Big Navi to be 80CUs but with lower clocks than Navi 10 according to Chiphell leaker

[Remij 4,670]

Basically this was posted on ResetERA, but it comes from a Chiphell user.  Boards are apparently in testing so there should be an announcement soon if everything goes well.  We pretty much already knew this would be the case.. but the certainty is piling up.

 

https://www.chiphell.com/thread-2181303-3-1.html

 

They should finally be able to edge out the 2080ti.  

 

Just in time for Ampere to razzle-dazzle them  

[Sergio Perez SP11 173]

80 CU @ 1500 MHz would be 15.4 TF

 

that should blow away a 2080 Ti

[Remij 4,670]

10 minutes ago, Carlos Vela said:

80 CU @ 1500 MHz would be 15.4 TF

 

that should blow away a 2080 Ti

Not necessarily.  The 2080ti is 13.4TF.. but AMD's RDNA TFs still aren't quite as performant as Nvidia's.

 

The 5700XT = 9.754 TF

The 2070 Super = 9.062 TF

 

The 2070 Super is ~10% faster on average. 

 

Based on your TF estimate.. if it IS that (it could be faster yet) then it should be a bit faster than the 2080ti... nothing I would call "blowing it away"

 

There WAS a supposed leak of "big navi" beating the 2080ti in a VR benchmark by 17% though... so that would fall in line with what I would expect the general performance differential to be.  Seems like it might be legit if all this is indeed true.

 

 

Still that will be a great GPU and will force Nvidia to compete.

Edited January 19, 2020 by Remij_

[Sergio Perez SP11 173]

80 CU x 64 ROPS x 2 (operations per cycle) = 10,240 (but we still haven’t added clock speed to TF calculation)

 

10,240 x MHz would give you the teraflop for this beat of a GPU

 

10,240 x 1500 MHz  = 15.36 TF

10240 x 1550 MHz = 15.88 TF

10,240 x 1600 MHz  = 16.38 TF

10240 x 1650 MHz = 16.9 TF

10,240 x 1700 MHz = 17.4 TF
 

 

thia thing has potential to blow away a 2080 Ti

Edited January 19, 2020 by Carlos Vela

[Sergio Perez SP11 173]

I wouldn’t expect more than 1650 MHz for something so big tho 

[Remij 4,670]

37 minutes ago, Carlos Vela said:

80 CU x 64 ROPS x 2 (operations per cycle) = 10,240 (but we still haven’t added clock speed to TF calculation)

 

10,240 x MHz would give you the teraflop for this beat of a GPU

 

10,240 x 1500 MHz  = 15.36 TF

10240 x 1550 MHz = 15.88 TF

10,240 x 1600 MHz  = 16.38 TF

10240 x 1650 MHz = 16.9 TF

10,240 x 1700 MHz = 17.4 TF
 

 

thia thing has potential to blow away a 2080 Ti

The potential is there.. but at lower frequencies, it's not blowing anything away.  15-20% faster is my guess.

 

But hey... if they hit high clocks.. it'll do what we need it to do.  

[lynux3 2,049]

Novideo about to get bodied. 

[Remij 4,670]

Just now, lynux3 said:

Novideo about to get bodied. 

[lynux3 2,049]

1 hour ago, Remij_ said:

[DynamiteCop 2,085]

This is exactly what I told Dark AMD would do and he was calling me an idiot 

[jehurey 3,227]

Wait a minute, is Big Navi just a big ass chip consisting of alot more compute units?

 

Was there a video by some AMD fanboy saying that is how they were going to win the GPU wars against nVIdia because they were just going to stack more and more CUs as die shrinks continued?

 

That video was like from 2 years ago.

[Sergio Perez SP11 173]

16 minutes ago, jehurey said:

Wait a minute, is Big Navi just a big ass chip consisting of alot more compute units?

 

Was there a video by some AMD fanboy saying that is how they were going to win the GPU wars against nVIdia because they were just going to stack more and more CUs as die shrinks continued?

 

That video was like from 2 years ago.

Even a 80 CU Navi is likely smaller than 2080 Ti Die Size 

[The Mother Fucker 26]

RX 5800 > RTX 2080 Super

RX 5900XT > RTX 2080 Ti

RX 5950XT > RTX Titan

 

 

I'll be happy for AMD if that happens.  

also. The RTX ray tracing will have to deliver.   Currenly only the RTX 2080Ti can deliver straight 60fps with ray-tracing.    So once again the onus is on AMD.   They will have to beat the 2080Ti's ray tracing frame rate, not match it.    

Edited January 19, 2020 by The Mother Fucker

[Remij 4,670]

2 hours ago, Carlos Vela said:

Even a 80 CU Navi is likely smaller than 2080 Ti Die Size 

A lot smaller.

[Remij 4,670]

2 hours ago, DynamiteCop! said:

This is exactly what I told Dark AMD would do and he was calling me an idiot 

It's not going to be 2x the performance of the 5700XT though.  So as long as you didn't say that.. you should be fine. 

[The Mother Fucker 26]

Just now, Remij_ said:

It's not going to be 2x the performance of the 5700XT though.  So as long as you didn't say that.. you should be fine. 

Can you imagine.  Twice the performance of 5700XT.   

 

NVIDIA would be in the same boat as INTEL.  SOL

 

 

 

[jehurey 3,227]

3 minutes ago, The Mother Fucker said:

Can you imagine.  Twice the performance of 5700XT.   

 

NVIDIA would be in the same boat as INTEL.  SOL

 

 

 

According to that graph, the GTX 970, a video card that came out around August/Sept 2014, the 5700XT has only managed to offer double the power of THAT card, 5 years later.

 

And you think AMD, within one year of releasing the 5700 XT, is going to double that power of that?

[DynamiteCop 2,085]

4 minutes ago, jehurey said:

According to that graph, the GTX 970, a video card that came out around August/Sept 2014, the 5700XT has only managed to offer double the power of THAT card, 5 years later.

 

And you think AMD, within one year of releasing the 5700 XT, is going to double that power of that?

You say managed to as if that means that's all they're capable of, the 5700 line is just a mid-range offering. 

[jehurey 3,227]

1 minute ago, DynamiteCop! said:

You say managed to as if that means that's all they're capable of, the 5700 line is just a mid-range offering. 

the highest range card, an nVidia 2080 TI, has only managed to perform 2.5x better than my GTX 970 from 5.5 years ago.

[Remij 4,670]

Just now, jehurey said:

the highest range card, an nVidia 2080 TI, has only managed to perform 2.5x better than my GTX 970 from 5.5 years ago.

In that single benchmark...