AMD "Matisse Refresh" Processor SKUs Include 3900XT, 3800XT, and 3600XT

[btarunr]

Rumors of AMD refreshing its 3rd generation Ryzen desktop processor family are growing louder. On Friday (22/05), reports of the "Matisse Refresh" processor family surfaced, with talk of "Ryzen 7 3850X" and "Ryzen 7 3750X" processors headed for a June 2020 announcement followed by July availability. Turns out AMD has a different naming scheme in mind, targeted at wooing gamers. The company is reportedly bringing its "XT" brand extension over from its Radeon graphics card family over to the Ryzen line.

There are three SKUs AMD is developing, the Ryzen 9 3900XT, the Ryzen 7 3800XT, and the Ryzen 5 3600 XT. All three are likely to retain core counts of the SKUs they are displacing from current price points - with the 3900XT likely being a 12-core/24-thread part; the 3800XT an 8-core/16-thread part, and the 3600XT a 6-core/12-thread part. AMD is likely to give the three a major clock speed increase to shore up gaming performance. It won't surprise us if AMD tinkers with boost algorithms, either. GIGABYTE has already referenced "Matisse Refresh" in its motherboard product roadmaps, which adds plenty of credibilty to this rumor. With "Zen 3" based 4th gen Ryzen processors unlikely to relieve the embattled 3900X, 3800X, and 3600X in the wake of Intel's 10th gen Core "Comet Lake" launch until Q4-2020, it makes sense for AMD to plan a product stack refresh to bolster its competitiveness. AMD is reportedly planning a June 16 product announcement, followed by July 7 availability.



View at TechPowerUp Main Site

 

[btarunr]

My expectations:

• 400-700 MHz speed bumps
• Mothballed AMD XFR tech could be revived. With the right coolers the XT chips could hit/cross 5 GHz, but AMD will be careful not to write "5 GHz" on its specs sheet
• XFR and Intel TVB are similar in what they do (short-burst opportunistic auto-overclocking dictated by cooling)
• Possible PBO revival

 

[Vya Domus]

My expectations:

• 400-700 MHz speed bumps

No way, that's way outside the optimal power/frequency curve of current 7nm Zen chips. 200-300 Mhz is way more realistic (for base clocks that is) and as far as boost clocks 100-200 is the best you can expect.

 

[Deleted member 185088]

Probably better binned with a 100/200mhz bump in speed.
I believe AMD needs to wait till devs start optimising for the Zen architecture, they'll get a huge boost with the PS5 Series X.

 

[Crackong]

Maybe an optimized stepping ?
My 3900x manufactured after week 32 2019 can do 4.5GHz at 1.35v

With some Tweak maybe they could add ~200MHz to base / boost clocks ?

 

[phanbuey]

if it's cheaper and faster maybe, but why bother to get one now if you're they're gonna release a faster chip in 3 months.

 

[Flanker]

In a couple year's time the 5700XT could mean both a GPU and CPU

 

[EzioAs]

if it's cheaper and faster maybe, but why bother to get one now if you're they're gonna release a faster chip in 3 months.

Because there are people who want things now?

 

[Xuper]

Why Can't CPU stay stable at 4.6ghz for sustained speed ?on all cores , they can only stay at 4.3ghz.

 

[TigeronStarfire]

These are potentially tempting! I have a first gen Ryzen 7 1700 and was thinking of upgrading my CPU. I figured I'd hold off until AM5, and I still may, but knowing I could also use these in my B350 board is nice.

 

[Deleted member 185088]

Why Can't CPU stay stable at 4.6ghz for sustained speed ?on all cores , they can only stay at 4.3ghz.

Good question, I think someone with more knowledge could go into the electric details on why this is the case.
Maybe someone like Ian Cutress of Anandtech can explain it in an article.

 

[Vya Domus]

Why Can't CPU stay stable at 4.6ghz for sustained speed ?on all cores , they can only stay at 4.3ghz.

Power limitations, this is the best you can do on this 7nm node.

 

[HwGeek]

If this is true then I am wondering if the Ryzen 4000 will remain the same price as Ryzen 3000...
I think this is a hint that the Ryzen 4000 will launch at higher prices then the current parts[launch prices], if the rumor's are true and they can get ~25% ST performance uplift(IPC+Clocks) then there is no more reason to keep selling cheap.

 

[Xuper]

Power limitations, this is the best you can do on this 7nm node.

I doubt it's from node.hell even with 14nm , First Ryzen couldn't pass 4.1Ghz (My 1600X can stay at 4ghz for 24/7 but my CPU cooler is suck) . Right Now with 7nm , Third Ryzen can't pass 4.3ghz ( OC all cores not XFR or PBO ).
One told me It's because Of L3 cache which can't do synchronized with Cores.

 

[Vya Domus]

I doubt it's from node.hell even with 14nm , First Ryzen couldn't pass 4.1Ghz.now with 7nm , Third Ryzen can't pass 4.3ghz ( OC all cores not XFR or PBO ) , My 1600X can stay on 4ghz for 24/7 but my CPU cooler is suck.
One told me It's because Of L3 cache which can't do synchronized with Cores.

Different nodes, different architectures and Zen 2 has way more cache, cache scales horribly with size in terms of power. 7nm wasn't like other nodes before which came with a intrinsic reduction in power and lower operating voltage, Zen 2 hardly uses any less voltages than Zen 1, AMD chose higher clocks and that came at the cost of power. I am sure there are problems with keeping 3 separate dies in perfect sync as you increase clocks.

Not to mention the issue of thermal density.

 

[TranceHead]

These are potentially tempting! I have a first gen Ryzen 7 1700 and was thinking of upgrading my CPU. I figured I'd hold off until AM5, and I still may, but knowing I could also use these in my B350 board is nice.

Pretty sure refresh won't work in a B350 board

 

[Mussels]

These chips were probably AMD's plans before the backlash over B450 not support 4000 series... "hey you cant upgrade to those, but have a refresh to upgrade to!"

 

[Camm]

There was a significant bin difference between the 3900 and 3950. I expect 3950 binning to be pushed into the XT line.

 

[Space Lynx]

i just dont understand this, i mean september is only 3 months away and that is confirmed release target date of rdna 2 gpu, ampere, they would have better off saving rnd, production lines and moving their november 4000 series cpu release to september, so people could do a full new build in time for cyberpunk 2077

 

[Vya Domus]

There is nothing strange about this, it's just a refresh, the products are pretty much the same. Refreshes can arrive at any time.

 

[ShurikN]

Is it possible that these chips are on 7nm+, the peoces that's supposed to go into Ryzen 4000. Maybe AMD was experimenting with the node, see how it behaves, and then saw an opportunity to release it alongside b550?

 

[TigeronStarfire]

Pretty sure refresh won't work in a B350 board

The graph seems to imply I could, and this is probably it, since it lists A320. Still will wait and see for sure.

 

[ARF]

This is a disturbing and bad news. Does it mean something bad happened to Zen 3 and it is pushed back from September ?

This is a marketing disaster for AMD. People who buy today won't buy in 3 months, and they will be screwed and left sitting with much inferior products.

 

[TranceHead]

The graph seems to imply I could, and this is probably it, since it lists A320. Still will wait and see for sure.

Consider only the 3000 APU's are compatible with B350 (graph shows 3000G as compatible but 3000X as incompatible)

 

[EzioAs]

This is a disturbing and bad news. Does it mean something bad happened to Zen 3 and it is pushed back from September ?

This is a marketing disaster for AMD. People who buy today won't buy in 3 months, and they will be screwed and left sitting with much inferior products.

I'm pretty sure I just read recently AMD said that Zen 3 is on track. Don't know where you're getting this disturbing news idea. I forgot the source though. It could be TPU.

Some of you act like CPU refreshes like this is new. It's been done for years and will probably happen again in the future.