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"

 

[Det0x]

^^ continuation from last page (max 10 pictures per post)

2033MT/s is the tipping point for this cpu.. above this speed i'm clearly getting reduced scaling.

2066: (t2)
Linpack extended 8gb = 349.2 average gflops
Y-cruncher 2.5b = 96.755s



2100: (t2)
Linpack extended 8gb = 342.7 average gflops
Y-cruncher 2.5b = 101.177s


Maxed aida numbers @ 2100:4200 T1 GDM disable

Pretty sure i will, but i haven't tested if i see performance scaling in light programs like games all the way up to 4100MT/s yet.

----------------------------------------------------------------------------------------------------------------------

Stability testing 2033:4066:

~430 whea over a 30min 1usmus testmem run


Started karhu ram test before i went to work, lets see what results i'm getting when i get home:
(25 cycles testmem 1usmus cfg already done without error)

 

[ZGR]

Just loaded up my X4: Foundations save. The 5775C only got around 20ish FPS during this battle. The 5800X3D is keeping my game above 80fps no problem. This kind of performance increase was something I could not have imagined. Feels surreal!

 

[DrMrLordX]

It's the Price, They are selling a 16 Core Milan-X for $4,000 and That CPU is Just Owning the 5950X in HPC based applications thanks to the stacked 3D$, so a 5950X3D would be undermining that Milan-X Segment.

Not directly. Yes, it means AMD would lose some CCDs to a commercial product that would sell for less than $1k, but it would not replace a Milan-X CPU in someone's workstation or server since the 5950X3D system would be missing features essential for serious work machines.

Just look at this. Milan-X vs Milan and Alder Lake

Sorry to sound like a broken record, but when you post Phoronix geometric mean results, the data you're getting is pretty skewed based on the benchmarks selected by Michael Larabel. It is not, for example, representative of the workloads an enthusiast "power user" typical of these forums might choose to run on their systems. Though it is curious that Phoronix's test suite loves extra L3 that much. It likes ST performance quite a bit as well.

 

[JoeRambo]

Just loaded up my X4: Foundations save. The 5775C only got around 20ish FPS during this battle. The 5800X3D is keeping my game above 80fps no problem. This kind of performance increase was something I could not have imagined. Feels surreal!

Yup, and i bet it would not show up if there were benchmarks on the web, testing some built in benchmark, or a save file 5 minutes after start of game. It is really the same with plenty of other games like Anno 1800, Stellaris etc. It takes a strong CPU with great memory subsystem to shine in those worst cases.

 

[nicalandia]

Sorry to sound like a broken record, but when you post Phoronix geometric mean results, the data you're getting is pretty skewed based on the benchmarks selected by Michael Larabel.

It's actually from the openbenchmarking.org Website and info posted by many contributors/testers/enthusiasts that upload their results into the data base. It's not a single source posted by Mr. Larabel. It has been proven to be a very well balanced source of data for Linux/Unix/BSD performance metrics.

 

[DAPUNISHER]

Yup, and i bet it would not show up if there were benchmarks on the web, testing some built in benchmark, or a save file 5 minutes after start of game. It is really the same with plenty of other games like Anno 1800, Stellaris etc. It takes a strong CPU with great memory subsystem to shine in those worst cases.

Preach it Rambo. How good a CPU can be in those situations, in this instance, the 3D, is where big reviewers fail miserably.

A member was looking to go from a 7700K to a 5900x a couple months back. Said he mostly played games like Planet Coaster. Another member posted results from a well known reviewer, and there was no big uplift between them in the game. I find a guy on youtube that did the same upgrade, and it more than doubled his performance. Of course, he had a massive park with 1000s of customers, which took many hours of game play to build up. The reviewer? Not so much.

To conclude my weekly rant about how everyone has been unintentionally gaslit into a false paradigm. Throwing around reviewer numbers in discussions carries too much weight with most. Stop relying on them as the last word on CPU performance. They should be considered more of a ABCs of the new product, not the holy book for it.

 

[Insert_Nickname]

Of course, he had a massive park with 1000s of customers, which took many hours of game play to build up. The reviewer? Not so much.

You'd think it'd be easy to get a late game save from someone to test late game. Too much effort apparently.

 

[Hitman928]

You'd think it'd be easy to get a late game save from someone to test late game. Too much effort apparently.

One of the worst sites for this is GameGPU, or at least they used to be. When Doom (2016) came out, they had results that were a bit different than most. If you dug into what their test scene was, it was at the very beginning of the game where they ran around a small, empty room for about 20 seconds and then shot an exploding barrel. That was it, literally. Not even a single enemy on screen or anything happening in the game. It was a terrible test case.

 

[Schmide]

Sorry to sound like a broken record, but when you post Phoronix geometric mean results, the data you're getting is pretty skewed based on the benchmarks selected by Michael Larabel. It is not, for example, representative of the workloads an enthusiast "power user" typical of these forums might choose to run on their systems. Though it is curious that Phoronix's test suite loves extra L3 that much. It likes ST performance quite a bit as well.

Did you read the article? They explain on the last page.

Anyhow, from some 285 benchmarks carried out under Linux and ignoring the Linux gaming mess, the side-by-side comparison for workloads with a measurable difference included (benchmarks with less than a 2% difference either way omitted):

For workloads able to properly utilize AMD #D V-Cache, the 96MB L3 cache on the Ryzen 7 5800X3D really pays off. But 61% of the time, the Ryzen 7 5800X was faster particularly for all the general purpose workloads/benchmarks:

They didn't select the benchmarks. They did all the benchmarks and reported on the outliers.

It is good data and good reporting.

 

[nicalandia]

Did you read the article? They explain on the last page.

I believe he was saying that when I reference The geometric mean result from a Pseudo 5950X3D whaling on The regular 5950X. and Alder Lake The thing is that I did not take it from Phoronix but directly from the source

Intel Core i9-12900K vs. AMD EPYC 7373X vs. AMD Ryzen 9 5950X Benchmarks - OpenBenchmarking.org

openbenchmarking.org

 

[gdansk]

The X3D is particularly hard to know if it will help. Some games love it and others don't. You can look in reviewer tests for games you play.
But you really need someone to compare a 5800X or 5800X3D in the same saved file in your specific game. Obviously that would be difficult.

And even if AMD made a cache pressure tool available it looks too complicated for most people (I mean me) to interpret the results correctly.

 

[ZGR]

Running R6 Shadow Legacy with 100 AI on Clubhouse. My i7 got around 90 FPS, the 5800X3D is getting a bit over 2x. I still need to tweak the RAM though.

Hitting 200 FPS with 100 AI is crazy. Hitting GPU bottleneck. Wish I got to benchmark more CPU's for this game.

I am still experiencing occasional stutters, but it is a lot better than the old CPU.

 

[Schmide]

I believe he was saying that when I reference The geometric mean result from a Pseudo 5950X3D whaling on The regular 5950X. and Alder Lake The thing is that I did not take it from Phoronix but directly from the source

Intel Core i9-12900K vs. AMD EPYC 7373X vs. AMD Ryzen 9 5950X Benchmarks - OpenBenchmarking.org

openbenchmarking.org

View attachment 60710

That's on you then. Source your source and context. Remember a lie by omission is still a lie.

Regardless the Dr referenced Michael Larabel which put the context directly on the article which if read had complete context.

 

[nicalandia]

That's on you then. Source your source and context. Remember a lie by omission is still a lie.

Regardless the Dr referenced Michael Larabel which put the context directly on the article which if read had complete context.

But I did put the source.

He just thought I was posting it directly from Phoronix and that Dr. Larabel just picked the benchmarks.

Here is the Original Post I made with reference: post number #3,172

 

[Schmide]

But I did put the source.

He just thought I was posting it directly from Phoronix and that Dr. Larabel just picked the benchmarks.

Here is the Original Post I made with reference: post number #3,172

Note: The post the Dr referenced was before that and just images with no link. post #3,169.

Edit: Regardless bypassing the article and going directly to the database is omitting data. The context should be clear to everyone here. If one did not read the article or chooses to pick data on their own out of context, that is on them.

 

[biostud]

We can run all kinds of benchmarks, and personally I don't do anything that is very CPU intensive. And sometimes I wonder how many are really relevant to the majority of users.

For me all I would need to know are World of Tanks, Borderlands 3 and NEF to jpg conversion performance. Then 95% of my PC usage would be covered, besides office and browsing.

 

[Justinus]

We can run all kinds of benchmarks, and personally I don't do anything that is very CPU intensive. And sometimes I wonder how many are really relevant to the majority of users.

For me all I would need to know are World of Tanks, Borderlands 3 and NEF to jpg conversion performance. Then 95% of my PC usage would be covered, besides office and browsing.

For a while I've had an idea of reviewing CPU's/platforms with a completely subjective test suite. How snappy does it feel? Opening windows, closing them, opening everyday programs like browsers/office/etc and doing basic tasks in them and closing them. I think the problem with this would be consistency, although with some of the advanced scripting reviewers have done for other test suites before perhaps a "normal user" workflow could be scripted and the time it takes measured.

I think most people assume the fastest CPU/platform in benchmarks will be the most snappy and responsive for everyday computing, and that might be right.... but how do we know?

 

[Shamrock]

quick question:

You think I would benefit from a 5800X3d (would also have to buy a heatsink) with only a Sapphire RX 6700xt? Running a 5600x now.

 

[igor_kavinski]

quick question:

You think I would benefit from a 5800X3d (would also have to buy a heatsink) with only a Sapphire RX 6700xt? Running a 5600x now.

Depends on the games you intend to play and the resolution. 1440p, maybe not that great a difference. 1080p, could be a significant boost depending on how well the game responds to the additional cache.

 

[BTRY B 529th FA BN]

Running R6 Shadow Legacy with 100 AI on Clubhouse. My i7 got around 90 FPS, the 5800X3D is getting a bit over 2x. I still need to tweak the RAM though.

Hitting 200 FPS with 100 AI is crazy. Hitting GPU bottleneck. Wish I got to benchmark more CPU's for this game. View attachment 60711

I am still experiencing occasional stutters, but it is a lot better than the old CPU.

Do you have fTPM disabled in bios as well as AMD PSP 11.0 Device disabled through Device Manager? If not you need to do both to get rid of system stutter caused by fTPM and AMD PSP.. IF that's what's causing it.

 

[dr1337]

quick question:

You think I would benefit from a 5800X3d (would also have to buy a heatsink) with only a Sapphire RX 6700xt? Running a 5600x now.

What games do you primarily play?

 

[ZGR]

Do you have fTPM disabled in bios as well as AMD PSP 11.0 Device disabled through Device Manager? If not you need to do both to get rid of system stutter caused by fTPM and AMD PSP.. IF that's what's causing it.

Thanks for the tip, I will check tomorrow! I am still on mostly default BIOS settings...Thanks for all the tips so far BRTY, you have been so helpful!!

I am fairly confident it is just because Siege wasn’t ever intended to have more than 16-20 alive AI at a time. We love the pve coop mods for it though.

If I place 100 AI inside each other, the game will slow down to 30fps and stutter like mad. But as soon as they spread out, the stutters go away.

If I did that on the 5775C, I would have crashed at 50 AI and above. I can finally host 10 player coop with 100 AI!

But so far my performance in other games is fantastic. 1080 ti is finally the bottleneck in most games.

 

[Shamrock]

What games do you primarily play?

Mechwarrior Online and Space Engineers

 

[DrMrLordX]

It's actually from the openbenchmarking.org Website and info posted by many contributors/testers/enthusiasts that upload their results into the data base. It's not a single source posted by Mr. Larabel.

Yeah but I'm pretty sure the benchmark selection is just Phoronix Test Suite. Openbenchmark just records the data submitted by people running PTS.

If I'm wrong I'm wrong, but if you're linking a geometric mean from an article published on Phoronix, then it's usually a segment of PTS.

 

[igor_kavinski]

Mechwarrior Online and Space Engineers

From https://forums.overclockers.co.uk/posts/35576145/

Slightly vague and anecdotal, but the 5800X3D has eliminated fps lows in the online competitive game 'Mechwarrior Online' that I play. And that's 3440x1440 low settings with a 1080Ti. It's now pegged at the screen refresh rate and never moves, it's noticeably smoother.

Space Engineers already hits 90+ FPS with Skylake so I'm guessing you are enjoying more than that. You want even higher fps in this game?