Info 64MB V-Cache on 5XXX Zen3 Average +15% in Games

Page 47 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Kedas

Senior member
Dec 6, 2018
355
339
136
Well we know now how they will bridge the long wait to Zen4 on AM5 Q4 2022.
Production start for V-cache is end this year so too early for Zen4 so this is certainly coming to AM4.
+15% Lisa said is "like an entire architectural generation"
 
Last edited:
  • Like
Reactions: Tlh97 and Gideon

coercitiv

Diamond Member
Jan 24, 2014
7,354
17,422
136
I'd even go further and say this is the most exciting chip Intel has had in a long while, maybe even since Zen was released.
I'm not trying to downplay the chip, I bought one. My point was Alder Lake needs time to reach it's potential in the market, things will probably change with the second part of the launch in 2022: by then boards will be plenty, coolers will come with proper brackets, Win11 will get fixes.
 

Hans Gruber

Platinum Member
Dec 23, 2006
2,516
1,357
136
I don't think this is true. AMD never had an overall performance lead over Intel until Zen 3 and they sold fine even when they were further behind.

If Zen3D didn't happen at all I don't think AMD would be any worse off if they had nothing new until Zen 4.

Are people really going to make a new build with Zen 3D on a platform that's got to future beyond that as opposed to going with Intel when you can get in at the start of one? AMD fans would hold out for Zen 4 anyways especially with the GPU market still being the way it is.
I am the wrong guy to be asking that question. I bought a rog Strix B450-F for $90 simply for the ability to upgrade my PC to Zen 3. I bought a 32GB kit of E-die for $104 last year on Amazon. I had two 16GB hynix kits that I was running in one PC. Put in the E-die in the B450-F and it runs 16-19-16-16 timings @ 3800ghz completely stable. My B350 was never stable with memory over 3600mhz. I bought the Hynix gold P31 NVMe (1.3 gen) 1tb drive, couldn't justify the price of B550 boards considering Zen 4 will be AM5. The BeQuiet case was $80. I had a new Cooler Master 240MM AIO RGB sitting in the shelf.

My old P400 Phanteks case and AIO 240mm now run with the old 3600mhz 16GB kit and my B350 and the original R3 1200 quad core Ryzen CPU. When I get a Zen 3 CPU, this R5 3600 goes into the old build. I bought replacement fans for my GTX 970 for $11. The card runs like new again. Almost forgot I got a 550w evga GA gold power supply for $40 within the last year. So I had a lot of new components sitting on the shelf.

So basically I spent $170 + tax for a second system plus whatever the Zen 3D chip costs. I know a lot of people here use their systems for crypto mining or trying to find cures for cancer or folding teams. Zen 3 overcomes the memory latency issues that plagued Zen 2 and previous Zen CPU's. The Zen 3D talks of 15% gaming improvement. So from a Ryzen 3600 to a Zen 3 is 20% better gaming performance plus 15% for Zen 3D which should be good for 35% better gaming performance over Zen 2 for gaming.

I will keep the Zen 3D CPU for two or three years. At that point the DDR5 prices should be lower. The BeQuiet 500DX is pretty great considering it comes with 3x 140mm fans and can handle 280mm AIO's on the top of the case and 360mm in front. This is the 1st case that I bought that I could see reusing for a new build.
 

Shivansps

Diamond Member
Sep 11, 2013
3,918
1,570
136
Well, the reason is actually pretty obvious.

hmm no it isnt, if it was about Picasso and Matisse running out and still having a ton of unsold motherboards (that are still being produced btw) AMD wouldnt have blocked Cezanne from posting if the detected chipset is a A320. AMD pushed this update under those conditions for a reason and that reason seems to be Renoir. In fact OEMs are only officially supporting Renoir, they are leaving Veemer out of the official support list, even trought it works. Why Renoir seems important now i dont know.

But this discussion is for another thread, i was just pointing out what happened to the first generation AM4 boards, and it is something to keep in mind for AM5. I could also add that they only got VRM setups right with the 400 boards, specially the VRMs for IGP/SOC power.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,695
136
I could also add that they only got VRM setups right with the 400 boards, specially the VRMs for IGP/SOC power.

You can say that again. My Asrock AB350M-Pro4 has never been stable with any Ryzen-based APUs. Even lowly Athlons just crash randomly. So it can't be because of excessive power consumption.

IGP-less Ryzen CPUs work just fine.
 
  • Wow
Reactions: Ajay

Hans Gruber

Platinum Member
Dec 23, 2006
2,516
1,357
136
You can say that again. My Asrock AB350M-Pro4 has never been stable with any Ryzen-based APUs. Even lowly Athlons just crash randomly. So it can't be because of excessive power consumption.

IGP-less Ryzen CPUs work just fine.
Confirmed. I am playing around with my original Ryzen tester CPU R3 1200 (4-core) on my MSI B-350 Carbon Gaming Pro board. Known good to run any ram and worked well with my 3600. The 1200 is crashing with my Hynix 3600mhz kit. I guess it's only good to 3200mhz. I have 8GB of B-die 2 x 4GB sticks and now I know why it wouldn't go higher than 3400mhz with the R3 1200. It's a little POS CPU. Works great as long as you don't push the memory timings. My 3600 will find it's way back into this build once Zen 3D CPU's are released.

Credit to AMD. They came a very long way in a short period of time since the original Zen CPU's were released. The Rog Strix B450-F runs 3800mhz 24x7 without any stability issues. The B350 always would boot 3733-3800mhz but was like a fainting ghost. It seemed stable and the ghost in the machine would faint into a crash without BSOD.
 
  • Like
Reactions: Tlh97 and krawcmac

Topweasel

Diamond Member
Oct 19, 2000
5,437
1,659
136
Confirmed. I am playing around with my original Ryzen tester CPU R3 1200 (4-core) on my MSI B-350 Carbon Gaming Pro board. Known good to run any ram and worked well with my 3600. The 1200 is crashing with my Hynix 3600mhz kit. I guess it's only good to 3200mhz. I have 8GB of B-die 2 x 4GB sticks and now I know why it wouldn't go higher than 3400mhz with the R3 1200. It's a little POS CPU. Works great as long as you don't push the memory timings. My 3600 will find it's way back into this build once Zen 3D CPU's are released.

Credit to AMD. They came a very long way in a short period of time since the original Zen CPU's were released. The Rog Strix B450-F runs 3800mhz 24x7 without any stability issues. The B350 always would boot 3733-3800mhz but was like a fainting ghost. It seemed stable and the ghost in the machine would faint into a crash without BSOD.
While some of the Zeppelin issues with memory were on the licensed controllers they used for Zen and Zen+ and silicon lottery. Most of the issues with first gen Ryzen was down to all the mobo makers half butting their first wave or two of boards. I think I posted here that one even suggested it was on AMD because they launched close to an Intel product line launch.

The 400 line wasn't really needed for technology sake but so important just for AMD to disassociate their product stack from those 300 series boards.
 
  • Like
Reactions: Tlh97 and coercitiv

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,695
136
It seemed stable and the ghost in the machine would faint into a crash without BSOD.
The 400 line wasn't really needed for technology sake but so important just for AMD to disassociate their product stack from those 300 series boards.

I will say to AMDs credit that Ryzen has always been completely stable at stock speeds. It was only when overclocking memory 1st gen Ryzens had trouble with early AGESAs.

F.x: At launch my 1700 would not POST with anything higher then 2400MHz set on boot. It'd just restart and complain that overclocking failed. With the lastest ComboAM4 1.0.0.6 AGESA it chucks along happily with memory set at 2933MHz. Tried up to 3200MHz, but couldn't get it completely stable.

The APU stability issues on B350-based boards seems to be a different matter, since it's happening at completely stock settings with up-to-date AGESA. The very same Athlon 200GE runs more then fine in my A320M-K with 3533MHz memory. So it can't be the APU itself there is anything wrong with.
 

Topweasel

Diamond Member
Oct 19, 2000
5,437
1,659
136
I will say to AMDs credit that Ryzen has always been completely stable at stock speeds. It was only when overclocking memory 1st gen Ryzens had trouble with early AGESAs.

F.x: At launch my 1700 would not POST with anything higher then 2400MHz set on boot. It'd just restart and complain that overclocking failed. With the lastest ComboAM4 1.0.0.6 AGESA it chucks along happily with memory set at 2933MHz. Tried up to 3200MHz, but couldn't get it completely stable.

The APU stability issues on B350-based boards seems to be a different matter, since it's happening at completely stock settings with up-to-date AGESA. The very same Athlon 200GE runs more then fine in my A320M-K with 3533MHz memory. So it can't be the APU itself there is anything wrong with.
That's why I mentioned first and second wave 300 series. X370 and B350 launches were pretty early. The 320 and 300 releases being later. While many bad vrm and layering choices where made with those.

But that's my point. $300 boards with no overclocking. Bios updates that killed $500 boards. About half the memory issues where early board related. Then just about every sub $90 board had VRMs that struggled with 2700x's not to mention what a 3900x or 3950x would do to it.

That said the launch Agesa was terrible and it did take like 2 months for AMD to sort it out. But considering what the board makers did i am not sure it's then trying to engineer around their shortcomings.
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,695
136
Bios updates that killed $500 boards.

My Crosshair VI had a nasty little BIOS bug in the release version which would brick the board on update if you weren't careful. It was fixed, and the later boards shipped with a newer AGESA, but still. That first update was done with your heart in your throat, because it was readily replaceable due to a shortage of AM4 boards back then. They simply didn't produce enough to satisfy demand. I think AMD underestimated Ryzen immediate success rather badly.

But it has always been completely stable, if a very oddly behaving board. Runs a 3600 with 3200MHz memory now without issues.
 

Ajay

Lifer
Jan 8, 2001
16,094
8,114
136
My Crosshair VI had a nasty little BIOS bug in the release version which would brick the board on update if you weren't careful. It was fixed, and the later boards shipped with a newer AGESA, but still. That first update was done with your heart in your throat, because it was readily replaceable due to a shortage of AM4 boards back then. They simply didn't produce enough to satisfy demand. I think AMD underestimated Ryzen immediate success rather badly.

But it has always been completely stable, if a very oddly behaving board. Runs a 3600 with 3200MHz memory now without issues.
Thanks for reminding that I didn't buy a CH8 x570 board when they were reasonably priced :p
 

Hans Gruber

Platinum Member
Dec 23, 2006
2,516
1,357
136
I will say to AMDs credit that Ryzen has always been completely stable at stock speeds. It was only when overclocking memory 1st gen Ryzens had trouble with early AGESAs.

F.x: At launch my 1700 would not POST with anything higher then 2400MHz set on boot. It'd just restart and complain that overclocking failed. With the lastest ComboAM4 1.0.0.6 AGESA it chucks along happily with memory set at 2933MHz. Tried up to 3200MHz, but couldn't get it completely stable.

The APU stability issues on B350-based boards seems to be a different matter, since it's happening at completely stock settings with up-to-date AGESA. The very same Athlon 200GE runs more then fine in my A320M-K with 3533MHz memory. So it can't be the APU itself there is anything wrong with.
Here you go. 1st generation Ryzen R3 1200 CPU with technically not B-die but Samsung 4GB sticks 14-15-15-15 timings @ 3200mhz. I said enough with the Hynix 3600mhz kit and went to my Team Dark Pro sticks with the heaviest duty heatsinks I have ever seen on a memory stick. This Ryzen 1200 idles in the low 20's and maxed out @ 40C under a 240mm AIO. In hours long gaming it still will not touch 50C. One more thing, both the ram and the CPU have hard walls they hit with no exception. It's pointless trying to add more voltage or mhz. This chip will do 3850mhz and that is it. I settled in on 3800mhz.
MSI B350 GAMING PRO CARBON (MS-7B00) Performance Results - UserBenchmark
 

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,695
136
This chip will do 3850mhz and that is it. I settled in on 3800mhz.

The last 50MHz doesn't seem worth the bother. You have a decent chip there, a 700MHz OC on 1st gen Ryzen is nothing to scoff at. Original gen usually top out in that range (3800-3900MHz) anyway, apart from the rare gem that can do 4GHz.
 

tomatosummit

Member
Mar 21, 2019
184
177
116
I will say to AMDs credit that Ryzen has always been completely stable at stock speeds. It was only when overclocking memory 1st gen Ryzens had trouble with early AGESAs.
Overclocking and I found xmp itself to be the main incompatability.
Using a launch1700 with the above mentioned asrock b350pro4 and corsair veng 3200cl16;
Any xmp profiles just wouldn't boot like yours but dialing in similar settings manually I had 3200 memory on day one. Then agesa came out improving xmp combatability but breaking my 3200 settings. Now it won't go above 2933 albeit cl14, even with a new 3600cpu.
 

Ajay

Lifer
Jan 8, 2001
16,094
8,114
136
According to Igor's Lab, there is going to be a Threadripper without V-Cache after all.

Makes no sense to me to release it:
- this late
- at all
Chagall lives! AMD Ryzen Threadripper PRO 5995WX and its 4 brothers 5975WX, 5965WX, 5955WX and 5945WX with technical data | Leak | igor'sLAB (igorslab.de)
Makes total sense for an engineering/data science or CGI station. All WRX80 sockets now. That motherboard makes me salivate - wish I had extra $$s for DC projects.
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
16,798
7,249
136
According to Igor's Lab, there is going to be a Threadripper without V-Cache after all.

Makes no sense to me to release it:
- this late
- at all
Chagall lives! AMD Ryzen Threadripper PRO 5995WX and its 4 brothers 5975WX, 5965WX, 5955WX and 5945WX with technical data | Leak | igor'sLAB (igorslab.de)

You have to factor in where the Zen 3 chiplet dies go that aren't good enough for Milan. Demand for higher end Vermeer should dry up once Zen 3D is available.

And apparently they will now support dual socket.

I thought this was a typo but Dell does sell workstations with the 2P Xeon.
 

Joe NYC

Diamond Member
Jun 26, 2021
3,630
5,172
136
You have to factor in where the Zen 3 chiplet dies go that aren't good enough for Milan. Demand for higher end Vermeer should dry up once Zen 3D is available.

One of the articles mentioned that the new TR will be stepping B2, so only new dies, between Milan-X, Zen 3D and non-V-Cache Vermeer-X
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
WTF?? This makes no sense to me at all. Is someone punking us??

Sounds like one of those low core number with maximum cache and maximum speed SKUs for automated trading and special snowflake app environment where performance is critical and each of those cores has up to 96MB of L3 to themselves and shares that cache only with another core.

2P of 16C with 2C per CCD enabled with each CCD having 96MB of L3 => perfectly valid SKU for those traders.

EDIT: google by SKU returns price of $5700, so quite palatable for what this CPU gives to the right people.
 
Jul 27, 2020
27,944
19,099
146
AMD's strategy formulation teams are amazing. They are listening to customers and giving them exactly what they want. No wonder they are thriving and executing flawlessly. It's also possible they slaughtered a few virgins at some specific underground energy points around the world. I mean, there's just no stopping them!

They have a pact with the Devil or they have a perfect meritocracy in their offices/headquarters, unlike Intel, who let internal politics hasten their downfall.