FINAL FANTASY XIV: Dawntrail Official Benchmark

[RootinTootinPootin]

1024x768

 

[BNSoul]

Edit: re-did benchmark with g-sync off

New monthly Windows 11 update (22631.4169), new Dawntrail benchmark run (1080p maximum settings), marginally better this time than previous W11 build (36247 vs 36596, +3 avg fps gain), same config, same GPU and chipset drivers, etc. Stock 5800X3D and also stock 4080.

 

[Det0x]

So i have finished my dualrank 64 gigs performance exploration and went back to this benchmark, i found SR 2x16gigs adie @ 6600MT/s CL26-37-32-30-62 + 2200mhz FCLK 1:1 mode to about ~400points faster than SR 2x16gigs adie @ 8000MT/s CL32-45-40-44-84 + 2200mhz FCLK 2:1 mode in this benchmark.

So the verdict is that this benchmark prefers latency, atleast for my current X3D setup


With some further tweaking i also managed to break 51k, but to my surprise, i still seem to be GPU limited even at 1080p lowest settings, on a tuned watercooled 3090
(saw nice scaling with +30mhz GFX core)
View attachment 345989

Same settings, mid oc on the 9950X ES
(2 months old run, performance could have improved with latest agesa)
View attachment 359107

Now with more mature agesa / software
(latest agesa 1.2.0.2 fixes the cross CCD latency)

Above is done on my bad 9950X (my main 9950X ES is waiting for a new delid frame).
Is that the new CPU highscore in this thread ?

 

[soulrider4ever]

1920x1080 maximum defaults

 

[BNSoul]

was curious about the post above this (Shadowbringers benchmark) so I tested with stock Zen 3 5800X3D + stock RTX 4080 + DDR4 3800 on Windows 11 Pro 24H2 and mine is just 4% slower, not bad considering the PC above is a Zen 5 9950X + DDR5 + RTX 4090 also on Win 11 24H2.

 

[Knight47]

R3 1200, 4x8GB DDR4 2666MHz, RTX 4070 Ti Super.

R7 7800X3D, 2x16GB DDR5 4800MHz, RTX 2070 Super.

 

[Paramount]

i put all setting as low as possible to know effect of CPU

1- default setting ( latest drivers / CPPC : Driver) :

2- force using CCD1 ( by Process lasso)

3- Force use CCD0 ( By process lasso )

 

[soulrider4ever]

was curious about the post above this (Shadowbringers benchmark) so I tested with stock Zen 3 5800X3D + stock RTX 4080 + DDR4 3800 on Windows 11 Pro 24H2 and mine is just 4% slower, not bad considering the PC above is a Zen 5 9950X + DDR5 + RTX 4090 also on Win 11 24H2.

View attachment 366391

Very nice indeed - I do have SMT/Hyperthreading disabled - could be why my scores are lower than others with 9950x

 

[BNSoul]

Very nice indeed - I do have SMT/Hyperthreading disabled - could be why my scores are lower than others with 9950x

With regard to the 7950X/9950X you should be getting a better score with just one CCD enabled and maybe SMT disabled on top of that (as in just 8 threads in the same CCD), also make sure you have the latest chipset drivers and the AGESA which improves Zen 5 latency.

Since some people are posting custom benchmarks I'm just going to post the settings I actually play the game with, basically 1440p maximum settings with just two tweaks ("Radial Blur" and "Depth of Field" disabled). Also, for this game I set my 5800X3D to 114-75-115 power limits and -30 all-core in the curve optimizer. In addition, SMT (simultaneous multi-threading) is disabled. Thus, performance improves by 23% compared to stock CPU limits and stock voltage curve. On the other hand, I run my RTX 4080 at stock settings with the BIOS switch on "gaming" mode.

 

[AVATARAT]

i put all setting as low as possible to know effect of CPU

1- default setting ( latest drivers / CPPC : Driver) :

View attachment 366839


2- force using CCD1 ( by Process lasso)

View attachment 366840

3- Force use CCD0 ( By process lasso )

View attachment 366841

Yes, the second CCD1 has a sort of "lag" and the result is always worse than the CCD0.
The CCD1 is faster than Intel's E-cores, but in the end it's just "cores for something that isn't important".

 

[Knight47]

R3 1200, 4x8GB DDR4 2666MHz, RTX 4070 Ti Super.

So I put the 4070Tis in my main rig, before I return it.

 

[yzonker]

Last run before the new stuff gets released. I guess we'll see if ARL can match RPL in this bench. Intel as showing it 8% slower, so maybe not. But maybe 9800x3D.

64/0/53, 8400C36

 

[Gideonic]

Have you guys seen this run by a presumed 9800X3D with an unknown GPU?

Page 843 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 843 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

(this is from the same source that leaked early 9950X scores, so the leak is legit)

Could anyone of you run this bench at the same settings with your monster RTX 4090 rigs, for comparison?

 

[Det0x]

Have you guys seen this run by a presumed 9800X3D with an unknown GPU?

Page 843 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 843 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

(this is from the same source that leaked early 9950X scores, so the leak is legit)

Could anyone of you run this bench at the same settings with your monster RTX 4090 rigs, for comparison?

@yzonker ran Raptor Lake @64/0/53, 8400C36 + 4090 with the same settings right above, i think

 

[Gideonic]

@yzonker ran Raptor Lake @64/0/53, 8400C36 + 4090 with the same settings right above, i think

While it's close, it's 1080p vs 720p (FSR vs DLSS probably doesn't matter much)

Scores over at anandtech forums (with the exact same settings) on 7800X3D + 7900 XTX rigs are only in mid 45000ies for 720p.

It doesn't seem 100% CPU limited even at 720p, so it would he nice to get a RTX 4090 score for that resolution as well.

 

[yzonker]

Have you guys seen this run by a presumed 9800X3D with an unknown GPU?

Page 843 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 843 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

(this is from the same source that leaked early 9950X scores, so the leak is legit)

Could anyone of you run this bench at the same settings with your monster RTX 4090 rigs, for comparison?

I said before the new stuff, didn't know it would be this close. I'll mentally count that when it's not redacted. LOL

Kinda look like it's fully unlocked based on the R23 run. Probably used a chiller based on the R23 temp, but I'm using one too.

While it's close, it's 1080p vs 720p (FSR vs DLSS probably doesn't matter much)

Scores over at anandtech forums (with the exact same settings) on 7800X3D + 7900 XTX rigs are only in mid 45000ies for 720p.

It doesn't seem 100% CPU limited even at 720p, so it would he nice to get a RTX 4090 score for that resolution as well.

Someone already did 51k on 7800x3d,

FINAL FANTASY XIV: Dawntrail Official Benchmark

The 7800x3d is impressive. What memory settings were you running here? My 14900kf is pretty much tapped out on this board (Apex running 8800 is gone). Nice run :toast: I were running my daily #1 profile (6600/2200) settings together with daily CO tune on the 7800X3D (my daily #2 profile is...

www.techpowerup.com

 

[Gideonic]

Someone already did 51k on 7800x3d,

FINAL FANTASY XIV: Dawntrail Official Benchmark

The 7800x3d is impressive. What memory settings were you running here? My 14900kf is pretty much tapped out on this board (Apex running 8800 is gone). Nice run :toast: I were running my daily #1 profile (6600/2200) settings together with daily CO tune on the 7800X3D (my daily #2 profile is...

www.techpowerup.com

Yeah, but both his and your score are at 1080p. The leak is 720p. And @Det0x did mention this:

With some further tweaking i also managed to break 51k, but to my surprise, i still seem to be GPU limited even at 1080p lowest settings, on a tuned watercooled 3090
(saw nice scaling with +30mhz GFX core)

So there could be additional uplifts at 720p

 

[yzonker]

Yeah, but both his and your score are at 1080p. The leak is 720p. And @Det0x did mention this:



So there could be additional uplifts at 720p

I could run again, but there was very little GPU dependence in my run on the 4090. As a test I ran it at stock clocks and OC on the 4090. Only gained 300pts.

On 3090 there certainly could be though.

 

[mouacyk]

7800x3D ECLK 102 = 5.15GHz max boost, 6400C28 GDM Off = 51,265:

Looking like 21% uplift to 9800X3D, but that leak is tweaked through the roof (with BCLK overclock?)

 

[Makaveli]

I see a 53k 7950X3D + 6900XT score over at anandtech which is faster than a 7800X3D + 4090??

 

[lexluthermiester]

at anandtech

Isn't that site dead an gone?

 

[Athlonite]

Just a couple of runs to waste some time
1080p

1440p

 

[yzonker]

Isn't that site dead an gone?

Page 847 - Discussion - Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)

Page 847 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Makaveli]

Isn't that site dead an gone?

Site is dead not the forum's which are very much alive.

I've been posting there alittle longer than I have here.

Im still curious what is up with the RDNA 2 scores is it due to larger infinity cache?

 

[mouacyk]

Pushed RTX 4090 to 3,015MHz and 25Gbps = 51,577