News Nvidia RTX 3060 Ti Gets Few Benefits from GDDR6X, Says Reviewer

[Admin]

Polish news outlet reviews the GeForce RTX 3060 Ti GDDR6X and compares it to the regular version.

Nvidia RTX 3060 Ti Gets Few Benefits from GDDR6X, Says Reviewer : Read more

 

[OneMoreUser]

3060 still one of the best???

It is what, like the ten fastest or something like that so how does it count as "one of the best".

I feel like such statements reflect the quality of reviews here, clearly they are some of the best.

 

[pclaughton]

3060 still one of the best???

It is what, like the ten fastest or something like that so how does it count as "one of the best".

I feel like such statements reflect the quality of reviews here, clearly they are some of the best.

Best != Fastest. Speed is obviously the biggest ingredient there, but there are plenty of other factors.

 

[OneMoreUser]

Best != Fastest. Speed is obviously the biggest ingredient there, but there are plenty of other factors.

Well, yes - however what would that be?
It is not like it offers the best performance per $ since AMD has that covered*, it doesn't offer anything special with regards to number of monitors, noise level or looks so what then?

* I'm ignoring RT here, since really RT on anything than something close to a top end card is really to slow to matter.

 

[The Historical Fidelity]

That’s about what I expected for the gddr6x swap. Nvidia would need to be really bad at designing gpu’s if they purposefully bottlenecked the compute cores with insufficient memory bandwidth. That 3% increase in performance compared to the regular 3060 Ti shows that (in rough terms) the compute cores were not fully saturated by the gddr6 maybe 5-10% of the time (by that I mean the compute cores could only work at like ~95% max performance for 10% of the time as bandwidth needs are not static when the load type is dynamic) , and now that the compute cores have 35% more bandwidth they are fully saturated 100% of the time. Makes you think, could nvidia have truly made the compute cores the rate limiting step by simply using 16 Gbps gddr6 instead of cheaping out on 14 Gbps?
Btw please don’t look too hard into my napkin math, it’s been a 14 hour day and I’m brain dead lol.

 

[watzupken]

The results should not come as a surprise. Previously we saw this change with the RTX 3070 to 3070 Ti. Even with a mild CUDA core bump, the resulting performance improvement is less than 10% at best, if I am not mistaken. Given that the RTX 3060 Ti is targeted for a max of 1440p, or, low RT performance, the extra bandwidth is not going to make a huge difference. If the upgrade does not result in an increase in price, then its fine. However, I've seen higher prices for the GDDR6X over the GDDR6 version, which makes the upgrade look silly.

 

[digitalgriffin]

The results should not come as a surprise. Previously we saw this change with the RTX 3070 to 3070 Ti. Even with a mild CUDA core bump, the resulting performance improvement is less than 10% at best, if I am not mistaken. Given that the RTX 3060 Ti is targeted for a max of 1440p, or, low RT performance, the extra bandwidth is not going to make a huge difference. If the upgrade does not result in an increase in price, then its fine. However, I've seen higher prices for the GDDR6X over the GDDR6 version, which makes the upgrade look silly.

Higher speed ram usually has higher latency and strobe line refreshes. That's why we have CAS numbers on RAM. This is why CAS numbers go up as speed does typically.

So higher latency + higher bandwidth = wash performance wise.

 

[salgado18]

That 6.6% higher boost clock of the GDDR6 model nearly invalidates the test, in my opinion.

 

[russell_john]

Polish news outlet reviews the GeForce RTX 3060 Ti GDDR6X and compares it to the regular version.

Nvidia RTX 3060 Ti Gets Few Benefits from GDDR6X, Says Reviewer : Read more

The biggest problem with this test is he didn't mention the difference in memory clock speeds ..... I've never even heard of KFA2 and looking at their website they don't even list the memory speed in their specs .... So if they didn't actually tune the memory speeds then the card isn't going to gain a damned thing going to GDDR6X

I'll wait to see what one of the more respected manufacturers puts out before I'll judge it ..... I don't really expect much because the 3060 just can't make use of the extra speed

 

[KyaraM]

The biggest problem with this test is he didn't mention the difference in memory clock speeds ..... I've never even heard of KFA2 and looking at their website they don't even list the memory speed in their specs .... So if they didn't actually tune the memory speeds then the card isn't going to gain a damned thing going to GDDR6X

I'll wait to see what one of the more respected manufacturers puts out before I'll judge it ..... I don't really expect much because the 3060 just can't make use of the extra speed

KFA2 is a pretty common brand in Europe and Asia, the farther east you go the more widespread they get. Just like Palit, for that matter. Excellent cards, but Americans are apprehensive because they often don't know the name, while in Europe their rep is pretty good, and for good reasons. In fact if I remember correctly, KFA2 is a sub-brand of Palit just like GALAX and Gainward is nowadays. Just because they aren't as well-known in the US does not mean they are bad. Palit is the biggest manufacturer of Nvidia cards in the world.

Edit:
Techpowerup gives the memory clock as 1188MHz:
https://www.techpowerup.com/gpu-specs/kfa2-rtx-3060-ti-1-click-oc-plus-gddr6x.b10257

(Edited for spelling mistakes)

 

[Friesiansam]

As there is no explanation of what the figures represent, the table, "GeForce RTX 3060 Ti GDDR6X Benchmarks" means nothing.