QUOTE(cawan @ Mar 17 2022, 05:02 PM)
Just got RTX3060 at 2498. 

3050 at that price is just daylight robbery.
NVIDIA GeForce Community V19, RTX 5000 unveiled
|
|
Mar 17 2022, 08:49 PM
Return to original view | IPv6 | Post
#1
|
|
Elite
15,694 posts Joined: Mar 2008 |
|
|
|
|
|
|
Mar 17 2022, 09:13 PM
Return to original view | IPv6 | Post
#2
|
|
Elite
15,694 posts Joined: Mar 2008 |
|
|
|
Mar 18 2022, 10:05 AM
Return to original view | Post
#3
|
|
Elite
15,694 posts Joined: Mar 2008 |
QUOTE(TristanX @ Mar 17 2022, 10:03 PM) Let's hope Intel is smart enough to try gaining market share instead profiteering. That will crash the prices. Not so soon, their driver is still in infancy. QUOTE(gacktleong @ Mar 18 2022, 09:27 AM) The GDDR6X is a real heating wire, can easily consume 100W. |
|
|
Jun 5 2025, 09:24 PM
Return to original view | IPv6 | Post
#4
|
|
Elite
15,694 posts Joined: Mar 2008 |
QUOTE(PJng @ Jun 5 2025, 09:19 PM) NVIDIA reportedly preparing GeForce RTX 5050 desktop GPU for July release 5060/Ti 8GB will have hard time with 9060XT 8GB/16GB.https://videocardz.com/newz/nvidia-reported...or-july-release don like 5060? nvidia give more option 5050 5050? 199 USD please. This post has been edited by kingkingyyk: Jun 5 2025, 09:25 PM |
|
|
Jun 6 2025, 08:31 AM
Return to original view | Post
#5
|
|
Elite
15,694 posts Joined: Mar 2008 |
|
|
|
Jun 6 2025, 01:49 PM
Return to original view | Post
#6
|
|
Elite
15,694 posts Joined: Mar 2008 |
QUOTE(babylon52281 @ Jun 6 2025, 09:21 AM) A 96bit mem bus would kill all reason to have a bigger VRAM. Its similar back in the days where 64bit GPU has huge 512MB VRAM but still loses to 128bit GPU with smaller 256MB but faster VRAM. That is not a general rule. AMD's design has less dependent on RAM bandwidth and you have already seen it the current 9060/9070. It is achieved through better precomputation pruning and compression.--- If you shift to the LLM scene, larger memory still has the advantage of loading entire network layers into the GPU. Lower bandwidth only causes the inference to run slower, but still way faster than having some of the layers handled by CPU. Well in this scenario, 5060 Ti 16GB is actually quite a lot faster than 9060 XT 16GB with it's GDDR7. Games are just simply having more well-known, opportunistic & human perception-based optimizations than the brute forcing in AI. This post has been edited by kingkingyyk: Jun 6 2025, 01:52 PM |
|
|
|
|
|
Jun 6 2025, 02:14 PM
Return to original view | Post
#7
|
|
Elite
15,694 posts Joined: Mar 2008 |
QUOTE(babylon52281 @ Jun 6 2025, 02:03 PM) Actually its the opposite where Radeons are more affected by mem architecture. Heres what TPU Wiz has to say: "Historically, it has also been the case that AMD's memory management isn't as good as NVIDIA's, which probably means that the 8 GB 9060 XT will have a harder time dealing with memory shortages than NVIDIA's card." Nope, it is not about 8GB or 16GB. It is relative comparison of 9060XT vs 5060 Ti. QUOTE(babylon52281 @ Jun 6 2025, 02:07 PM) And as for using consumer grade GPU for AI... I would say that I will condemn such practices as much as using GPU for mining. Both incidents which had shortages in gaming GPU and hiking skyhigh prices. I curse those crypto bros and now AI bros. Consumer card is not for AI bros. They opt for used workstation cards instead for way larger amount of memory. This post has been edited by kingkingyyk: Jun 6 2025, 02:18 PM sai86 liked this post
|
|
|
Jun 6 2025, 09:13 PM
Return to original view | IPv6 | Post
#8
|
|
Elite
15,694 posts Joined: Mar 2008 |
QUOTE(babylon52281 @ Jun 6 2025, 06:01 PM) Its what Wiz is talking about, between 8GB 9060XT and 5060TI, the memory limitations (VRAM size, bus width, PCIE speed) will hit that Radeon harder compared to Geforce. You can only strangle so much the bandwith just to cram in as much VRAM that it becomes detrimental instead. Bus width shouldnt g below 128bit (much like cutting PCIE lanes too). 12GB will effectively offset that data exchange issue. 4090 and 5090 are consumer cards yet go figure why are they constantly in shortage and where they actually end up. Even 3090TI in 2nd hand market are getting hot for that particular reason. I hate jerkwads that buys up these GPUS for other purpose than gaming. This post has been edited by kingkingyyk: Jun 6 2025, 09:14 PM |
|
|
Jun 18 2025, 10:48 AM
Return to original view | Post
#9
|
|
Elite
15,694 posts Joined: Mar 2008 |
QUOTE(adamtayy @ Jun 17 2025, 05:58 AM) [attachmentid=11514861] 4090 is still the way to go. https://www.slashgear.com/1878188/nvidia-rt...d-which-better/ Rtx 4090 vs Rtx 5080 QUOTE(chocobo7779 @ Jun 17 2025, 08:04 PM) IINM TristanX found an Intel B580 at RM1.1K - which pretty much invalidates this card considering the larger VRAM pool 1.1K is with Shopee discount, original price was 1269, but still a nice price nevertheless.This post has been edited by kingkingyyk: Jun 18 2025, 10:50 AM |
|
|
Jun 18 2025, 04:08 PM
Return to original view | Post
#10
|
|
Elite
15,694 posts Joined: Mar 2008 |
QUOTE(babylon52281 @ Jun 18 2025, 01:16 PM) Even prediscount its still a very good price dy since 5060 is starting at RM 1500. Only problem is its not suitable for older systems thats on Gen3, no Rebar and with CPU overhead. 5060 will still be bottlenecked by PCI-E 3.0 due to tons of data swapping between RAM/VRAM happening. chocobo7779 liked this post
|
| Change to: | 0.0460sec
0.27
7 queries
GZIP Disabled
Time is now: 2nd December 2025 - 09:55 PM |