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

[Kedas]

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"

 

[MadRat]

All sorts of websites (mostly tech. some health. some youtube tabs) on my Thinkpad. Should really look into Opera. At work, google searches lead to opening up tons of tabs and Edge reaches more than 4GB RAM. Even though that system has 16GB RAM and an 860 EVO SSD, sometimes it feels it is struggling under the load of those tabs. But it is somewhat infrequent. Haven't had that much thrashing on it recently. Could be they have improved Edge in the last update?

Not judging but pr0n and GLP bring every system to a crawl. Its all the drive-by ghost miners loading in the background tabs.

 

[igor_kavinski]

Not judging but pr0n and GLP bring every system to a crawl. Its all the drive-by ghost miners loading in the background tabs.

Nah.

CPU utilization doesn't go more than 20-30% at worst.
When the system starts thrashing, disk usage in Task Manager goes to 100%.
pr0n sites are the best way to get all sorts of hard to get rid of malware (kinda like STDs) so nope.
Too old for them, anyway.

What is GLP?

 

[eek2121]

It's actually not a bad time to buy 64GBs worth of DDR4 right now. Price is actually below what I paid for 32GB a couple of years ago. 64GB DIMMs are probably going to be expensive for a while yet of course.

I was tempted to pick up a matching 64GB set for one of my 32GB kits. For 96GB total. Not that I need it, but 22 years ago my PC had 96MB RAM. Seems fitting.

It is a bad idea if you are upgrading your system soon and don’t have use for the old one. DDR5 is the future.

What if we go with all-core clock instead? Current all core workload clocks for AMD CPUs are anemic, maybe 3.9Ghz or so for 5950x. Does not take anything exotic to take that to 20% higher clock of 4.8Ghz? It's not something today's 5950x can't do with tuning and cooling for 300W for 16C, but it takes new process node and new arch to cut power usage to more palatable region.

I think the 3.9 GHz all core speed for the 5950X is due to the TDP of the chip more than anything else. The 5900X has a higher all-core turbo and both of the single chiplet CPUs can hit around 4.5 GHz all core. There's some data from an AT article that shows the 5800X, 5900X, and 5950X all hitting the same power levels when boosting, with the only difference being the clock speed that all of the cores hit.

The details we have on TSMC 5nm don't suggest that provides either enough of a frequency or a power uplift for a Zen 3 shrink to hit something like a 5 GHz all core turbo within the same 125 W TDP. The transistor density improves a lot, but it's a 15% frequency boost or a 20-30% power improvement. I'm not sure where the person who originally made the post got 20% - 40% improvement from, but it doesn't line up with other published figures. Some articles do talk about other possibilities for additional performance gain, but those parts are light on details or specifically indicate that they come at the expense of density.

If we use the 5N shrink for a 30% power reduction and the figures from the AT article, we'd only expect an all-core turbo of around 4.5 GHz for a 16-core chip. The 8-core chip could be pushed so that the all-core turbo is above 5 GHz, but I question the usefulness of that since games largely still rely on having a single fast thread as opposed to being able to boost all threads to some amount and anything that's massively parallel like Cinebench is going to do far better with twice the cores even if they're not as fast.

Even if we could get a magic 40% frequency boost, at that point the chip is probably outrunning the memory system and won't realize a 40% performance gain from those clocks. We've already seen how much Alder Lake benefits from DDR5 memory in some benchmarks, so it's likely that AMD is going to see a lot of the performance uplift from Zen 4 come from faster memory as well.

This.

Also, I could definitely see a 5.2-5.3 ghz single core clock. If not, power consumption will simply be lower.

Zen 3 already hits ~5.1 ghz on single core workloads (especially if curve optimizer is involved)

 

[gdansk]

I know AMD keeps finding ways of pushing clocks higher but at some point (Zen 5?) don't we expect clocks to regress again?

I guess even Vermeer X has a slight regression

 

[Insert_Nickname]

It is a bad idea if you are upgrading your system soon and don’t have use for the old one. DDR5 is the future.

Sure, but after being an AM4 early adopter (literally launch day), I'm avoiding first-gen AM5. Maybe I'll upgrade once it hits 2nd or 3rd gen. DDR5 should also be a lot cheaper by then.

With my new 5700X I'll be set up for years. On a mature platform, where all the bugs, quirks and kinks have been worked out.

 

[Mopetar]

On the ~40% performance improvements for Zen 4... AMD is not using straight N5. They are using a custom variant of N5p which affords a ~20% frequency clock speed improvement at the same iso power (over N7). And then there is an expectation of ~20% IPC improvement with all the additional transistors that the brainy AMD engineers might find a use for.

But yeah just guesses in the end.

The post I was responding to originally wasn't talking about IPC increases, but merely about a hypothetical Zen 3 CPU on TSMC 5N getting 20-40% frequency boost.

I think Zen 4 will see more performance uplift come from IPC gains than straight frequency gains.

 

[igor_kavinski]

Lisa Su showed off Zen 4 doing 5 GHz on all cores. Seems very likely it can hit 5.5 GHz boost on a single core. That and IPC gains should easily allow it to put Raptor Lake to shame. And if the top 170W SKU is paired with V-cache, Oh Dear God.

 

[DisEnchantment]

Also, I could definitely see a 5.2-5.3 ghz single core clock. If not, power consumption will simply be lower.

The more interesting thing about N5 is not the frequency tbh. My 5950X can regularly do 5.1 GHz stock with a 280mm AIO.

In my mind what is more interesting would be how much tradeoffs AMD would have to make this time, or rather what things they need not have to sacrifice this time.
Even if AMD did not go with a 3 fin design with N7, they made a lot of the cells taller and increased the number of tracks in order to get the needed drive performance for ~5GHz operation. (I think it is 2x poly)
This degraded density, and the taller cells also degrade RC characteristics leading to lower energy efficiency.
Then there are those additional metal layers needed to deliver the signals and power which is necessitated by the high frequency design which also introduced additional RC losses. If I remember well, all the way to 19 metal layers compared to 13-14 for mobile SoCs.

If on their custom N5 they can go with standard 2 fin layout with compact/shorter cells and keeping more or less similar frequency it would already be a big leap.
Far higher efficiency than what TSMC is advertising for N7--> N5 transition because basically the N7 customizations AMD made nerfed efficiency in no insignificant amount.
The other double whammy is that the taller cells cut density a lot, so if they can stick to a compact 2 fin cell layout on N5 would really boost Xtor density allowing a greater density gain than TSMC's advertised gains for N7-->N5 transition. (In performance optimizations the cells can go almost 30% larger than usual)

This was the reason I found that Zen 4 process reveal from last year's accelerated DC event to be interesting, it indicated that they indeed did not have to sacrifice density and efficiency much and because of reasons above got better gains vs N7 than TSMC's N7-->N5 transition (using mobile as standard)
They will gain base frequency and all core frequency simply due to being more efficient. And the density gain is a wild card, it is very plausible that they literally meant it when they said 2x density gain. which could be up to 60% MTr gain for a Zen 4 CCD with the rumored die sizes. What are they doing with so much more MTr on a arch refresh?

But the rumored 170W SKUs are still baffling to me though, can't explain what is happening there. Could be just over provisioning or some special SKU.

Anyway we will know soon enough.

And bonus patent applications for practically doubling SRAM density with creative cell layouts (could have been used on 3D V Cache I think, but would be interesting if they could use this on the Zen4 L3)

20220093504 - FOLDED CELL LAYOUT FOR 6T SRAM CELL

https://www.freepatentsonline.com/y2022/0093504.html

20220093503 - DUAL-TRACK BITLINE SCHEME FOR 6T SRAM CELLS

https://www.freepatentsonline.com/y2022/0093503.html

 

[Ranulf]

Sure, but after being an AM4 early adopter (literally launch day), I'm avoiding first-gen AM5. Maybe I'll upgrade once it hits 2nd or 3rd gen. DDR5 should also be a lot cheaper by then.

With my new 5700X I'll be set up for years. On a mature platform, where all the bugs, quirks and kinks have been worked out.

Ditto on the early adopter thing and all the other issues that AMD has caused with their game playing on cpu support and delayed releases. Nevermind all the "ram standard is the future" nonsense that happens every 5 or so years when a new standard comes out. Not worth the price nor the early teething issues. DDR5 isn't the future for 1-3 years at this point.

 

[MadRat]

What is GLP?

A conspiracy and urban legend (ufo, bigfoot, etc) site.

 

[Mopetar]

But the rumored 170W SKUs are still baffling to me though, can't explain what is happening there. Could be just over provisioning or some special SKU.

I suspect part of it will be used by the new parts out of the gate. Intel pushing their chips beyond 200W let them benchmark a bit better. AMD doesn't need to go nearly that far, but an extra 50W will let them stretch the bars a bit more.

It also gives some extra room for all core boost on the two chiplet parts. Those were constrained a fair bit because when 16 cores all want to use that TDP and there's an IO die to consider as well, 125W doesn't go quite as far as one would think.

There's also the rumors that all of the Zen 4 CPUs will have a small amount of integrated graphics on the IO die this time around. That will add to the power draw if it's being used and they probably want extra room for it.

Eventually they will release Zen 4D parts and the v-cache is going to need to be powered. With the 5800X3D they lowered the boost clocks in part because the TDP needed to stay the same.

So I don't think there's just one reason for the 170W TDP, but a lot of little reasons. We may not even see the initial Zen 4 CPUs use all of that 170W either, but it will be there in case they need it later and it will ensure that boards will support future parts that actually will draw that full amount of power.

 

[gdansk]

I find 170W regrettable but if your competitor is doing it you have to play that game. Really, AMD shouldn't follow but they will because halo parts work with irrational consumer behavior.

But it'll be really neat to see if the 5800X3D is as competitive as AMD claims while not using more power. On a 4 year old process. On an old platform. With old memory.

 

[Mopetar]

I find 170W regretable but if your competitor is doing it you have to play that game. Really, AMD shouldn't follow but they will because halo parts work with irrational consumer behavior.

The funny part would be if it makes Intel go to 300W. Probably won't happen, but it's kind of funny to imagine.

But it'll be really neat to see if the 5800X3D is as competitive as AMD claims while not using more power. On a 4 year old process. On an old platform. With old memory.

The early results from the site in Peru that got an early CPU suggest it will be, at least in games and a few one off other applications.

 

[Markfw]

The funny part would be if it makes Intel go to 300W. Probably won't happen, but it's kind of funny to imagine.



The early results from the site in Peru that got an early CPU suggest it will be, at least in games and a few one off other applications.

Intel will do whatever is required to get to number one DIY/OEM, 400W ? 500W ? 12900ks already uses 500w+.

 

[jamescox]

I'm not alone. Browsers should logically put the inactive tabs to sleep and save their data to disk. But instead, they keep leaking memory. Just last night, I "end tasked" Firefox and my RAM went from 91% to 26%. Restored the tabs on relaunch. It's now at 43%. I didn't close any tabs. Actually opened quite a few more.

I think the people who have this "100+ tab" syndrome have a problem that can be solved with a to-do list in browsers. There are definitely tab management add-ons out there but too lazy and too busy to explore them. Plus, I don't want to depend on an add-on that might get abandoned by its developer at some point in the future.

I have never noticed Firefox leaking memory, but I run on Linux with Firefox plus noscript. This takes very little memory. I probably have hundreds of tabs open, and memory usage is in the hundreds of MB. I don’t think it loads them automatically on restart unless you click on them, so most of the tabs are effectively bookmarks. I generally enable scripts for domains as temp trusted so they don’t load on a restart. I would recommend this for anyone using an older system since it makes it very lightweight. It is extra work to enable sufficient scripts to get websites working though, so it isn’t for everyone. You only have to do it once though, unless you use the temp trusted setting; good for unknown websites. It probably also blocks almost all add networks unless you explicitly enable them.

 

[jamescox]

Lisa Su showed off Zen 4 doing 5 GHz on all cores. Seems very likely it can hit 5.5 GHz boost on a single core. That and IPC gains should easily allow it to put Raptor Lake to shame. And if the top 170W SKU is paired with V-cache, Oh Dear God.

It doesn’t seem like they have really talked about clock speeds much at all with pst releases. Since they did mention clock speed in this case, I am wondering if they are achieving very good clocks. Perhaps they expect to have higher clocks than intel can achieve so they don’t mind talking about them.

 

[igor_kavinski]

But it'll be really neat to see if the 5800X3D is as competitive as AMD claims while not using more power. On a 4 year old process. On an old platform. With old memory.

When you put it like that, it sounds really impressive to pull that off.

 

[biostud]

Ditto on the early adopter thing and all the other issues that AMD has caused with their game playing on cpu support and delayed releases. Nevermind all the "ram standard is the future" nonsense that happens every 5 or so years when a new standard comes out. Not worth the price nor the early teething issues. DDR5 isn't the future for 1-3 years at this point.

When zen4 launches DDR5 has been out for a year, and while I'm not going to buy it on day one, but first when next-generation video cards are out, I really don't believe the platform to be a bug ridden menace. Sure if you prefer to buy two years old technology, that is fine, personally I prefer PCIe5, DDR5 and zen4

 

[Saylick]

Alright, dig in!

AMD Ryzen 7 5800X3D Review – The last gaming gift for AM4

AMD's last gaming gift for AM4 platform. Ryzen 7 5800X3D review. Benchmarks and work in progress of the new AMD's gaming focused product.

xanxogaming.com

 

[Shamrock]

Looks like I'm buying a 5800x3d

 

[lobz]

would have been nice of them to have some nice DDR5 just to show the 12900K in its best light too, but you can't really argue about the cost efficiency - oh, how the turntables (sic!) - of the 5800X3D here

 

[Timorous]

Me 2, will go well with my 3600 C16 B-Die.

Tying or beating the 12900KF with the same ram is a good result. Doubt the games that are tied will improve much with DDR 5 and the ones where AMD wins you are paying a lot of money in ram to match that performance.

Can't see 12700K + Mobo + fast DDR5 being cheaper than 5800X3D + Mobo + 3600-4000 capable DDR4. For gamers who don't need the productivity benefits of ADL it looks like a good option and for existing AM4 owners it is even better.

 

[Gideon]

Me 2, will go well with my 3600 C16 B-Die.

Tying or beating the 12900KF with the same ram is a good result. Doubt the games that are tied will improve much with DDR 5 and the ones where AMD wins you are paying a lot of money in ram to match that performance.

Can't see 12700K + Mobo + fast DDR5 being cheaper than 5800X3D + Mobo + 3600-4000 capable DDR4. For gamers who don't need the productivity benefits of ADL it looks like a good option and for existing AM4 owners it is even better.

It will be an excellent upgrade to pretty much any AM4 owner, but a gigantic one to those still using Zen or Zen+. Particularily if rumors about B350 BIOS are true.

As for RAM it will probably show less gain from ultra-fast memory than other CPUs (and also less degradation from slower ram), as there will be other bottlenecks that will present themselve, but we'll have to see benchmarks for that.

 

[Timorous]

It will be an excellent upgrade to pretty much any AM4 owner, but a gigantic one to those still using Zen or Zen+. Particularily if rumors about B350 BIOS are true.

As for RAM it will probably show less gain from ultra-fast memory than other CPUs (and also less degradation from slower ram), as there will be other bottlenecks that will present themselve, but we'll have to see benchmarks for that.

MSI have announced that 300 series boards will have support by end of April.

Not sure if I should buy immediately at launch or wait until BIOS support is added. Thinking of buying launch day in case it is very popular and supply is limited. If that does not happen I can always return unopened and buy for cheaper when the BIOS is released.

It will be a very nice upgrade from my 2200G though.

 

[epsilon84]

Me 2, will go well with my 3600 C16 B-Die.

Tying or beating the 12900KF with the same ram is a good result. Doubt the games that are tied will improve much with DDR 5 and the ones where AMD wins you are paying a lot of money in ram to match that performance.

Can't see 12700K + Mobo + fast DDR5 being cheaper than 5800X3D + Mobo + 3600-4000 capable DDR4. For gamers who don't need the productivity benefits of ADL it looks like a good option and for existing AM4 owners it is even better.

So, in theory, the games where the 5800X3D has the big wins would also be games where ADL would make significant gains with DDR5? That makes sense to me. I don't think whatever price differences between the 2 combos you mentioned would be a huge factor for anyone looking at these for the absolute best in gaming, since they'll probably be rocking a 3090 or 3090 Ti as well, so a $50 price difference is literally loose change at the bleeding edge of PC gaming systems which costs thousands of dollars.

Based on this review of the 12900KS showing gains up to 20% with DDR5, I'll wait for comparisons with DDR5 based ADL systems before drawing my conclusions, though it seems with DDR4 a 5800X3D at worst ties with a 12900K and beats it handily in a few games which is a great result, though in line with my expectations all along.