibility and just better quality control regarding motherboards. This isn’t meant to be a current observation it’s an over time observation

 

[Atari2600]

Not quite sure I follow you. Stability and temps are king in my house,. Is that what you were saying ? Uptime on linux boxes is virtually forever. It would be on windows if windows would quit patching everything, and rebooting.

Stability is not king in your house Mark. If it were you wouldn't be undervolting.

Unless you are proposing that large scale deployments should go around undervolting their CPUs to pull down temps (and thus cooling costs)?

I'm saying you have quite different freedoms to optimise as you see fit that commercial companies simply don't have.

 

[Atari2600]

I think you should check up on Watergategate.

Was that not ChilledWaterCoolerGate?

 

[Zucker2k]

Read again. MANUAL overclocking.

What else could it be about? My whole response was anti-factory overclocking. If your chip is auto-boosting higher than you can manually overclock, that's a problem for manual overclocking, is it not?

Just watch this space when the 9900KS drops with it's all core 5GHZ clocks and then we should know what people's true feelings are about factory overclocks.

 

[ondma]

You are such a bum with that stupid username. Who cares about frequency? The P4 clocked very high, but it was so bad it was tossed for a lower frequency, higher performing chip.

Personal insults and an irrelevant analogy. Is that the best you can do?

Edit. This "who cares about frequency" argument on this forum has reached absurd levels. Performance is obviously a combination of frequency and ipc. When the ipc is similar, as it is for Ryzen and intel, especially in gaming(my primary use), where intel at least as fast per clock, frequency certainly does matter.

 

[Markfw]

Stability is not king in your house Mark. If it were you wouldn't be undervolting.

Unless you are proposing that large scale deployments should go around undervolting their CPUs to pull down temps (and thus cooling costs)?

I'm saying you have quite different freedoms to optimise as you see fit that commercial companies simply don't have.

So the fact that it runs perfectly stable at 100% load 24/7/365 for months does not make it stable ?

Edit: I think (but can't prove it) that AMD set the vcore so high, so ONE core could hit 4.6 ghz for a nanosecond. I want ALL cores@4.1, so I can use less voltage.

 

[Atari2600]

What else could it be about? My whole response was anti-factory overclocking. If your chip is auto-boosting higher than you can manually overclock, that's a problem for manual overclocking, is it not?

Ugh. No.

1. Your original remark was around folks purchasing expensive motherboards which were supposedly useless since you declared "overclocking" dead.
2. Can you OC single cores manually? I don't believe so.
3. If you gain a better overclock from letting the algorithms tune per core and you just put more cooling and wattage to it, then who needs manual overclocking?

I never ever mentioned anything about overclocking being dead - I said manual overclocking is now redundant. I don't know why you cannot differentiate between the two.


edit: https://www.techpowerup.com/259688/...wareness-increase-single-threaded-performance

A further nail into the coffin of manual overclocking. When the OS, scheduler and CPU are working in harness to best exploit the silicon at a level of fidelity a one-size-fits all manual overclock cannot match, then manual overclocking is dying.

 

[Atari2600]

So the fact that it runs perfectly stable at 100% load 24/7/365 for months does not make it stable ?

It does for you.

But will you lose potentially tens/hundreds of thousands of dollars on an outage? If you stood to lose your house if the system ever crashed, would you still be undervolting?

Edit: I think (but can't prove it) that AMD set the vcore so high, so ONE core could hit 4.6 ghz for a nanosecond. I want ALL cores@4.1, so I can use less voltage.

Pretty sure Vcore scales with clock.

 

[Markfw]

It does for you.

But will you lose potentially tens/hundreds of thousands of dollars on an outage? If you stood to lose your house if the system ever crashed, would you still be undervolting?




Pretty sure Vcore scales with clock.

So you agree with me that if you don;t care about single core high ghz, undervolting is good to reduce temps ?

 

[Atari2600]

So you agree with me that if you don;t care about single core high ghz, undervolting is good to reduce temps ?

Yeah, its fine.

My point was and is - how many people do that though? Your niche of a niche of a niche! [undervolting/underclocking side of enthusiasts of HEDT]

Its quite possible you could count the number of people across the world that run Threadrippers in the same way you do on one hand!

 

[Thunder 57]

Personal insults and an irrelevant analogy. Is that the best you can do?

Edit. This "who cares about frequency" argument on this forum has reached absurd levels. Performance is obviously a combination of frequency and ipc. When the ipc is similar, as it is for Ryzen and intel, especially in gaming(my primary use), where intel at least as fast per clock, frequency certainly does matter.

I don't care what you think. Your handle already shows your bias. I like to hear all points of view, which is why I've never blocked anybody, but you are clearly biased.

 

[Markfw]

Yeah, its fine.

My point was and is - how many people do that though? Your niche of a niche of a niche! [undervolting/underclocking side of enthusiasts of HEDT]

Its quite possible you could count the number of people across the world that run Threadrippers in the same way you do on one hand!

I actually overvolt and overclock my 2990wx (one of them) to 3.8 ghz all the time. Its on custom water. It will do 4 ghz, but I don;t like the temps.

So I overclock/overvolt/undervolt, whatever it takes to get the best all-core speed on a particular box.

 

[Markfw]

Personal insults and an irrelevant analogy. Is that the best you can do?

Edit. This "who cares about frequency" argument on this forum has reached absurd levels. Performance is obviously a combination of frequency and ipc. When the ipc is similar, as it is for Ryzen and intel, especially in gaming(my primary use), where intel at least as fast per clock, frequency certainly does matter.

Your point of "Performance is obviously a combination of frequency and ipc. When the ipc is similar, as it is for Ryzen and intel, especially in gaming(my primary use), where intel at least as fast per clock, frequency certainly does matter." has a bit of truth.

But what you fail to see, is that the ghz is so small, and due to other reasons (big L3?) the end result is so close that at this level, it really doesn;t matter MUCH. Also, on Intel, to get those speeds, a lot has to go into the cooling and overclock. For AMD, at stock, its very close and sometimes wins on stock cooling, runs cooler, and is less expensive.

And yes, your username tells it all.

 

[Vattila]

Evidence is mounting that there will be — not a spring refresh — but a summer refresh of Ryzen 3000, coinciding with the anniversary of the 7/7 launch last year. Do you think it will amount to much? Just binning, or a new stepping, re-spin of the die, perhaps, using the learning from the first version on 7nm? We've seen what Intel has been able to squeeze out of their 14nm products, with process optimisation. Then again, not much to gain? Better to focus on Vermeer?

"We initially believed that Matisse 2 would end up being a new die for the 3100 and 3300X, but we later reversed our stance. Then, fellow YouTuber Moore’s Law is Dead corroborated our leak saying that Matisse 2 was a refresh and would be coming soon. Finally, just a few hours ago 2 more leaks came out that strongly confirm the idea of a Matisse 2 refresh: this slide from WCCFTech writer Hassan Mujtaba and this leak about a 3750X and 3850X launching July 7th."

Matisse 2 Is Actually Real, 3750X and 3850X To Launch In July

Technology Vision

adoredtv.com

Just heard a rumor that AMD would release better binned version Ryzen2 in coming summer (3600XT, 3700XT, 3800XT, 3900XT, 3750X, 3850X or something similar). Stay tuned.

 

[gdansk]

I don't see the point but I wouldn't be surprised if there are updated chips. After the 1600AF anything seems possible from AMD land.

 

[Vattila]

Here is a hypothesis: The Ryzen 3000 refresh is very much part of AMD's plan to cover the mainstream desktop market. Reason is that Zen 3 will be the start of the interposer era for CPU packaging, and it will initially be high-end and expensive, with initial capacity only sufficient to cover server (relatively low unit numbers) and high-end desktop (likewise). With announcement of Zen 3 in September, then launch and availability in October, or even later in the fourth quarter, and initially deployed in EPYC 3 "Milan" and Threadripper 4000 "Genesis Peak", with Ryzen 4000 "Vermeer" scheduled for CES 2021, alongside Ryzen Mobile 5000 "Rembrandt" (Zen 3 + Navi 2), the Ryzen 3000 refresh will meanwhile fight for dominance in the mainstream desktop market.

If this should turn out true, it will be interesting to see what AMD has done with the interposer for Zen 3. Is it active or passive? Has it enabled more CCDs in the package? Or perhaps space for on-package memory? And it will be interesting to see how much of an improvement the Ryzen 3000 refresh is. Can it take on Intel in single-threaded performance?

 

[amrnuke]

Evidence is mounting that there will be — not a spring refresh — but a summer refresh of Ryzen 3000, coinciding with the anniversary of the 7/7 launch last year. Do you think it will amount to much? Just binning, or a new stepping, re-spin of the die, perhaps, using the learning from the first version on 7nm? We've seen what Intel has been able to squeeze out of their 14nm products, with process optimisation. Then again, not much to gain? Better to focus on Vermeer?

"We initially believed that Matisse 2 would end up being a new die for the 3100 and 3300X, but we later reversed our stance. Then, fellow YouTuber Moore’s Law is Dead corroborated our leak saying that Matisse 2 was a refresh and would be coming soon. Finally, just a few hours ago 2 more leaks came out that strongly confirm the idea of a Matisse 2 refresh: this slide from WCCFTech writer Hassan Mujtaba and this leak about a 3750X and 3850X launching July 7th."

Matisse 2 Is Actually Real, 3750X and 3850X To Launch In July

Technology Vision

adoredtv.com

3750 and 3850
Could be ported to N7P for some minor performance bump
Doubt it, but wouldn't surprise me

 

[lobz]

Yeah, its fine.

My point was and is - how many people do that though? Your niche of a niche of a niche!

So is heavy OC and then caring about extreme strict stability, to be fair.

 

[dpnelson]

Here is a hypothesis: The Ryzen 3000 refresh is very much part of AMD's plan to cover the mainstream desktop market. Reason is that Zen 3 will be the start of the interposer era for CPU packaging, and it will initially be high-end and expensive, with initial capacity only sufficient to cover server (relatively low unit numbers) and high-end desktop (likewise). With announcement of Zen 3 in September, then launch and availability in October, or even later in the fourth quarter, and initially deployed in EPYC 3 "Milan" and Threadripper 4000 "Genesis Peak", with Ryzen 4000 "Vermeer" scheduled for CES 2021, alongside Ryzen Mobile 5000 "Rembrandt" (Zen 3 + Navi 2), the Ryzen 3000 refresh will meanwhile fight for dominance in the mainstream desktop market.

If this should turn out true, it will be interesting to see what AMD has done with the interposer for Zen 3. Is it active or passive? Has it enabled more CCDs in the package? Or perhaps space for on-package memory? And it will be interesting to see how much of an improvement the Ryzen 3000 refresh is. Can it take on Intel in single-threaded performance?

AMD is going to be selling Matisse dies for a long time with Rome. It's could be that they made a new stepping for the server market and are taking advantage of the fact that it clocks better to push out a minor update. It could also be that this was planned as a peace offering to B450 and X470 motherboard owners after they announced their original plan to not offer Ryzen 4000 support.

 

[LightningZ71]

I feel that this may be related to some of the more recent “F” Epyc SKUs that have appeared on the market for HPC and special purpose uses. It seems to me that SKUs such as 3750 and 3850 could imply that these are two CCD 8 active core processors (it’s possible that these are 2+2 ccds or 4+0 ccds). With the 2+2 arrangement, they could leave all the L3 cache enabled to give them the maximum possible cache per die to maximize single core performance. With the 4+0 arrangement, it’s possible that these are special binned chips that have an outstanding CCX where AMD decides to fuse off the second CCD to spread the heat out as much as possible.

There are interesting possibilities out there...