+- hashcat Forum (https://hashcat.net/forum)
+-- Forum: Misc (https://hashcat.net/forum/forum-15.html)
+--- Forum: Hardware (https://hashcat.net/forum/forum-13.html)
+--- Thread: Thoughts on RTX 5090 (/thread-12395.html)
Pages:
1
2
Thoughts on RTX 5090 - TrowTrow - 12-17-2024
Hi guyz,
As you all know, Nvidia is going to announce RTX 5090 in the coming weeks.
What are your thoughts on hashcat performance on the new flagship ?
As RTX 3090 to RTX 4090 was like a bit more twice more performant (on classic hashes) can we expect the RTX 5090 to be like twice more performant as an RTX 4090 ?
RE: Thoughts on RTX 5090 - penguinkeeper - 12-17-2024
All 5090 specs are only rumours at this point, there's nothing confirmed as of yet
RE: Thoughts on RTX 5090 - monyanus - 12-17-2024
I highly doubt its 2X.
17% more shaders (cores) and about 33% more memory with better memory speed
https://nanoreview.net/en/gpu-compare/geforce-rtx-5090-vs-geforce-rtx-4090
So what to expect? For most hashes cores are limiting, so expect 25%-33% improvement, for memory hard hashing algorithms expect the most improvement. One extra benefit is that with the extra memory, larger rule sets such as when you combine rules, will less easily result in out of memory errors, something I still frequently run into.
RE: Thoughts on RTX 5090 - TrowTrow - 01-07-2025
Nvidia just announced the RTX 5090 specs, considered the new architecture it's quite hard to estimate performances.
When do you think we will have the first hashcat benchmarks on this one ?
RE: Thoughts on RTX 5090 - Chick3nman - 01-08-2025
Given what we've seen, I'd guess we see 1.5-1.75x depending on the algorithm and how everything behaves. It's tough to estimate still unfortunately. Benchmarks will be out as soon as I (or someone else) gets ahold of one of them.
RE: Thoughts on RTX 5090 - TrowTrow - 01-08-2025
(01-08-2025, 01:18 AM)Chick3nman Wrote: Given what we've seen, I'd guess we see 1.5-1.75x depending on the algorithm and how everything behaves. It's tough to estimate still unfortunately. Benchmarks will be out as soon as I (or someone else) gets ahold of one of them.
Which version are you planning to get?
RE: Thoughts on RTX 5090 - wgmmmx - 01-23-2025
(01-08-2025, 01:18 AM)Chick3nman Wrote: Given what we've seen, I'd guess we see 1.5-1.75x depending on the algorithm and how everything behaves. It's tough to estimate still unfortunately. Benchmarks will be out as soon as I (or someone else) gets ahold of one of them.
Great I want too see 5090 results then will wait for your post, thanks
RE: Thoughts on RTX 5090 - ZerBea - 01-25-2025
Some hashcat 6.2.4 benchmarks (SHA1, SHA512, TrueCrypt RIPEMD160 +XTS):
https://www.phoronix.com/review/nvidia-geforce-rtx5090-linux/6
RE: Thoughts on RTX 5090 - afd - 01-26-2025
(01-25-2025, 02:29 PM)ZerBea Wrote: Some hashcat 6.2.4 benchmarks (SHA1, SHA512, TrueCrypt RIPEMD160 +XTS):
https://www.phoronix.com/review/nvidia-geforce-rtx5090-linux/6
SHA 1 (-m 100)
-------------------------------
RTX 4090 ~ 50.4 Ghash/s
RTX 5090 ~ 68.9 Ghash/s
+ 36.7 %
-------------------------------
SHA512 (-m 1700)
-------------------------------
RTX 4090 ~ 6.3 Ghash/s
RTX 5090 ~ 8.9 Ghash/s
+ 41.3 %
-------------------------------
TrueCrypt RIPEMD160+XTS (-m 29311)
-------------------------------
RTX 4090 ~ 1.9 Mhash/s
RTX 5090 ~ 2.8 Mhash/s
+ 47.3 %
-------------------------------
RE: Thoughts on RTX 5090 - Chick3nman - 01-28-2025
I would caution against using these benchmarks without information about the way hashcat was setup for Phoronix. We saw this with the 4090 and we are seeing it again with the 5090. Failure to add the device to the aliases in the tuning database can lead to inaccurate benchmark numbers for quite a few of the modes. For the 4090, this difference was easily spotted on fast algorithms like NTLM, where the out of box speed benchmarks at ~250GH/s but adding the device to the tuning aliases shows ~285GH/s, which is a ~15% difference. I assume Phoronix is using an older version of hashcat (6.2.4) to keep consistency across older cards they have previously benchmarked but that will cause issues like this with benchmarks moving forward.