RX 6xxx(XT) Thread, Lieferbarkeit, Kompatibilität, Leistung

Zitat von kaneske

Ergo: liegt am BlackMagic RAW Speed Test, ist nix Halbes und nix Ganzes.

So doof ist der nicht. Ist natürlich interessant, welche Blackmagic RAW Codecs in welcher Geschwindigkeit, also wieviel Streams maximal wiedergegeben werden können. Zum Beispiel in DaVinci Resolve. Und somit verständlicherweise nicht nur Grafikleistung abhängig, sondern auch Durchsatz des Systems gemessen wird.
Und das erklärt dann auch, wieso verschiedene Versionen auch unterschiedliche Werte anzeigen. Denn diese sind auch an die jeweilige Technologie von Resolve gekoppelt. Gerade das Update 7.3.1 verspricht zum Beispiel auf M1 einen stattlichen Zuwachs um Faktor drei. Was immer dieser Marketing-Sprech konkret bedeutet, zeigt es doch, das mit Optimierungen eben mit gleicher Haedware deutlich mehr rauszuholen ist.

fabiosun

I know that benchmarks don't actually measure real world performance, but although the results may vary from run to run, even on the same system, we're only talking about slight differences. I asked for Geekbench results to check if my own figures differ significantly here too. Since your machine has far more PCIe lanes and uses PCIe4 thanks to the Threadripper CPU, it makes total sence that your results are a lot higher (what's interesting though is, that my Metal score is almost the same as yours).

On the other hand Aluveitie system should at least benefit from the PCIe4 capabilities of his processor but still his RawSpeedTest results almost correspond to mine.

Therefore I think we can neglect the Geekbench results, since the vary too much from run to run (but at least almost match at their peak). But RawSpeedTest seems to be much more reliable. Where I get about 104fps every time i try, toasta gets between 170 and 180 everytime he tries, although our rigs are only one generation apart, with mine beeing the newer one.

So what does his system have, that mine and Aluveities systems don't?

Zitat von TNa681

DerTschnig bei mir leider sehr ähnlich, liegt es eventuell am SMBIOS?

Edit: laut Luxmark und Geekbench müsste alles i.O. sein...

Alles anzeigen

Ja TNa681

Werte sind bei mir auch sooooo..

Was aber komisch ist..... bei mir ist die CPU 11900 K auch langsamer als mein 9900K......

hmmm

TheWachowski my was not a criticism to the choice of that kind of benchmark.

Only in the past it was proven it does not reflect well "power and speed" real life and it was possible also to cheat to have better performance, I do not remember but I think downloadfritz explain here or in other forum as data hub could interfere in this benchmark

About FPS difference, it is weird because I thought a 3950x (but also your CPU @5Ghz) could do better with that NVME (I use a sabrent rockets 4.0)

I have also better result with s sabrent rocket 3.0 (better than Aluveitie speed) but I do not know if he uses PBO and in that condition... in my case it is only enabled on bios

I don’t use PBO, and my RX 6900 XT runs with a slightly lower frequency limit with an undervolt so it is about 2-3% slower than stock, but uses considerably less power.

may I ask how do you check low frequency and voltage on AMD?

fabiosun I used MSI Afterburner to find suitable values and the. Set up a soft power play table with those values.

thank you kaneske and Aluveitie

Very interesting thread

I am concentrating on how to read temperature and utilisation device on OSX with 6900Xt

Found some way with Fakesmc and radeon monitor kext, but this way misses many important data as volt or RPM fan

As I understood MacOS does not report that of newer cards. I think CMMChris knows more about that.

Correct, Apple dropped native sensor readout starting from Vega20 because their own MPX module sensors deliver data via the SMC. Third party support has zero priority for Apple. Someone would need to develop a VirtualSMC or FakeSMC plugin to get readout of temps, fan speed and clock from non Apple GPUs again.

It is possible to read some data from 6900 Xt in OSX

Fans rpm is impossible for reson said above from Chris

and this is with sensei app:

fabiosun From where did you get the radeon monitor kext? Is it the regular FakeSMC_GPUSensors.kext from the FakeSMC package?

here you go hackmac004

https://github.com/CloverHackyColor/FakeSMC3_with_plugins

fabiosun I havn't test it. Is it safe to change VirtualSMC with FakeSMC?

Zitat von karacho

Is it safe to change VirtualSMC with FakeSMC?

I'm using it too. For me it works better with iStats, because I use a SMBIOS which is not supported by VirtualSMC. Bjango - developer of iStats - told me, that it supports FakeSMC natively.

Zitat von karacho

fabiosun I havn't test it. Is it safe to change VirtualSMC with FakeSMC?

On my rig I have no problems with fakesmc

I used it daily with latest Monterey beta 6 and Big Sur 11.6 release

To explain better (I hope)

with:

FakeSMC.kext

RadeonMonitor.kext

I have exposed:

1)Temperature*

2)Memory utilisation

3)Device utilisation

*In my Knowledge sensor exposed is equivalent to GPU Z GPU Temperature hot spot and not to GPU Temperature (usually lower than hot spot)

Thank you for your response Guys. I will test it too.