Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[scineram]

I think it's coz Zen 4 chips boost quite aggressively, always banging their heads against the thermal limit.

What is that snappy thing? First time I have heard of that.

 

[biostud]

This really must be only your third anniversary (congrats! ) Because how it works is, there's our money and her money.

Besides cost of living, she uses most of her money on stock investments, so it'll work out all right.

Do you think she will get jealous of your PC because of how much you spend time with it, instead of spending with your wife? Or do you crawl out of bed in the dead of night when she's out cold?

I spend most of my time here on AT, and very little time with the computer

 

[Markfw]

You know there are some other things that benefit greatly when you have a large L3. Look at Milan-X. Does anybody have examples of those ? Looking to use my 7950X3D on some. I know some primegrid really rock with a large L3 , but looking at other apps.

 

[Joe NYC]

Kinda makes you wonder why AMD isn't testing with those titles.

The AMD marketing for gaming is contributing a negative value.

If everyone was fired, things would actually improve.

 

[Joe NYC]

That would be red meat for the haters. They'd go HAM screeching about cherry picking. However, in a 50+ game suite, having a subset devoted to cache friendly titles makes sense.

Agreed. Some of the CPU intensive simulations would be great addition to the mix.

Whenever I see a "Gaming review" and everything is about FPS, I just reach for the bucket under my desk, just in case the review makes me throw up.

 

[biostud]

Agreed. Some of the CPU intensive simulations would be great addition to the mix.

Whenever I see a "Gaming review" and everything is about FPS, I just reach for the bucket under my desk, just in case the review makes me throw up.

View attachment 77145

So how is Stellaris running on your 5930k?

 

[Joe NYC]

So how is Stellaris running on your 5930k?

I came across these test results while looking for test results for another simulation made by the same company Paradox - Europa Universalis- with a particular mod.

I have not played Stellaris for years, so I don't remember.

BTW, the EU with M+T mod is nor really enjoyable on the 5930k, due to slowness, barely on 5800x.

I am looking forward to a substantial increase in performance on Zen 4 V-Cache, similar to Stellaris (from V-Cache).

 

[igor_kavinski]

What is that snappy thing? First time I have heard of that.

I got my first dose of snappiness with i5-12400 with DDR4-3200. It seemed like it chewed through common tasks like a piranha. Everything felt really quick, especially the snappiness of the UI. With my 12700K and DDR5-4600, I can see that almost every app in Start menu opens like it was already preloaded in RAM. Even Librecalc that usually takes a few seconds at startup, the logo showed up, progress bar zipped from start to finish in the blink of an eye and it was ready to be used. It's not just coz of NVMe coz I was using that with my i7-5775C before.

 

[Mopetar]

Yes, I already said to do it for big test suites. Let's not move the goalpost by injecting small sample sizes as a scenario. Let's keep debating my contention. In which case, I could not disagree more about the tarnishing remark. That ain't even close to how I see it.

Almost every game is niche. MSFS has a very hardcore user base. So do racing sims like AC. Why test Shadow of the Tomb Raider, Metro Exodus, or a bunch of the other titles still included? They are not played by many people anymore. Yet many test them. The answer is because they reveal different aspects of how different game engines perform with different CPU architectures, and their platforms. That should include big cache friendly titles. And not just a single example.

If we start eliminating outliers where do we stop? A few titles heavily favor AMD 3D right now. Does that mean we can get rid of the same number of games that heavily favor Intel's platform and CPUs? It's the slippery slope fallacy. Either you include all of the types of games that can reveal different aspects of the CPU and platform performance, or it's a bad look. Furthermore, If we are going to throw out all the games where the 3D has the biggest lead, then again, do the same for Intel. Hell while we are at it, throw out the same number of big leads for Zen 4 too. Even the AM4 3D looks a lot better if we do that.

For myself, there is no logical reason to exclude a decent sample size of games that reveal the benefits of dat big cache. It's a new CPU feature that should be fully examined. Not put through a test suite that fails to show its full potential. Imagine if RTX was treated that way. No one tested Cyberpunk'd or Metro or a couple of other games where RTX makes its best showings. That would look super sus, as if downplaying its dominance when properly utilized. Yet that seems to be what is happening with the 3D in test suites. All of its biggest wins are absent from the charts?

I'm don't really disagree with anything you say, but if a site has been using a consistent test suite for some time, modifying it in any way is sure to shake up the results in one way or another.

If you're going out of your way to pick titles you know will favor some product, it's not really a good way to go about benchmarking as it's only working towards a desired conclusion.

If you wanted to give a good review, by all means find some games where a particular product does well. Even if they are niche, some people will still care about those results and the review is valuable. Of course the reverse is also true and a reviewer ought to show the cases where a product performs poorly.

If you do this over a wide enough set of results, the average performance won't move much, and if you try to add about an equal number of good and bad games, it probably doesn't move at all.

But even averages themselves can be misleading. Imagine a scenario where v-cache is all or nothing and we have 3 games that get a 33.3% boost and 7 games that get a 0% boost. The average is 10% even though no individual title sees that performance difference.

Any serious gamer would have to be living under a rock not to realize there are some titles where the v-cache is a 30+ percent performance uplift. Or they're disingenuous dolts determined to dismiss disagreeable data.

If Zen 3D didn't do well in some particular titles I don't think anyone would be clamoring to have them included in the list of titles benchmarked by any reviewers. Only including them after the fact is by definition moving the goal posts. There's presumably dozens or hundreds of other titles that could also be added, but won't because they're uninteresting.

If the big reviewers don't cover niche titles where Zen 4D performs well, expect smaller reviewers to highlight those titles, even if they're only doing it for the traffic. I wouldn't want to see sites changing their suite to benefit AMD any more than I'd want to see them change it to the detriment of AMD. As always, no one reviewers suite will be perfect for any number of reasons, but we're not restricted to any one reviewers results either.

If someone already has a test suite they've been using for the past year or so, I don't think they should modify it one way or another to make a particular product look good or bad. Look how much flak Geekbench has been getting for changing how MT performance is measured. It's still possible to mention games that aren't part of the usual suite, but I wouldn't add a title just because it's discovered to be a performance outlier after the fact.

 

[H T C]

If we start eliminating outliers where do we stop? A few titles heavily favor AMD 3D right now. Does that mean we can get rid of the same number of games that heavily favor Intel's platform and CPUs? It's the slippery slope fallacy. Either you include all of the types of games that can reveal different aspects of the CPU and platform performance, or it's a bad look.

In reviews with a huge number of titles, remove the 2 best / worst, and reduce that to 1 if the number of titles is "on the smaller side".

This should be announced up front in the testing methodology, before the 1st benchmarks are even shown, and it doesn't matter if the best / worst results are "by a few FPS" or "by "200+ FPS".

While this may not remove all of the outliers or even remove titles that ARE NOT outliers, this should be done in ALL reviews, for the sake of consistency: it's when reviewers do it in review A but not in review B that it brings "mistrust".

My 2 cents.

 

[DrMrLordX]

That would be red meat for the haters. They'd go HAM screeching about cherry picking. However, in a 50+ game suite, having a subset devoted to cache friendly titles makes sense.

Well yes, but as you said, they can include it in their media material along with all the other titles they already use. Not that AMD is currently using 50+ games to showcase their hardware.

Whenever I see a "Gaming review" and everything is about FPS, I just reach for the bucket under my desk, just in case the review makes me throw up.

Keep in mind that such performance metrics are a bit on the niche side. You probably have more people playing Fortnight on PC than every turn-based strategy game released in the same timeframe. Yes some people are going to want to know what produces the fastest turns in Europa Universalis or Stellaris, but not very many.

 

[stuff_me_good]

Agreed. Some of the CPU intensive simulations would be great addition to the mix.

Whenever I see a "Gaming review" and everything is about FPS, I just reach for the bucket under my desk, just in case the review makes me throw up.

View attachment 77145

Would be nice to see real benchmark made with supreme commander forged alliance with FAF client. It is still popular RTS game with active development. You could always download some replay with heavy CPU usage and see how fast the replay play through with the time slider at the fastest setting. 5800x3D apparently speeds replay considerably over vanilla CPU's even though the clocks are much lower and the lobby CPU test isn't that impressive.

 

[eek2121]

It will be interesting to see non-gaming benchmarks.


Also, I hope some reviewers also test both chips at the 120W TDP.

 

[Kaluan]

It will be interesting to see non-gaming benchmarks.


Also, I hope some reviewers also test both chips at the 120W TDP.

Worthwhile outlets certainly will. I feel like good cooling will play an even bigger role with the X3D parts, in part because of the new 89C soft limit.

Here's 120/162W (stock) R23 result:
37.973 points


But we're gonna need more than CB and GB lol

 

[DAPUNISHER]

Well yes, but as you said, they can include it in their media material along with all the other titles they already use. Not that AMD is currently using 50+ games to showcase their hardware.



Keep in mind that such performance metrics are a bit on the niche side. You probably have more people playing Fortnight on PC than every turn-based strategy game released in the same timeframe. Yes some people are going to want to know what produces the fastest turns in Europa Universalis or Stellaris, but not very many.

There we go with the niche claims again. How niche, and what does that even mean? A game not showing up on the Steam top played games doesn't mean there isn't a big number of players worldwide.

For example: WoW still has something like 8.5 million monthly subscribers. It is difficult to assess games available through game pass too. MS Flight Sim had an estimated 10 million players across all platforms in December. That's why it should be in every major review. Yet it isn't. Dispensing with the "I can't log in!" excuse, it's because it adds time to the testing process. Nonetheless they did it before "I can't login derp!" That it favors Intel by not having it appear is a nice ancillary benefit for them however. Those average game charts lacking most of the biggest wins for 3D cache skews things.

If reviewers wanted to be logically consistent about why they choose games for their test suites, they'd include more "niche" titles. Because again, every type of game is a niche. Arguing any given title is an outlier due to its small player base also rings hollow. The leaked bar graphs for the new Zen 4 3D have Tomb Raider and Far Cry 5. Rise of the Tomb Raider averaged 800 players on Steam the last month. Assetto Corsa averaged 10 times that, let me type that again, 10 times. Yet which is the staple of review suites?

Not testing titles like MSFS, Assetto Corsa, Factorio, Stellaris, or Planet Coaster over other games no one plays like Tomb Raider, is a bummer. Limit the scope, limit the findings. Which creates doubt concerning the reviewers desire to truly explore the reaches and effects of the new cache technology. Otherwise they'd include more non gaming tests that are cache sensitive, as suggested by others here. And include more cache sensitive games as I have suggested.

I'll end my remarks by saying it is beneficial to at some point, spend the time on those titles for the loyal fan bases they have if nothing else. While serving the dual purpose of investigating the performance, which is a big part of their mission statements. And it certainly isn't impossible to do. Here is Wendell testing the 3D cache in Stellaris, Factorio, and Dwarf Fortress. Spoiler, it made the contemporary Intel CPU look pedestrian.

Note for the anticipated responses about scientific methodology and repeatable results. They have plenty of those already. A bit more real gameplay like in Stellaris and Planet Coaster would be most welcomed.

 

[Timmah!]

Worthwhile outlets certainly will. I feel like good cooling will play an even bigger role with the X3D parts, in part because of the new 89C soft limit.

Here's 120/162W (stock) R23 result:
37.973 points
View attachment 77179

But we're gonna need more than CB and GB lol

If the reviews tomorrow reveal this to be clocked like 200 MHz less than vanilla chip and this to be fixable by some basic manual tuning (not Det0x inhuman level of tinkering ) to make it run at more or less same clocks that out of the box 7950x runs, i will be on the phone to the local reseller i got the 7950x from, whether they wont take it back (since its still not used and box un-opened/sealed) and sell me 3D instead.

 

[maddie]

There we go with the niche claims again. How niche, and what does that even mean? A game not showing up on the Steam top played games doesn't mean there isn't a big number of players worldwide.

For example: WoW still has something like 8.5 million monthly subscribers. It is difficult to assess games available through game pass too. MS Flight Sim had an estimated 10 million players across all platforms in December. That's why it should be in every major review. Yet it isn't. Dispensing with the "I can't log in!" excuse, it's because it adds time to the testing process. Nonetheless they did it before "I can't login derp!" That it favors Intel by not having it appear is a nice ancillary benefit for them however. Those average game charts lacking most of the biggest wins for 3D cache skews things.

If reviewers wanted to be logically consistent about why they choose games for their test suites, they'd include more "niche" titles. Because again, every type of game is a niche. Arguing any given title is an outlier due to its small player base also rings hollow. The leaked bar graphs for the new Zen 4 3D have Tomb Raider and Far Cry 5. Rise of the Tomb Raider averaged 800 players on Steam the last month. Assetto Corsa averaged 10 times that, let me type that again, 10 times. Yet which is the staple of review suites?

Not testing titles like MSFS, Assetto Corsa, Factorio, Stellaris, or Planet Coaster over other games no one plays like Tomb Raider, is a bummer. Limit the scope, limit the findings. Which creates doubt concerning the reviewers desire to truly explore the reaches and effects of the new cache technology. Otherwise they'd include more non gaming tests that are cache sensitive, as suggested by others here. And include more cache sensitive games as I have suggested.

I'll end my remarks by saying it is beneficial to at some point, spend the time on those titles for the loyal fan bases they have if nothing else. While serving the dual purpose of investigating the performance, which is a big part of their mission statements. And it certainly isn't impossible to do. Here is Wendell testing the 3D cache in Stellaris, Factorio, and Dwarf Fortress. Spoiler, it made the contemporary Intel CPU look pedestrian.

Note for the anticipated responses about scientific methodology and repeatable results. They have plenty of those already. A bit more real gameplay like in Stellaris and Planet Coaster would be most welcomed.

I have a couple friends that only play/use MSFS and even built their systems purposefully for that program.
The concept of average is a mathematical construct that often has no real world reality. 2.2 children/family? How many earns the average salary?

 

[DrMrLordX]

There we go with the niche claims again.

It isn't a claim, it's just a fact. Stuff like CS:GO gets featured because more people play it than other games. I'm sure someone playing 4x games would like to see their titles featured as well, but

a). it isn't quick and easy to benchmark that stuff in a relevant fashion and
b). reviewers only reach a small audience by featuring those titles.

Anyone interested in knowing how well Stellaris runs on a 7950X3d will just have to buy the hardware and do a comparison on their own. Ideally published for the benefit of their community on a forum somewhere.

 

[Thibsie]

It isn't a claim, it's just a fact. Stuff like CS:GO gets featured because more people play it than other games. I'm sure someone playing 4x games would like to see their titles featured as well, but

a). it isn't quick and easy to benchmark that stuff in a relevant fashion and
b). reviewers only reach a small audience by featuring those titles.

Anyone interested in knowing how well Stellaris runs on a 7950X3d will just have to buy the hardware and do a comparison on their own. Ideally published for the benefit of their community on a forum somewhere.

Yeah but at the same time, that CS:GO ran @144fps or @35645 fps, nobody cares but 'I wanna have a bigger one than you type of kidos.
It changes nada to the CS:Go experience. So it is useless in the end.

Let's put CS:Go and type of games on a lone 4K@ultra settings. It plays @ 500fps ? Cool. Now interesting stuff.

 

[biostud]

If you're already testing 50 games, 20 of which all the other sites also tests, then it would be obvious to attract the "niche" crowd by change a couple of games to differentiate yourself.

 

[Timmah!]

Yeah, this catering to lowest common denominator is so annoying. There is crapton of various outlets doing the reviews, but majority of them is usually hell-bent on doing exactly the same basic stuff everyone else does. Cause why bother and offer something special and different, with added value, right?
Words cant describe my surprise when few years ago i found out Tech-gage actually reviews the application performance of GPUs, and not just games.

 

[In2Photos]

Yeah but at the same time, that CS:GO ran @144fps or @35645 fps, nobody cares but 'I wanna have a bigger one than you type of kidos.
It changes nada to the CS:Go experience. So it is useless in the end.

Let's put CS:Go and type of games on a lone 4K@ultra settings. It plays @ 500fps ? Cool. Now interesting stuff.

I personally think that reviewers keep testing CS:GO to see which CPU hits the 1,000 fps barrier at 1080p.

 

[misuspita]

My wife reminded me

Mate, that's as off topic as it gets, but if you want your wife happy, put a yearly reminder on your calendar, preferably a day before, to buy at least a flower

 

[DAPUNISHER]

If you're already testing 50 games, 20 of which all the other sites also tests, then it would be obvious to attract the "niche" crowd by change a couple of games to differentiate yourself.

I don't even get this continued talk of niche. As if only a few 1000 people own or play these titles. MSFS has over 10 million players total. That means there are millions playing on PC. Hardly a small niche title. Next, the goalpost gets moved to "It's hard.". Professionals being paid to test games, shouldn't include games that are hard to test? LOL Nonsense. If they can run around Hogsmeade, then they can save a file flying around NY City.

Factorio disappeared from testing suites. I was seeing some reviewers featuring that one before. Odd how both favor 3D cache and are no longer tested. I'd be very interested to see how Zen 4 3D does in both. Hopefully at least one reviewer will include these titles again.

 

[Markfw]

I don't even get this continued talk of niche. As if only a few 1000 people own or play these titles. MSFS has over 10 million players total. That means there are millions playing on PC. Hardly a small niche title. Next, the goalpost gets moved to "It's hard.". Professionals being paid to test games, shouldn't include games that are hard to test? LOL Nonsense. If they can run around Hogsmeade, then they can save a file flying around NY City.

Factorio disappeared from testing suites. I was seeing some reviewers featuring that one before. Odd how both favor 3D cache and are no longer tested. I'd be very interested to see how Zen 4 3D does in both. Hopefully at least one reviewer will include these titles again.

You mean you don't believe the stories of Intel people bribing people to test CPUs in a way that Makes Intel CPUs look good ???