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

Page 82 - 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

leoneazzurro

Golden Member
Jul 26, 2016
1,052
1,716
136
Pretty sure its the stacked 3d memory that cant run faster than 4.5ghz. Remember L3 run at same clockspeed (MP) as the cpu itself..
(3dstacked memory was originally developed for Milan X, which also happen to cap out at 4.5ghz max boosting)

Only solution for Vermeer X was to limit the clockspeed to 4.5ghz maximum and remove all overclocking options :disrelieved:
("temperature issue" is just PR BS)

See no other logical explanation for this :confused2:

I think that the temperature issue may not be a non-issue there.
Yes, cache may be the culprit but, AMD already realized on-die caches which can operate at frequencies >4,5GHz, even in mobile environment (Rembrandt reaches 4,9GHz turbo, Vermeer has been seen boosting to 5+GHz single core turbo without OC). With Vermeer-X we have stacked dies and that means that indeed the connection by TSV while being close to an on-die interface it will have for sure some limitation, but we have also a thermal interface between the two stacked dies. Which is a not trivial issue as AMD is adding the cache die thermal resistance, plus the resistance of said interface to the mix. So this too is contributing to the decision to limit frequency.
 

DrMrLordX

Lifer
Apr 27, 2000
22,065
11,693
136
Please excuse my ignorance but dumb question: who eats the loss in the price drop? The retailer or AMD?

I'm gonna side with Mark on this one, though personally I don't exactly know. I do know that yearly/seasonal product (like automobiles) goes through a pricing cycle where it can sell for MSRP for awhile, before it goes into discount status, after which point (if still unsold) it goes to liquidation at much lower prices for everyone involved.

If you've watched AMD's product launches over the years, you can usually tell when they're trying to get stuff moved off shelves and out of supply channels, but it's hard to know exactly how AMD applies leverage to get everyone on board with price cuts.

See no other logical explanation for this :confused2:

Didn't someone already demo a Milan-X @ 4.8 GHz max boost by setting unlimited EDC + TDC?
 

Det0x

Golden Member
Sep 11, 2014
1,299
4,234
136
Didn't someone already demo a Milan-X @ 4.8 GHz max boost by setting unlimited EDC + TDC?
Haven't seen or heard anything about this, and i try to follow the Vermeer X news/rumors pretty tight as i'm playing with the idea about buying one just for fun. (already have a very good binned 5950x)

Could you please share link if you know where you saw it ?

*edit*
One quick google search later: AMD's EPYC 'Milan-X' overclocked to 4.8 GHz - TechnoSports
He also attempted to overclock the CPUs. Because server platforms aren’t built for overclocking, the specialist had to use a specialised motherboard designed for hyperscale data centres and containing an upgraded voltage control module to push the frequency of AMD’s EPYC 7773X processors to 4.80 GHz (boost).

They upped power limitations to 1500W (up from 280W by default) and increased voltage to 1.55V using the AMD EPYC Overclocking software. However, the CPUs were cooled using air coolers, which were running at high speeds and so making a lot of noise.
But not other proof then purely text/rumors as far as i can see ? Whole story sounds a little bit fishy..

*edit2*
Found the answer: https://videocardz.com/newz/amd-milan-x-7773x-epyc-cpus-with-3d-v-cache-get-overclocked-to-4-8-ghz

Overclocked to 4.8 GHz… but not really
The creator is using a special EPYC Milan/Rome ES/QS overclocking tool by Uat4 and ExecutableFix. This tool unlocks power limits and increases maximum all-core frequencies and voltages much higher than AMD intended, but only for engineering samples. However, this is not true overclocking. ExecutableFix explained to us that the frequency reported by CPU-Z software is not a real frequency, but rather a target. This is probably why Kenaide did not provide any benchmark results using a 4.8 GHz OC session, it would have simply shown smaller than expected performance uplift.
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
22,065
11,693
136
But not other proof then purely text/rumors as far as i can see ? Whole story sounds a little bit fishy..

Point is, it set that target, and it very well could have reached it in ST workloads. Unless AMD does something to limit the AGESA releases for AM4, it's probable that correctly-tuned PBO could set higher targets for Vermeer-X as well.
 

deasd

Senior member
Dec 31, 2013
567
922
136
As I mentioned 5800X3D which is a consumer product is more of a experimental product rather than grabbing share from competitor.
But as for the OC features removing... I thought it's just being LOCKED multiplier...? Would AMD also going to lock everything related to Overclock, that said not only multiplier but also RAM, Bus speed? That sounds infeasible...
 

LightningZ71

Golden Member
Mar 10, 2017
1,798
2,156
136
For those pointing out that AMD has L3 caches that currently exceed 4.5Ghz in volume production, you need to realize that the process used for the stacked L3 cache is not exactly the same as the ones used for the CCDs. AMD stated publicly that they use a cache density optimized version of the N7 process to achieve greater cache density on the stacked L3 die, fitting 64MB in a space slightly larger than the 32MB on the CCD. There was likely a trade-off made with respect to maximum reliably achievable frequency of the L3 cache in making that decision. It was a wise one when you consider that the target market for the project was Milan-X and that the 5800X3d is a prestige product. I think that, with a solid cooling setup, a well tuned 5900x will likely prove to be almost even with respect to gaming performance while being significantly better in highly multi-threaded tasks. I'm still waiting for benchmarks, but, I'm leaning towards the 5900x being a better value for the long term if you already have a good X570 board. IF you don't, I think that the Alder Lake platform is marginally superior AT THE MOMENT.
 

Hitman928

Diamond Member
Apr 15, 2012
6,186
10,693
136
I think that the temperature issue may not be a non-issue there.
Yes, cache may be the culprit but, AMD already realized on-die caches which can operate at frequencies >4,5GHz, even in mobile environment (Rembrandt reaches 4,9GHz turbo, Vermeer has been seen boosting to 5+GHz single core turbo without OC). With Vermeer-X we have stacked dies and that means that indeed the connection by TSV while being close to an on-die interface it will have for sure some limitation, but we have also a thermal interface between the two stacked dies. Which is a not trivial issue as AMD is adding the cache die thermal resistance, plus the resistance of said interface to the mix. So this too is contributing to the decision to limit frequency.

The stacked V-cache is using a different library and maybe even different process tweaks that allows it to have roughly 2x the density of the on chip cache. These kind of changes could easily be limiting the max frequency of the stacked V-cache. Hopefully we get a deep dive from a reviewer or more info from AMD themselves at some point.

Edit: @LightningZ71 just beat me to it.
 

leoneazzurro

Golden Member
Jul 26, 2016
1,052
1,716
136
Interesting, do you have any data about the process used? Also, I did not want to entirely dismiss the idea that cache is frequency limited, instead I wanted to say that thermals are having a big influence, too.
 

biostud

Lifer
Feb 27, 2003
18,700
5,431
136
Gimmick or no gimmick. If you don't like the product, don't buy it. If you do, buy it. I can't really see any downside to giving consumers an extra choice for the AM4 platform. For me it is en enthusiast product like the intel KS line except this processor gets extra performance from huge amount of cache while Intel push frequencies. Same for intel, if you want top of the pop, buy the KS line, if not get one of the other processors.
 

JoeRambo

Golden Member
Jun 13, 2013
1,814
2,105
136
There is one thing in CPU having no overclockability at all. Another is having a hard wall in scaling. Luck of draw can move that wall by a hundred or so mhz. Removing OC option on what should be ultimate enthusiast CPU on AM4 is disgraceful and disgusting action by AMD.
 
  • Love
Reactions: igor_kavinski

Makaveli

Diamond Member
Feb 8, 2002
4,810
1,277
136
There is one thing in CPU having no overclockability at all. Another is having a hard wall in scaling. Luck of draw can move that wall by a hundred or so mhz. Removing OC option on what should be ultimate enthusiast CPU on AM4 is disgraceful and disgusting action by AMD.

lol disagraceful and disgusting such colorful wording. Why does this have you up in your feelings so much?
 
  • Like
Reactions: Tlh97 and Elfear

nicalandia

Diamond Member
Jan 10, 2019
3,331
5,282
136
There is one thing in CPU having no overclockability at all. Another is having a hard wall in scaling. Luck of draw can move that wall by a hundred or so mhz. Removing OC option on what should be ultimate enthusiast CPU on AM4 is disgraceful and disgusting action by AMD.
They are about to release the Gaming King CPU and you got your feathers ruffled? OC is Dead. It's been dead for quite a while now. Intel and AMD have become expert at binning their own CPU.


 

nicalandia

Diamond Member
Jan 10, 2019
3,331
5,282
136
ADF warriors move forward! Defend the undefendable. Sure, locking a single CPU is a 3D chess move by AMD!
Chill Out Dude.

It's a Niche product to regain the "Gaming" Crown(for reviewers at 1080P using a 3090 lol, because at 1440 and Up there is no difference between current Zen3 and Alder Lake), that will give people an upgrade path if they so desire on their AM4 Platform.

This Chip will become an Afterthought when AM5 Zen 4 land and star laying waste..

 
Last edited:
  • Like
Reactions: Tlh97 and Makaveli

Makaveli

Diamond Member
Feb 8, 2002
4,810
1,277
136
Chill Out Dude.

It's a Niche product to regain the "Gaming" Crown(for reviewers at 1080P using a 3090 lol, because at 1440 and Up there is no difference between current Zen3 and Alder Lake), that will give people an upgrade path if they so desire on their AM4 Platform.

This Chip will become an Afterthought when AM5 Zen 4 land and star laying waste..

i'm at 3440x1440p Ultrawide so I need to see final reviews cause I'm not sure the gains would be there in my setup going to this chip. Will find out soon enough.
 

Hitman928

Diamond Member
Apr 15, 2012
6,186
10,693
136
i'm at 3440x1440p Ultrawide so I need to see final reviews cause I'm not sure the gains would be there in my setup going to this chip. Will find out soon enough.

Even at 1440p with quality settings, the performance of just about all of the modern enthusiast CPUs are so close together that it doesn’t really matter which one you have.
 

Makaveli

Diamond Member
Feb 8, 2002
4,810
1,277
136
PBO on my Zen2 and MSI motherboard sucks. Too much added power for a small bump in performance. Seems like PBO2 on 5000 series Ryzen is actually worth while.
i've used both on my board having a 3800X previous to my 5800X and yes PBO2 is much better on Zen 3.
 

nicalandia

Diamond Member
Jan 10, 2019
3,331
5,282
136
i'm at 3440x1440p Ultrawide so I need to see final reviews cause I'm not sure the gains would be there in my setup going to this chip. Will find out soon enough.
There will be None... Come On...! You could use a 9900K with a 3080/3090 and would not notice any difference from 9900K and the 5800X3D at those settings(3440 x 1440)
 
Last edited:
  • Like
Reactions: Tlh97 and Makaveli

Insert_Nickname

Diamond Member
May 6, 2012
4,971
1,693
136
This Chip will become an Afterthought when AM5 Zen 4 land and star laying waste..

I'm getting distinkt K6-3 vibes from the 5800X3D. The original Athlon had the performance crown, but the K6-3 was an excellent upgrade for Socket7 boards. If you could find one that is.

The K6-3 was also a high-cache design for it's time. It had 256KB of on-die cache, something the SlotA Athlons didn't even have at the time. Further it could use the on-board L2 as L3 cache. Which was highly unusual to have at the time.
 

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136
Don't care about OC (as in going above spec, as little that means nowadays) per se. But I want to be able to undervolt and optimize a chip that way. Hope that ability is not permanently included as collateral damage.
 
  • Like
Reactions: Tlh97 and Elfear