Install Steam
login
|
language
简体中文 (Simplified Chinese)
繁體中文 (Traditional Chinese)
日本語 (Japanese)
한국어 (Korean)
ไทย (Thai)
Български (Bulgarian)
Čeština (Czech)
Dansk (Danish)
Deutsch (German)
Español - España (Spanish - Spain)
Español - Latinoamérica (Spanish - Latin America)
Ελληνικά (Greek)
Français (French)
Italiano (Italian)
Bahasa Indonesia (Indonesian)
Magyar (Hungarian)
Nederlands (Dutch)
Norsk (Norwegian)
Polski (Polish)
Português (Portuguese - Portugal)
Português - Brasil (Portuguese - Brazil)
Română (Romanian)
Русский (Russian)
Suomi (Finnish)
Svenska (Swedish)
Türkçe (Turkish)
Tiếng Việt (Vietnamese)
Українська (Ukrainian)
Report a translation problem
You can have multi frame gen in any game with Lossless Scaling app. it offers up to 20x multi frame gen.
Wrong again they use FP8 for FSR 4 although it shows it will run on INT8 on older GPU's, Nvidia and Intel use INT8 for DLSS/XES
Matrix cores are just normal cores with a fancy name "AMD's matrix cores support a wide range of precision formats, including FP64, FP32, and lower precision formats like INT8 and FP8"
"AMD's matrix cores are part of their CDNA architecture, found in server-grade Instinct accelerators, and have been introduced in consumer-grade RDNA architectures starting with RDNA 4"
You read some stuff and thought you understood it but alas you didn't.
yes, RX 9070 has matrix cores — but they’re limited in scale, purpose, and software support compared to AMD’s true AI accelerators .So , AMD use them on RX 9000 only for FSR 4 but not for frame gen
old Radeons ,dont have Matrix cores at all
edit
in RX 9000 there re Matrix cores but they are not so powerful as in AMD AI cards . limited functionality and cant do frame fen 4 without latecy increase
AMD Matrix cores have the same features as Nvidia tensor cores.
Also keep in mind: RTX5060 "supports" MFG at the cost of losing 1/2-1/3 of the real performance xD
Not trully x3 or x4, is it?
There is no frame generation without latency increase yet.
The best solution is Lossless Scaling app with a second GPU dedicated just for framegen so there is no loss of base FPS. Even AMD, Nvidia or Intel don't offer such option of using second GPU for FG. All current FG technologies increase latency.
Nvidia does it a bit different but same result, Nvidia didn't introduce FP8 too till 4th gen Hopper.
Matrix cores are integrated into the Compute Units (CU's) as specialised execution units.
Nvidia tensor cores are dedicated and separate processing cores within a streaming multiprocessor (SM), same for RT cores.
I'm going to guess it's better the way Nvidia does it because AMD is separating RT from the CU's in the upcoming "Radiance Cores" which will handle more of the ray tracing workload directly offloading it from the CU's, but that's just for RT not up-scaling in RDNA5, maybe RDNA6 will have a separate "tensor" solution for FSR to offload the CU's, makes sense to focus on RT cores and RT performance first as FSR 4 is decent for now.