I really feel something is wrong with rtx3080..

[watzup_ken]

Yes, but it creates alot of hype....

Normally, we just a small-moderate increase in cuda cores when moving from 1 generation to the next. Sometimes, we even see a reduction.

Eg. 2080FE is a tiny bit faster than 1080ti. 2080FR has 2944 cuda cores, 1080ti 3584. A bit less power too (250 vs 225W). So, less cuda cores but more efficient, so it can perform just as well and less power...

1080FE vs 980ti is quite similar....1080FE has 2560 @ 180W, 980TI 2816 @ 250W...Same, less cuda cores, less power but slightly faster...

Now, 3080 suddenly has 8704 cuda cores vs 4352 for 2080ti. [/B]DOUBLE!![/B]. Then over 300W TDP!! So whats your impression when you see this? Then nvidia claims up to 1.9x improvement in performance/watt!! WOW!

We never seen such a big jump in cuda cores. I don't think we seen new architecture being less efficient as well. Then Nvidia also said 3070 is faster than 2080ti! So, do you think this adds up to the hype?

1.9x performance is the best case scenario over the RTX 2080. I think during the announcement, Anandtech did highlight that the claim sounds dubious. At best, its closer to 50% efficient.

RTX 3070 may be faster than the RTX 2080 Ti in certain scenarios. I was pretty sure this is the case from the get go. The specs sounds great, but performance don't scale with the number of CUDA cores. I feel in most cases, it should be may be around 80% of what the GTX 2080 Ti capable of. Only in those highly optimized Nvidia titles like Control, Doom Eternal, etc, will you most likely see it out performing. Basically those titles that they showcase during the announcements are the ones that we may see the RTX 3070 do as good if not slightly better than the 2080 Ti is my guess.

 

[hg2]

Watch a few benchmarks videos, the only conclusion it's those games not fully optimised for RTX 3000 and making full use of its RTX accelerations capabilities at 4k...

My conclusion, U don't "REALLY" need RTX 3080 card if you:
(1) don't game.
(2) don't play RTX ON titles.
(3) don't have 4k or above monitors.
(4) Already own a RTX 2080

But if you have the spare cash to splash and want the lastest and greatest...why not?

 

[Swordsman]

i want a massive jump in improve from 1070..

how many % i wondered.

 

[Hailgod]

You shouldnt compare 2080ti to 3080. Then why not compare GTX 980ti to 1080? The performance increase is even lesser during that generation but i dont see anyone complaining.

why cherrypick the only graph that showed minor increase? techpowerup full list shows 35%+ improvement on average.

Fyi. The cuda cores at every gen is different. You cant count core count vs core count of a different gen. They are different to begin with. Doesnt mean the 3080 is 3x more cores than 2080Ti it should be faster by 3x. Doesnt work that way or rather it doesnt scale that way.

And fyi again. 3080 is not suppose to compare directly to the 2080Ti. But more towards the 2080.

This is or was at roughly the same price at launch.

getting tricked? 700$ is a 80ti price. 2xxx is special case.

1.9x performance is the best case scenario over the RTX 2080. I think during the announcement, Anandtech did highlight that the claim sounds dubious. At best, its closer to 50% efficient.

RTX 3070 may be faster than the RTX 2080 Ti in certain scenarios. I was pretty sure this is the case from the get go. The specs sounds great, but performance don't scale with the number of CUDA cores. I feel in most cases, it should be may be around 80% of what the GTX 2080 Ti capable of. Only in those highly optimized Nvidia titles like Control, Doom Eternal, etc, will you most likely see it out performing. Basically those titles that they showcase during the announcements are the ones that we may see the RTX 3070 do as good if not slightly better than the 2080 Ti is my guess.

the 1.9x number came from comparison of 130w ampere vs 240w turing. maybe u can get a 130w mobile gpu that runs like a 2080ti. maybe not.

 

[Nitrile]

3080 @ Computerbase.de: A 50W lower TDP would've meant only 4% lower performance and 25% better efficiency than 2080Ti, not 10%.

https://www.computerbase.de/2020-09/geforce-rtx-3080-test/6/

nvidia went all out, pushing the 3080 to its limits.
this is really brute force in action.

pressure from RDNA2?

 

[Clearnfc]

3080 @ Computerbase.de: A 50W lower TDP would've meant only 4% lower performance and 25% better efficiency than 2080Ti, not 10%.

https://www.computerbase.de/2020-09/geforce-rtx-3080-test/6/

nvidia went all out, pushing the 3080 to its limits.
this is really brute force in action.

pressure from RDNA2?

Wow, 50w for just 4%. They really squeeze every bit out of the card. 25% better efficiency would definitely be a step forward.

 

[Gattberserk]

I just dont understand why 3080 is just ~35% faster than 2080ti.. i was expecting something closer to 2x...this is because 3080 has 8704 cuda cores vs 4352 for 2080ti. Double... Memory also faster. Not to mention new architecture and new process.

So where is the improvement??

You are very dramatic, chill abit bro.

https://www.youtube.com/watch?v=csSmiaR3RVE&t=1638s


This video link will answer some of your question on why the 3080 seem to "under whelming".

 

[Alphas]

On the contrary, I find that the performance increase is a lot higher in 3080 vs 2080ti. Typically each generation of gpu, the performance increased by 10-20%. 30% increase is amazing considering it is a 3080, the performance would be higher with 3080ti vs 2080ti.

 

[hg2]

You are very dramatic, chill abit bro.

https://www.youtube.com/watch?v=csSmiaR3RVE&t=1638s


This video link will answer some of your question on why the 3080 seem to "under whelming".

Another video to defend the low performance bump.

Absolute hype dissappoints completely....
for gaming... maybe this card more suitable for productivity ?

 

[Clearnfc]

It's not bad I'd say considering the power consumption of the 1080ti and 2080ti AIB OCed version of those cards is guzzling about the same amount or slightly more power and with lesser performance than stock 3080.

Must remember that 2080ti is still on 12nm..... IF it goes to 8nm, I think the gap will be even smaller. Also, that 30% increase is against the FE version, not the oc ones.

 

[blurrry]

TS confirm some salty person with 2080ti LOL.

 

[watzup_ken]

3080 @ Computerbase.de: A 50W lower TDP would've meant only 4% lower performance and 25% better efficiency than 2080Ti, not 10%.

https://www.computerbase.de/2020-09/geforce-rtx-3080-test/6/

nvidia went all out, pushing the 3080 to its limits.
this is really brute force in action.

pressure from RDNA2?

I feel this may be dependent on the quality of the chip. Perhaps this is the target clockspeed, so with the low chip quality they needed to increase the TDP drastically to make sure all the 3090/ 3080 out there can meet the target clockspeed.

 

[m0n0n0ke]

TS confirm some salty person with 2080ti LOL.

Who wont dl

Nvidia slap those 2080ti user so hard this time

 

[ronnie_gogs]

3080 is compared to 2080ti as launch price of the card is cheaper than 2080ti launch and even pre ampere announcement retail prices... It gives a good performance bump and is superior to a 2080ti in every aspect dropping 2080ti aftermarket prices.

Its a beast of a card for 3d rendering due to pure increase in CUDA cores.

 

[Unic0rn]

waiting next Thursday for the 3090 unboxing and reviews! i feel the third party AIB manufactures were not given sufficient time to prepare and had to rushed out their designs (mostly familiar old designs) to capture the hype and market share for 3000 series, their custom higher end ROG cards were not even available at launch

 

[lumpyme]

by TS logic
1 pregnant woman needs 9 months to produce 1 baby
9 pregnant women should take 1 month to produce 1 baby

there might be more cores but there is still only 1 decoder to split the job among all the available cores