Speculation: Spring refresh for Ryzen

Page 7 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Vattila

Senior member
Oct 22, 2004
799
1,351
136
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?

1569662809984.png
 
Last edited:

AnandThenMan

Diamond Member
Nov 11, 2004
3,949
504
126
I think this is more about marketing than anything else, keeping Ryzen fresh in the minds of consumers as the 'newest' desktop cpu's.. Also as others have stated it forces re runs of the benchmarks showing Ryzen is the best all round choice.
Intel's gaming and ST lead is going to be minimal.
Sounds like good marketing to me.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,353
10,050
126
Well, I'm not going to say the new XT release are a bad thing in any way, I'm a little under-whelmed.

I'm also a bit under-whelmed about what I'm seeing with my newly-purchased "Fresh batch" Ryzen R5 3600. I had to boost the Vcore to 1.3500V to keep it stable at 4.20Ghz all-core, under a PrimeGrid load. I couldn't hit 4.3 or 4.4Ghz at all with it, even with 1.35V, and even without the DC load. No Windows boot.

So either I missed my purchase window, and all of the higher-clocking 3600 CPUs are going towards the 3600XT, with the higher single-core boost (which ANY core can boost too, correct?) At least according to HWMonitor when I first got my OG 3600 CPU, my max clock for every core would eventually show 4.2Ghz, which was the max boost as the time.

So, likewise, if the new 3600XT can single-core boost to 4.5Ghz, and that means ANY core, then there's hope for an all-core 4.5Ghz OC, though I wonder exactly how much you would have to juice the CPU to survive an all-core AVX2 load like PrimeGrid, and whether the processor would survive that. :p

Don't get me wrong, I'm happy with my new-batch 3600, I couldn't even get the OG 3600 to clock all-core to 4.00Ghz under PrimeGrid load, even under 1.3625V or whatever.

I'm on 240mm LC too. (Which I recently blew the dust out of, and at an all-core OC of 4.20Ghz, @ 1.28125V, which was sufficient for a 100% CPU load doing WCG, and had temps of 71C max, it wasn't enough for PrimeGrid. I hit 81C @ 1.3500V doing PrimeGrid, which was the lowest voltage that I've tried so far that doesn't spontaneously reboot on me.)
 
  • Like
Reactions: Vattila

blckgrffn

Diamond Member
May 1, 2003
9,128
3,069
136
www.teamjuchems.com
Well, I'm not going to say the new XT release are a bad thing in any way, I'm a little under-whelmed.

So, likewise, if the new 3600XT can single-core boost to 4.5Ghz, and that means ANY core, then there's hope for an all-core 4.5Ghz OC, though I wonder exactly how much you would have to juice the CPU to survive an all-core AVX2 load like PrimeGrid, and whether the processor would survive that. :p

Don't get me wrong, I'm happy with my new-batch 3600, I couldn't even get the OG 3600 to clock all-core to 4.00Ghz under PrimeGrid load, even under 1.3625V or whatever.

Running BOINC projects my 3600 boosts to 4.05 (all cores) and sits there at ~64C, all I did was set the XMP profile to enabled, zero other tweaks. I suppose I should test PrimeGrid.

Anyway, I somewhat agree. Even a 100 mhz base core clock uptick to go with the new turbo speeds would have been more compelling.

How long are we really seeing the 3800xt/3900xt hitting that 4.7ghz anyway? Is that supposed to be a single core sustained number?
 

Pumice

Member
Jan 17, 2011
63
1
66
Would most of you choose the Ryzen 3900XT over a Core i9-10900K ? Both are around $500.

Would the Ryzen 3900XT comfortably overclock to 5.3ghz like the I9-10900K?
 

DrMrLordX

Lifer
Apr 27, 2000
21,643
10,859
136
To achieve a higher frequency across all cores at the same time would either require an improvement in efficiency or different limits compared to the direct precursors.

I doubt it's going to boost beyond 142W, but we'll see. Remember that some people getting "late" samples of the 3600 are seeing up to 200-300 MHz higher clockspeed limits in all-core OCs. If that's the case, the boost algo may take advantage of that potential in workloads where clocks dipped due to high current/low voltage.

What is the core limit within a CCX for Zen 2?

Will any of these XT models have all the cores on one CCX and thus give the sorts of latency and performance improvements that we saw with the 3300x?

So far as any one knows, 4 cores per CCX is the "limit". I doubt they taped out an entirely different design but hey, you never know.

Don't like this - too much like an Intel move.

That was my take on it. It does not bode well for AMD customers when AMD is rehashing their existing core instead of hyping the next core. Granted, AMD also rehashed Piledriver (8320e, 8370e, 8370) and Kaveri (Godavari). Oh and they rehashed Carrizo (Bristol Ridge). But those were moves they made when they had nothing better to offer at the time.
 

misuspita

Senior member
Jul 15, 2006
401
452
136
So, this is essentially a little bit more single-thread performance for a little bit more cash. With a max of 4.7GHz, the 3900x 3900xt inherits the single-thread performance of the 3950x. What this says, for now, is that 4.7GHz is the ceiling for TSMC's 7nm. It's not clear if this is N7P as some members were suggesting earlier.
Also, if AMD is going to go to this length to monetize a 100MHz/200MHz single core boost, does this have any bearing on when Zen 3 will release?
At lauch, the 3xxx series did relatively poor with regards to the advertised max frequencies. so that "up to" was microseconds "up to". If they can get 3600XT work at 4.3-4.5GHz all the time, it's a huge bonus. Big IF...

Edit - I see VirtualLarry said same thing
 
  • Like
Reactions: Vattila

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,249
136
That was my take on it. It does not bode well for AMD customers when AMD is rehashing their existing core instead of hyping the next core. Granted, AMD also rehashed Piledriver (8320e, 8370e, 8370) and Kaveri (Godavari). Oh and they rehashed Carrizo (Bristol Ridge). But those were moves they made when they had nothing better to offer at the time.

So you'd suggest AMD kills their current sales by spouting out some performance figures from the next core? Isn't that the thing you'd want to do if your current core sucks and sales are lacking?
 

DrMrLordX

Lifer
Apr 27, 2000
21,643
10,859
136
So you'd suggest AMD kills their current sales by spouting out some performance figures from the next core? Isn't that the thing you'd want to do if your current core sucks and sales are lacking?

Building hype for your next gen product while clearing inventory on the existing product at reduced prices is a perfectly acceptable way of doing business. And it's pretty much what they did for Pinnacle Ridge. By this time last year, we already had some substantial leaks on Matisse performance.

For Vermeer, we have nothing.

Also, sadly, while Matisse Refresh seems to have been "announced" today, we have zero review samples going out, and availability isn't until July 7th. So nobody can review the product to see what the deal really is with MT clockspeeds.
 

Rigg

Senior member
May 6, 2020
472
974
106
Unless these SKUs have some XTra OC headroom in them it's going to be an XTremely boring refresh.

Considering my 7th week of 2020 3600 is real bench stable at 4.55/4.5 @ 1.28v load and 3dMark firestrike stable at 4.7 /4.65 @ 1.35, I'd say it's a pretty safe bet these new SKU's will have some OC headroom. At least the 3600XT and 3800XT should. The 3900XT will probably be too thermally limited. At the very least the 3900XT should have more closely binned chiplets and hopefully get better clocks on the second CCD. I can get my launch 3900x to OC per CCX 4.475/4.4/4.3/4.325 @ 1.325 load but it hit's mid 80's in real bench on custom water. I ran it daily at 75mhz and 100mv lower than that. A 3900XT with 2 CCD's as good or better than my 3600 would be pretty beast even if it is a bit thermally limited. All of these chips would be a lot more overclock-able if they'd give you a bios option to throttle them in high current loads when running fixed clocks.
 
Last edited:

thigobr

Senior member
Sep 4, 2016
232
166
116
I am eager to replace my 1700 but I am not touching these new XT parts for their current prices this late in the game...

For new buyers, if they do a quick research, they would rather best spend their money on any of the current non-XT offerings.

I just hope this refresh doesn't imply any Zen 3 delays!
 

amrnuke

Golden Member
Apr 24, 2019
1,181
1,772
136
Would most of you choose the Ryzen 3900XT over a Core i9-10900K ? Both are around $500.

Would the Ryzen 3900XT comfortably overclock to 5.3ghz like the I9-10900K?
Clock speeds aren't everything.

Case in point, the 10900K lost to the original 3900X in many benchmarks, even when overclocked to 5.1GHz all-core for multithreaded applications, even when allowed single-core 5.3GHz boost.

So, adding 2-4% to 3900X performance, that's a good reason why someone might choose it over the 10900K.

All depends on your use case. Find your use case, find the benchmark that is the best analog, and do your research. Who cares if the 10900K clocks to 5.3GHz if it's slower in the app you need speed in?
 

TitusTroy

Senior member
Dec 17, 2005
331
39
91
early rumors had Base clocks being a bit higher on the XT chips but they're the exact same as their X counterparts...the new CPU's have only a 100MHz increase in Boost clocks...I don't see the point unless there's something else I'm missing...a 3600XT with a Base clock of 4.0 would have been the bee's knees
 

VirtualLarry

No Lifer
Aug 25, 2001
56,353
10,050
126
early rumors had Base clocks being a bit higher on the XT chips but they're the exact same as their X counterparts...the new CPU's have only a 100MHz increase in Boost clocks...I don't see the point unless there's something else I'm missing...a 3600XT with a Base clock of 4.0 would have been the bee's knees
I don't know about "base clock", but if they gave an all-core boost clock a bump up, from say, 4.0Ghz or thereabouts (my all-core clock with my OG 3600 @ stock), to 4.1 or 4.2Ghz (preferably 4.2Ghz), then that would give me a much greater reason to purchase a 3600XT, as most of my loads are all-core, and just a 100-200Mhz boost to single-core boost doesn't really interest me all that much.

I wonder if AMD couldn't, with some firmware and chipset driver / power-management tweaks, do something akin to Intel Turbo Boost 3.0, where Intel designates a particular core as the "Golden core", that can do a single-core boost higher than the rest, and will boost that, thermals and power permitting, on particularly heavy single-threaded loads, as I understand it, but not just on loads that ONLY load one core. IOW, it will ST boost using THAT core, for ST loads, ALSO WHILE using the other cores to their max MT boost clocks.

I know that AMD designates a "Golden core" for each chiplet, and uses those for ST boosts that go above a certain level, but it would be nice if, perhaps, hypothetically, that a 3600XT could boost multi-core to 4.2Ghz, while AT THE SAME TIME boosting a particular single-threaded load to 4.5-4.7Ghz on the "Golden core".

IMHO that would actually be perfect for gaming, and allow AMD the clear lead in gaming, because most game engines have a single CPU-heavy "master thread", and while they are multi-threaded, and can put loads on all available cores, there is benefit for allowing this CPU-heavy master thread to clock higher than the other cores/threads.

(I saw a post or maybe it was a YT video link posted here, that was purporting to describing dropping the priority of the "secondary" threads, to allow the "master" thread priority over them, to allow it to have as much CPU time as it needed to operate.
 
  • Like
Reactions: Vattila and Tlh97

DrMrLordX

Lifer
Apr 27, 2000
21,643
10,859
136
Remove cooler, reinstate launch prices and offer new buyers a 100 MHz extra boost clock which they already are getting on late non XT models anyway.
Hmmm... 🤔

I'm still willing to hold out for some actual product reviews, but yeah I'm not all that on board with this Matisse refresh thing.

There are concomitant price cuts, official and otherwise. Even the 3950X is cheaper today.

I don't think that's what he means. If AMD launches the 3900XT today @ $499, then they launch the 4900x in October at a higher price.

The 4900x should be $499, and the 4950x should be (I guess) $749.

Clock speeds aren't everything.

They are within the same product range. A 3900x sold last July can sustain all-core clocks of . . . well okay, MY 3900x can sustain all-core clocks of 4175 MHz in CBR20 MT. If the 3900XT can do 4300 MHz or higher all-core in CBR20 MT then yeah, that's an improvement, and it's something that might be possible. I would consider paying an extra $90 for that today were I choosing an AM4 CPU.
 
  • Like
Reactions: Vattila