Zen 3

[Eschaton]

Looking for Benchmarks, anybody see any yet?

Specifically interested in 5950X versus 10900K at 4k resolution in fps for WoW.

[Thunderball]

Looking for Benchmarks, anybody see any yet?

Specifically interested in 5950X versus 10900K at 4k resolution in fps for WoW.

Benchmarks are out, pick your poison. To answer your question:

1) WoW doesnt scale with so many threads.
2) There's not going to be any difference because it's going to be GPU/draw calls bound.

To summ up Zen 3.

- First AMD chip since Athlon 64 to beat Intel stuff in average gaming FPS.
- AMD now leads in all types of production workloads, including the ones that are IPC/frequency dependant.
- Maximum frequency gains are unexpectedly good: 4.6-4.7 GHz all core is possible now, boosting up to 5 GHz on high quality silicon.
- 20% IPC increase was a hoax as expected. There is 20-25% performance uplift, but most of it is due to frequency. IPC increase is between 7-12%.
- Intel still leads in minimum FPS. This is due to inherent memory architecture drawbacks, which are very unlikely to be solved ever.
- 5600X might not be worth it considering the price, especially if 3600 actually starts costing meaningfully lower than 3600X again.

[mrgreenthump]

- Intel still leads in minimum FPS. This is due to inherent memory architecture drawbacks, which are very unlikely to be solved ever.

I agree with all but this, from what I saw minimums were about even exept a few exception. And the biggest oddity being CSGO where 5900X 95th %ile min is almost the same as 10900k average.

[Thunderball]

I agree with all but this, from what I saw minimums were about even exept a few exception. And the biggest oddity being CSGO where 5900X 95th %ile min is almost the same as 10900k average.

For CS:GO it literally doesnt matter. People play at 1024х768 stretched mostly and you get 500-800 FPS with modern systems. You just dont have monitors with such refresh rate. People that play with 240Hz monitors cap their FPS at 300.

[mrgreenthump]

For CS:GO it literally doesnt matter. People play at 1024х768 stretched mostly and you get 500-800 FPS with modern systems. You just dont have monitors with such refresh rate. People that play with 240Hz monitors cap their FPS at 300.

I know it doesn't matter, but it still doesn't make it not true.

[Eschaton]

Benchmarks are out, pick your poison. To answer your question:

1) WoW doesnt scale with so many threads.
2) There's not going to be any difference because it's going to be GPU/draw calls bound.

I understand WoW doesn't scale with many threads. But even in single threaded performance the 5950X seems to have an advantage on most benchmarks I have seen.

What I don't understand is that many people say WoW is a CPU bound game, not GPU....so would the 5950X not have an advantage? Or is simply a wash with the 10900K?

[Thunderball]

I understand WoW doesn't scale with many threads. But even in single threaded performance the 5950X seems to have an advantage on most benchmarks I have seen.

It does but in 4K regardless of the game that advantage wont manifest because it's going to be GPU bound.

What I don't understand is that many people say WoW is a CPU bound game, not GPU....so would the 5950X not have an advantage? Or is simply a wash with the 10900K?

It's a CPU bound game, yes, but:

- It's dependant mostly on single core performance. 10 or 16 core CPUs of any kind of help in any possible way, especially in 4K.
- The reason that your FPS drops in raids comes down to game's poor optimization and how it works with memory. The core of the engine and netcode is built for a single core CPU. So regardless of a CPU you pick you're going to bottleneck on draw calls.

[Ange]

What I don't understand is that many people say WoW is a CPU bound game, not GPU....so would the 5950X not have an advantage? Or is simply a wash with the 10900K?

Somewhere close to FS2020 numbers in comparison between AMD and Intel could be expected: at STOCK



But WoW scales in lower resolutions pretty well with CPU and with RAM OC. With a standard 300second benchmark in a controlled environment I get this numbers in WoW:

Intel 10900k 4.9GHz (stock) - 5.2GHz (OC) - 5.3GHz (OC)
32GB RAM 4000MHz (stock( - 4400MHz (OC) both with subtiming tuning

300 second RTSS benchmark, 1080p, graphics-7, controlled medium load, filtered loading fps peak

4.9 4000 min-FPS: 249 max-FPS: 255
5.2 4000 min-FPS: 270 max-FPS: 280
5.2 4400 min-FPS: 274 max-FPS: 280
5.3 4000 min-FPS: 274 max-FPS: 278
5.3 4400 min-FPS: 294 max-FPS: 298

=> loading FPS is filtered out (3500-4000 FPS peak, useless data)
=> CPU OC and RAM OC have certain sweetspots, look at the jump at 5.3GHz with 4000 vs 4400 Memory: +400MHz more memory was more than overclocking 10/20 core CPU from 4.9GHZ to 5.2GHz (+300MHz all-core)

This is something you cant do in this scale with ZEN3, because:
* memory tuning is limited by the FCKL
* CPU OC is limited with much less overhead in thermals
* CPU OC is limited by better bins only available in the high core count CPUs and high core count CPUs are over multiple CCDs (chiplets) distributed that causes latency issues in games, because windows will assign core changes between different CCD's (windows 10 task scheduler is still full intel optimized).

As soon as you count in the huge OC range with Intel, the comparisons with CPU heavy games change a lot.

[mrgreenthump]

4.9 4000 min-FPS: 249 max-FPS: 255
5.2 4000 min-FPS: 270 max-FPS: 280
5.2 4400 min-FPS: 274 max-FPS: 280
5.3 4000 min-FPS: 274 max-FPS: 278
5.3 4400 min-FPS: 294 max-FPS: 298

Somehow I doubt this test was done in a mythic raid or Boralus. Anything modern can run WoW at 100+ on average, but they all tank like hell in certain situations(raids and cities). Question now is which CPU tanks the least.

As soon as you count in the huge OC range with Intel, the comparisons with CPU heavy games change a lot.

From what I saw 10900K at 5.3GHz still lost to a 5900X at stock on many games.

Edit: Yup I suck at writing

[Temp name]

From what I saw 10900K at 5.3GHz still lost to a 5900K at stock on many games.

Yeah, from what I saw too. If by 5900k you mean 5900x

Fuck, even the 5600x was beating it (and also the 5900x in some games that aren't core-bound)

[Ange]

Somehow I doubt this test was done in a mythic raid or Boralus.

With a standard 300second benchmark in a controlled environment I get this numbers in WoW:

Boralus is useless for testing, way to many variances and FPS goes up to 500 on empty servers.
Keep in mind, I just use a 2070-Super for those FPS, my lows-/avrg is very close together. All I care about is to rise my LOW-FPS, just using a better GPU would rise my MAX-fps and my AVRG-FPS while my LOW's wouldnt even change in the positive direction.

The abosolute numbers are therefor meaningless in my test, its about the changes with CPU and RAM OC.

[mrgreenthump]

Boralus is useless for testing, way to many variances and FPS goes up to 500 on empty servers.
Keep in mind, I just use a 2070-Super for those FPS, my lows-/avrg is very close together. All I care about is to rise my LOW-FPS, just using a better GPU would rise my MAX-fps and my AVRG-FPS while my LOW's wouldnt even change in the positive direction.

The abosolute numbers are therefor meaningless in my test, its about the changes with CPU and RAM OC.

Ye well, I was just pointing out that most people only care about if the minimums are good. And general testing in a controlled enviroment wont tell you that. Even my 1700 runs at 200 fps+, but going to a raid or a town it's shit show. I mean it's playable, but my monitor doesn't have LFC, so it's not fun.

[Thunderball]

Somewhere close to FS2020 numbers in comparison between AMD and Intel could be expected: at STOCK

MSFS2020 is a very bad idea for a benchmark.

1) It's horribly unoptimized, same as most other sims.
2) You get big swings of performance with every patch, both up and down. Which is expected because the game is new.

[C20]

MSFS2020 is a very bad idea for a benchmark.

1) It's horribly unoptimized, same as most other sims.
2) You get big swings of performance with every patch, both up and down. Which is expected because the game is new.

This is true. The FPS in areas with buildings absolutely tanks. Even on a 3900X and RTX3080 (with 64GB of RAM, admittedly at 2666 - intending to upgrade to 128GB of 3600+ soon), if I try to land at Heathrow, the FPS at max settings with clear weather settings drops down to 15, at 3440x1440. If I drop settings to absolute minimum, that still only gains me 10 FPS. If I run it at 2560x1440, the FPS is only 2-3 frame higher. Add cloudy weather, rain, snow, or any other impressive weather effect, and I'm sure the performance would drop down to around 9 FPS.

To put perspective on that, FPS while flying at 7000+ ft altitude over land with clear weather manages to get up to about 45 FPS at max settings. That's still awful considering the horsepower that the system has available, but it's playable. It's a great experience from gameplay standpoint (one of the most relaxing games I've ever played in an entirely good way), but an awful one from performance perspective. It's the 2020 version of Crysis.

To compare, my older GTX1070 had significantly worse performance, but that's to be expected considering it's 2 generations old at this point, and built for the era when BF1 was a lead title.

The dev performance overlay shows it flickering between GPU limited and main thread limited, but primarily it's hogging CPU time on the main thread that's causing the performance issues. This is where the 10900K or 5900X would be superior, as any IPC increase is going to be a huge benefit, but the title itself isn't optimized enough for me to say that it should be the benchmark you rate systems against, not compared to actual benchmark tools or titles which are more evenly spread between CPU and GPU, such as RDR2. Maybe they'll optimize the title at some point, and we'll see FPS reaching 50+ in heavy areas, but I feel that we won't see those benefits for quite some time as they seem to be focused on bugfixing the initial release, and adding real world mapping data for more cities.

[Ange]

Even on a 3900X and RTX3080 (with 64GB of RAM, admittedly at 2666 - intending to upgrade to 128GB of 3600+ soon), i

Keep in mind, intel CPUs do above 25-30% more fps in FS2020 as AMD ZEN2. Its one of the worst outlier games for AMD in low and avg FPS.

I use 1080p ULTRA and getting 45-60fps (NY flyby included) with 5.2GHz CPU OC and with 4400MHz subtiming optimized memory. GPUs dont really do anything for FS2020 to get better FPS.

4k in 60fps is far far away in FS2020 for a very long time.

This is as good as it gets right now with stock CPUs.

[kail]

From what I've gathered via popular tech Youtubers (GN, Jayz, Pauls, etc), the Zen 3 lineup is killer. I won't get into specifics but the $300 CPU from AMD (5600k) is matching or beating the latest i9 from intel even with the i9 overclocked in a handful of demanding games.

One of the few outliers is RDR2 where Intel CPUs still hold top performance.

[gaaara]

Yea kind of keen to see how those single threaded improvements play in WoW as well. All those threads won't make a difference, but the overall improvements to the CPU itself would see a huge benefit for those on older AMD CPU's, but for those coming from intel side...that..might be the real question.

I personally will be upgrading my 8700k to 5900x, just waiting on a few more reviews and memory reviews to pick that path. Recently got my 3080 ASUS TUF OC, and this upgrade should see a decent benefit.

[Thunderball]

From what I've gathered via popular tech Youtubers (GN, Jayz, Pauls, etc), the Zen 3 lineup is killer. I won't get into specifics but the $300 CPU from AMD (5600k) is matching or beating the latest i9 from intel even with the i9 overclocked in a handful of demanding games.

One of the few outliers is RDR2 where Intel CPUs still hold top performance.

1) 10900K was already a CPU not worth buying. You could match 10900K performance (or get very close) with a simple overclock on a 10600K in most games.
2) It's going to be not worth for a lot of people. 5600X is fast, but it's going to be up to 2 times more expensive than 3600, and 3600 is already plenty for a lot of people. Getting $100-150 extra to invest into a GPU is going to be more important most of the time.
3) 5950X/5900X is going to take the crown as the ultimate CPU, but that's a very niche market already.

[Biernot]

Benchmarks are out, pick your poison. To answer your question:

1) WoW doesnt scale with so many threads.
2) There's not going to be any difference because it's going to be GPU/draw calls bound.

To summ up Zen 3.

- First AMD chip since Athlon 64 to beat Intel stuff in average gaming FPS.
- AMD now leads in all types of production workloads, including the ones that are IPC/frequency dependant.
- Maximum frequency gains are unexpectedly good: 4.6-4.7 GHz all core is possible now, boosting up to 5 GHz on high quality silicon.
- 20% IPC increase was a hoax as expected. There is 20-25% performance uplift, but most of it is due to frequency. IPC increase is between 7-12%.
- Intel still leads in minimum FPS. This is due to inherent memory architecture drawbacks, which are very unlikely to be solved ever.
- 5600X might not be worth it considering the price, especially if 3600 actually starts costing meaningfully lower than 3600X again.

All these points are ok, except the IPC increase. I don't remember who did it, but they compared a 3000 series cpu with the equivalent 5000 series (same core count) both locked at 4GHz and got pretty much exactly the same numbers of IPC increase as in the AMD demo slides: depending on the application between 9% and 19% increased performance.

[Afrospinach]

I am kind of interested to see the 8 core amd vs the 16 core. The reason being one is a monolithic core and the other uses 2 chiplets, the latter having some disadvantages in gaming previously. Maybe something to look out for.

8 cores is still a lot of cores