Official RTX 30x0 thread

Imho,


I think Amd did innovate as well; SAM and Infinity Cache are efficient abilities which improvers performance watt, too.

My constructive nit-pick has been their gaming ecosystem and noticed that they're working hard with FSR 1.0 and with a Global setting ability; FSR 2.0, bringing more performance for DirectX 11; hopefully bring broadcast abilities. Having the majority of gaming consoles' hardware is not enough to convince PC gamers to buy -- PC gamers/owners desire PC experiences.

Nvidia is a handful, they just don't stop -- AMD has resources and talent -- and can compete.


I believe AMD has the raw talent and may crawl out of the gutter one of these days but they haven't had the financial push they need to compete with the green giant. They stayed with slower VRAM and kept the memory controller limited. Their cache can't solve the problem with speed if the scene desires more bandwidth than the pipeline has. This is why 4k always falls on it's face even when AMD is winning at all other resolutions on some games. I want them to do better but I'm not going to sugar coat the truth about NV's dominance. They made a good series of cards last time but they still have to improve FSR and ray tracing while managing raster performance without the silicon advantage they had last time. It's an uphill battle for them.
 
qXLQi6.jpg


a little bigger

and no more SLI port :cry:
 
I believe AMD has the raw talent and may crawl out of the gutter one of these days but they haven't had the financial push they need to compete with the green giant. They stayed with slower VRAM and kept the memory controller limited. Their cache can't solve the problem with speed if the scene desires more bandwidth than the pipeline has. This is why 4k always falls on it's face even when AMD is winning at all other resolutions on some games. I want them to do better but I'm not going to sugar coat the truth about NV's dominance. They made a good series of cards last time but they still have to improve FSR and ray tracing while managing raster performance without the silicon advantage they had last time. It's an uphill battle for them.

I don't see, "falls on its face," but very competitive 4k raster performance overall from Amd.
 
I believe AMD has the raw talent and may crawl out of the gutter one of these days but they haven't had the financial push they need to compete with the green giant. They stayed with slower VRAM and kept the memory controller limited. Their cache can't solve the problem with speed if the scene desires more bandwidth than the pipeline has. This is why 4k always falls on it's face even when AMD is winning at all other resolutions on some games. I want them to do better but I'm not going to sugar coat the truth about NV's dominance. They made a good series of cards last time but they still have to improve FSR and ray tracing while managing raster performance without the silicon advantage they had last time. It's an uphill battle for them.

AMD confirms RDNA 3-based Navi 31 has 6 x MCDs + 384-bit memory bus

Read more: https://www.tweaktown.com/news/8738...i-31-has-6-mcds-384-bit-memory-bus/index.html



AMD Radeon RX 7900 XT With RDNA 3 'Navi 31' GPU May Possibly Over Over 6 TB/s of Effective Bandwidth According To Speculation

The user mentions prior information to set up their claim, such as that the AMD Radeon RX 6950 XT which is the company's top performer in its current GPU lineup, offering 16GB of GDDR6 memory that is clocked at 18 Gbps with the assistance of the cards 256-bit bus. This combination of technology results in bandwidth speeds of 576 GB/s. Bright Iron continues to discuss that the bandwidth is dwarfed by NVIDIA's GeForce RTX 3090 GPU, which offers 936 GB/s of bandwidth, and the company's RTX 3090 Ti's bandwidth of 1,008 GB/s. However, in several benchmarks, the AMD Radeon RX 6950 XT fights for close dominance with rival NVIDIA's powerful GPUs.


Bright Iron continues to explain that AMD's bandwidth performance does quite well compared to NVIDIA because of the Infinity Cache technology that the company uses. Readers will remember that the Infinity Cache is the memory separate from the main memory located on AMD's GPUs. It is usually minimal in size compared to the main memory found on the graphics card, but due to being close to the chip, the Infinity Cache can be pulled at a much higher rate.
The AMD Infinity Cache can offer as high as 128 MB memory capacity with 2 TB/s of memory bandwidth. The AMD Radeon RX 6850 XT offers a standard bandwidth of 512 GB/s until you add the 128MBs of Infinity Cache, boosting the bandwidth another 1,281 GB/s, giving users a total of 1,793 GB/s of bandwidth, which is much higher than NVIDIA can produce currently.

Newer RDNA 3 architecture will allow the Infinity Cache to reach capacities of 384 MB, which is three times higher than the previous RDNA 2 architecture. The chip can produce 192 MB through the standard Cache while utilizing another 192 MB of 3D V-Cache packaging to achieve 384 MB of the entire Infinity Cache.

http://www.rage3d.com/board/showpost.php?p=1338350044&postcount=708

i think they may well have that covered this time
:hmm:

........

and NV lost the ARM deal but AMD did get Xilinx

and if AMD was ever in the gutter that ended with ryzen
 
Looks good to me. The 3D cache is a good idea and gave them a speed advantage before which is why Nvidia has boosted their cache so heavily this time around. The problem is they were out of bandwidth and it couldn't fix that no matter how fast it was. If they expand it this time then that will be one less problem holding them back and a good call. I haven't heard any complaints about their drivers since the last card launch 2 years ago either. Sounds promising, I'm hoping they get it together this time.
 
Last edited:
I don't see, "falls on its face," but very competitive 4k raster performance overall from Amd.

There are a few games that are commonly benched where AMD has a 15-20FPS lead at 1080P and it drops off to a 10FPS disadvantage at 4K. Perhaps "falls on it's face" isn't an accurate description but they clearly had an advantage with 3D cache and ended up failing at 4k where it matters to most gamers that are buying $1000 cards. 4K and raytracing needs to be focused on if they want to pull gamers away from NV. That means expanding their ability to ray trace without it killing the raster performance and getting rid of hard bandwidth limitations, not effective memory limitations under certain benchmarks but really lifting the throughput per cycle on the card.
 
Last edited:
Looks good to me. The 3D cache is a good idea and gave them a speed advantage before which is why Nvidia has boosted their cache so heavily this time around. The problem is they were out of bandwidth and it couldn't fix that no matter how fast it was. If they expand it this time then that will be one less problem holding them back and a good call. I haven't heard any complaints about their drivers since the last card launch 2 years ago either. Sounds promising, I'm hoping they get it together this time.

one problem with a 384-bit memory bus is it will mine better :bleh:
 
Serious games care not about the 30 series any longer. The 40's are the only thing on my watch list.

But... if a 3080Ti happens to cost $400 or less, then that could potentially sway my opinion.

DEEEEEEEEEEP discount it MUST be.
 
Back
Top