Ryzen AM5 ; 7000

[RasmusP]

Thought since there's an Intel 13th gen thread and nvidia 4xxx thread, I'd create a Ryzen 7000 thread or rather a AM5 thread to have a new one for the new AMD chapter with DDR5.

In 3.5 hours AMD will tell us more about the B650 Motherboards.

I'm definitely planning an upgrade to the 7600X but I also have an excel sheet ready with "FPS per €" and won't buy anything if that ratio is too much worse compared to a 12400f+B660+DDR4 or the 5600+B550+DDR4.

100€ extra for more raw performance and future upgradeability (is this a word?!) is my limit.

My prediction is that the ratio will only become a thing when the X3D variant(s) will arrive next year.


Are you planning to buy a Ryzen 7xxx?
Which motherboards will you be buying and why?

Any comparison benchmarks, loads+temps etc. to your current/previous systems?

This is the thread for it

I'll keep you updated with my excel sheet (will probably move it to be a Google sheet) and if I'll buy anything, compare it to my 10600k.

 

[RasmusP]

Just got the first version of my Google sheet ready with the ACC benchmarks from PurePC.pl and Hardware Unboxed (written down at Techspot):

CPU Price/Performance

docs.google.com

I'll edit the Full System charts once the "value" B650 boards will become available.
"Value Mobo" and "Value ram" is what I would buy.

Mobo:
8x USB and quality look, backplate shroud preferred.
For B650, I put in 200€ as placeholders.
RAM: DDR4 = 3600 CL18 ; DDR5 = 5600 CL40 (for now).

Here are the charts!
PurePC 7600x is interpolated from other gens! That's why the 7700X is in this chart although it's not a budget CPU.

PurePC.pl:

Hardware Unboxed:

 

[Novopaine]

While the performance is better than the previous gen, I was hoping for a bit more from Ryzen 7000 out of the gates. I may hold out for the 3D cache variants. The 5800X3D is currently available to me locally right now, and the CPU cost sites right between the 7600X and 7700X, plus the platform is mature. Then there's Intel's upcoming products to check out. These decisions are killing me. lol

My wish list includes 2.5Gb Ethernet, WIFI 6E (possibly for wireless VR), and at least 8 rear USB ports. From what I've seen so far this should be an easy order to fill regardless of brand.

I haven't picked a mobo brand or model yet, but I've always had good experience with ASUS when it comes to quality. My only gripe with ASUS is their clueless tech support. I tried to report a BIOS bug to them in 2019. What a painful experience that was! All they kept offering me was to RMA my board.

EDIT: forgot to mention, I think I'm targeting 8 cores for my CPU, so the 7700X or a future 7800x would be a possibility for me.

 

[RobPhoboS]

I'm in the same boat as you. But I'm thinking 5800x3d now, it'll be a whole new pc for me as my current one is +10 years
But the non 3d chips don't appeal to me at all.

 

[RasmusP]

EDIT: forgot to mention, I think I'm targeting 8 cores for my CPU, so the 7700X or a future 7800x would be a possibility for me.

May I ask for what use case?

About 5800x3D vs going with the 7000:
Really difficult decision.. Building an awesome combo now or hoping for a better CPU down the road.
How much cheaper will it be to pay extra now and replace the CPU later on with "slow" ddr5 and a feature lacking mobo now vs simply buying a good combo now and buying a completely new combo in 2-4 years?


My personal issue is that I'd like to get some single thread workstation performance gains and the 5800x3D isn't great at this...

Seeing in clear charts how bad the value of the new CPUs is helps a lot for being happy to wait for a 7600X3D or 7600 non-x together with getting cheaper ddr5 6000 cl30 and hopefully cheaper mobos!

Let's see what the Intel 13400 brings. I didn't really had the 12400f in my brain but looking at my sheet, it looks quite awesome, lol

 

[Novopaine]

May I ask for what use case?

The goal for the new PC is as a dedicated rig capable of any Sim racing title, but once I finally get a new VR headset it will go double duty for stand up VR as well. While I don't stream or anything like that, I might decide to one day and the extra cores may come in handy. 6 cores is probably enough for my immediate needs, but eventually this will become a hand-me down PC that will replace one of the many others I use.

Currently in service
- R9 3900x - all-purpose/daily driver
- R5 5600x Truenas Scale server (I actually plan to swap the CPU with the 3900x eventually)
- i7 3770K - HTPC for the couch (my old daily driver)
- ????? - future dedicated Sim/VR PC

Yes... I know. I have issues.

 

[Neilski]

My personal issue is that I'd like to get some single thread workstation performance gains and the 5800x3D isn't great at this...

Well, the Passmark single-thread (and multi-thread in fact) performance for the 5800X3D is dramatically lower than the 7700X, and yet the former wipes the deck with the latter in the gaming data in your second post, proving that all of that extra cache changes the equation pretty dramatically in certain cases.
But I guess you have probably compared the workstation benchmarking for things reasonably similar to your intended use-case?

 

[RasmusP]

Well, the Passmark single-thread (and multi-thread in fact) performance for the 5800X3D is dramatically lower than the 7700X, and yet the former wipes the deck with the latter in the gaming data in your second post, proving that all of that extra cache changes the equation pretty dramatically in certain cases.
But I guess you have probably compared the workstation benchmarking for things reasonably similar to your intended use-case?

Yep, I did
It's also not every game and for some reason, ACC has one of the craziest gains from the big cache!
I sadly have no data about AC, rF2, AMS2 etc...

The 5800X3D might very well be a bit worse for these other simracing titles!

Counter-Strike GO for example:
7600X = 487 fps
5800X3D = 374 fps

12 games average shows:
7600X = 211/166 fps (avg/1% low)
5800X3D = 203/158 fps

As for "Workstation" tasks:

PurePC.pl did an Unreal Engine 4 compilation in Visual Studio:
7700X = 71.5 seconds
5800X3D = 90.2 seconds

7-Zip Test:
7700X = 120418 "Operations"
5800X3D = 98545

HWUB/Techspot Adobe Photoshop Benchmark (closest to Lightroom):
7600X = 1439 points
5800X3D = 1198 points


However, any of these CPUs would be quite the upgrade for me. Comparing all these to my 10600k

Techspot:
ACC:
5800X3D = 198 fps
7600X = 170 fps
10600k = 104 fps

CS:GO
5800X3D = 374 fps
7600X = 487 fps
10600k = 259 fps

12 games average:
5800X3D = 203 fps
7600X = 211 fps
10600k = 145 fps

Photoshop:
7600X = 1439 points
5800X3D = 1198 points
10600k = 841

PurePC:
7-Zip Test:
7700X = 120'418 "Operations"
5800X3D = 98'545
10600k = 57'043

Visual Studio:
7700X = 71.5 seconds
5800X3D = 90.2 seconds
10600k = 150.9 seconds

 

[Neilski]

It's also not every game and for some reason, ACC has one of the craziest gains from the big cache!

Ahhh, good. That makes a lot of sense (as does the rest of your data).
Based on that set of figures, my instinct is that the "excess" performance in ACC for the X3D is simply because ACC is kinda sucky. Or to be more specific: I reckon it has been poorly optimised for CPUs that don't have bloody enormous caches (i.e. almost all of them).
In the past, I've seen different compilers give enormous speed differences on a code that's particularly cache-size-sensitive. Alternatively a good programmer can prevent a code from being more greedy than necessary on that front.

 

[pattikins]

Here's an interesting take on the cost of making (materials) the 7000 cpu.

The AMD 7950X costs £764 in the UK.

 

[R8pilot]

That analysis doesn't include a myriad of other costs:

* Hiring and paying hundreds of expensive engineers and others to do the design, testing, and so on, plus the absurd number of other employees involved in every aspect of the whole process.
* Maintaining supercomputer-level simulation facilities and their uber-expensive CAD software licenses.
* Purchasing and maintaining the extremely expensive testing hardware and the time and effort going into testing.
* Logistics, other packaging materials, shipping, advertising.
* Paying bribes "consideration" to officials, Youtube "influencers" and others.
* Fat, fatter, fattest stock awards, salaries, private jets, etc., to the 1%ers in the company.
* Tons else that I haven't thought of in this brief response.

 

[Neilski]

Here's an interesting take on the cost of making (materials) the 7000 cpu.

Hehe. It cracks me up to see him estimating chiplet dimensions to the nearest micron, calculating the chip area to no less than *5* significant figures (wtf!), and then admitting that for the actual wafer fabrication cost, "estimates vary wildly" so he just went with a nice round $17k as a ballpark-sensible number. Oooook. So, you didn't really need that level of (spurious) precision on the chip size, did ya Mr. TechTechPotato?

Would be interesting to ponder if he got the marginal cost per chip to within a factor of 2 (seems quite possible) but for sure that's only a small part of the whole story, especially for the flagship product of the line.

 

[Wrath85_ita]

7700X here, i try to post some video for testing but i only have a 2070 Super so my system is pretty gpu limited.

Assetto corsa competizione - imola 29 cars

Project cars 2 - brand's hatch full grid

F1 22 - monza

started last with a slow pace to stay in the traffic as much as possible.

 

[pattikins]

ACC seems to be GPU bound. Maybe try @ 1080p resolution, a camera position where the mirrors are visible & increase the number of car visible to the max....

Do you have Assetto Corsa [installed]?

 

[Wrath85_ita]

ACC seems to be GPU bound. Maybe try @ 1080p resolution, a camera position where the mirrors are visible & increase the number of car visible to the max....

Do you have Assetto Corsa [installed]?

at reduced resolution obs made me mad to record the gameplay, for that reason i set low/lowest settings and DLSS ultra performance (which is really horrible to see at work).
the same happened with Pcars2.
now i'm uploading a clip from F1 22 which i was able to record at 2560x1440 with a gpu load around 90%.

Ac it's the main game i play, no problem to record a test BUT performances depends a lot by mods and settings, to me it's even harder to be cpu bound instead of gpu bound.

 

[pattikins]

...
now i'm uploading a clip from F1 22 which i was able to record at 2560x1440 with a gpu load around 90%.

Ac it's the main game i play, no problem to record a test BUT performances depends a lot by mods and settings, to me it's even harder to be cpu bound instead of gpu bound.

For AC, I was thinking of Lemans with 54 AI + one human driver - monitoring the framerate as the other cars pull off & disappear @ the start of a quick race. No post processing & just a few basic apps.

My set up is:
AMD 5800X3D
Nvidia 3060ti
Memory @ 3000mhz.
res. 1440p

 

[Wrath85_ita]

For AC, I was thinking of Lemans with 54 AI + one human driver - monitoring the framerate as the other cars pull off & disappear @ the start of a quick race. No post processing & just a few basic apps.

My set up is:
AMD 5800X3D
Nvidia 3060ti
Memory @ 3000mhz.
res. 1440p

custom shader patch disabled with following settings:

Spoiler

race session settings:

Spoiler

 

[RasmusP]

I'll chime in with a 10600k + 3080 later today. Since I'm getting the most annoyed in AC from cpu micro stutters, it's very interesting to see 5800x3D vs 7700x!

Thanks for the settings screenshot!
With this, csp disabled, no post processing and no big apps like substanding, acti etc., it should be very comparable!
I'll try to use exactly the same apps!

We should make it a bit more accurate though with the exact same session settings too!
So same cars, same time of day, same weather.

2070 super at around 30% is definitely not bottlenecking and 20% load on the 7700x perfectly shows how annoyingly single threaded AC is lol.


Edit:
Did the testing

Using MSI Afterburner Benchmark Tool, hitting start when the lights go out and ending it at 30.00 seconds.
Took a Screenshot right before starting it and just after ending it.

Here's the result:
31-10-2022, 20:15:26 acs.exe benchmark completed, 3609 frames rendered in 30.172 s
Average framerate : 119.6 FPS
1% low framerate : 75.0 FPS

From within Afterburner:

I sadly couldn't record a second run with OBS. AC keeps crashing as soon as I start the record. Probably to do with the resolution, disabled scaling, gsync, whatever... But the data is there anyway!


Btw, we could also just monitor the fps for 10 seconds before starting the race.. It seems to be pretty tough for the CPU!
31-10-2022, 20:40:01 acs.exe benchmark completed, 802 frames rendered in 10.172 s
Average framerate : 78.8 FPS
Maximum framerate : 86.8 FPS
1% low framerate : 63.8 FPS

 

[Wrath85_ita]

i'm at work now and can't do anything else at this moment but i updated last post with session's settings

 

[Wrath85_ita]

I'll chime in with a 10600k + 3080 later today.

great job!
i used capframeX because in afterburner i can't display every cpu spec (like temp, power and clock) even if i set "CPU.dll on" in the monitor page.

here AB bench with same settings, starting at the end of red lights, 2 run(1 of 10 sec and 1 of 30 sec):
31-10-2022, 21:23:18 acs.exe benchmark completed, 1936 frames rendered in 10.094 s Average framerate : 191.7 FPS
Minimum framerate : 165.2 FPS
Maximum framerate : 215.4 FPS
1% low framerate : 144.5 FPS
0.1% low framerate : 134.6 FPS

31-10-2022, 21:17:08 acs.exe benchmark completed, 7113 frames rendered in 31.391 s
Average framerate : 226.5 FPS
Minimum framerate : 171.6 FPS
Maximum framerate : 270.5 FPS
1% low framerate : 150.3 FPS
0.1% low framerate : 135.2 FPS

tomorrow i can try to run ACC e Pcars2 in windowed mode at 1080p or 720p testing that way if you want to collect more data.