*BREAKING* AMD Radeon "Navi" (RX 5000) will not use GCN, it will use RDNA

During Computex, CEO Lisa Su announced that "Navi" will using a brand new architect called RDNA. She also stated that GCN will continue on with Vega only.

• More energy efficient than GCN
• Family named RX 5000
• RX 5000 outpacing RTX 2070 using Strange Brigade

10 percent faster over RTX 2070 would land on ~62 fps while NVIDIA releases RTX 2070 Ti and RTX 2080 Ti-R.

CU improvement is okay, but raster is the main bottleneck.

Lol Strange Brigade... and AMD cards are horrible overclockers. A 2070 can hit 2040MHz with ease.

F*** this whole keynote has been a joke.

If STRANGE BRIGADE gets them 5-10FPS more than a 2070 then lets put it this way in 80% of the other games even at stock clocks the 2070 will out perform it or be on par.

If the pricing will be $399 and $499... This will be no different from the Vega 56/64 launch, too little too late. No wonder Nvidia is going just launch a refresh and no new chips next, because they just don't have to.

That's awesome for the new consoles.

Since when is a Radeon 5700 series their top end?

I think AMD are not giving information on their top GPU.

I do wonder what other features it has but I have a feeling some info are choke held until MS' E3.

Ports, their version of DLSS, compute performance, h/w agnostic raytracing performance compared to Vega/Polaris, rt performance compared to RTX, etc.

It's actually just a revised GCN. It's still shit.

Strange brigade is AMD optimised, it will end up being slower than a 2070. AMD are always behind, when are they actually going to dump some money into a complete new design and get some decent engineers.

https://twitter.com/RyanSmithAT/status/1132961332746055681?ref_src=twsrc%5Etfw

This is interesting, someone tell me what it means.

Since RX-5700 is a little faster over Vega 64 and RX-2070 on STRANGE BRIGADE

When mainline "Unreal Engine 4" based games are factored in.

RX-5700 being slightly faster than Vega 64 is okay. Mainline "Unreal Engine 4" based games are hitting AMD GPUs badly.

I don't see Microsoft's Gear of War 4 improvements with other UE4 games. NAVI's improvements for mainline UE4 remains to be seen.

I purposely selected GTX 980 Ti over Fury X due to multiple games being based on UE4.

Lower power consumption Vega 64 OC ~= RX-5700

They should use Run-D.M.C.

LOL! This is so random/unexpected/stupid comment, but i got a good laugh out of that, thanks.

@ronvalencia: What this about the 2080 TI-R ? Refresh?

Console gamers should find this depressing. Your GPU tech for the PS5/Xbox whoknows will be beat by a mid tier GPU released 2+ years before the consoles launch. 60 fps will still be the minority, ray tracing will be low-end tier, and even resolution won't improve too much. At least the games will load fast (lol)

This will be a smaller jump than what we saw from launch Xbox One to Xbox One X.

It's a rumor. I think it's more likely to see a 2070 Ti since AMD again has nothing to challenge Nvidia at the top-end. The 2080Ti refresh was speculated in case AMD came out swinging. They didn't, so there's no reason for Nvidia to respond at the enthusiast level

Can't say I'm surprise expecting a 2070Ti showing up and some of us thought we wouldn't see a 1070Ti. Nvidia is sure gonna milk this whole RTX.

So it's Vega 64 in performance more or less, remains to be seen how much lower the power draw is or did going from hbm2 to gddr6 eat most of the gains. It also needs to be cheaper than 2070, preferably by a 100 or bust

And Nvidia is laughing again.

It's like listening to Louis Rossman talk about G3HOT. I hear the words, but they mean nothing to my brain.

@goldenelementxl: bu... But... The fast load times? Cerny says it will blow away the PC? LOL

Isn't that like the same every gen? Discount ps4/x1, they were weak to begin with.

I would likely buy one of those.

The price will be a major competitive factor for Navi. If RX 5700 priced for $299, RTX 2070 and the below will all die off.

@acepilot: obviously price is going to be the make or break, it needs to be 100 or more cheaper than 2070

It won't happen. They goofed with the VII, they will goof again here. And lets not forget the Ryzen prices are all higher than the leaks...

@FireEmblem_Man: It's still "GCN". Hint: look in Linux GPU driver change logs.

AMD thinks VII's 16 GB VRAM feature is a fair swap for RTX 2080 8GB's Tensor and RT cores.

Both RTX 2080 and VII has rapid pack maths up to quad rate INT8.

RTX 2080 has variable shader rate feature while VII has higher FP64 performance (1/3 the rate from FP32). MI60 has FP64 is half rate from FP32.

Both RTX 2080 and VII has rapid pack maths up to quad rate INT8 and double rate FP16/INT16 features.

RTX 2080 at 1900 Mhz has 11.2 TFLOPS FP32 with discrete 11.2 TIOPS INT32. Turing CUDA cores splits FP and INT workload paths.

VII at 1750 Mhz has 13.4 TFLOPS FP32 shared with INT32.

----

RTX 2080 TI reached ~17.2 TFLOPS FP32 (Single Precision)

AMD doesn't have raw TFLOPS superiority during NVIDIA's Turing GPUs.

Many games haven't been optimized for Turing's new features.

@goldenelementxl: It's so depressing judging performance of games where black-box compilers are biased towards one GPU-architecture (cough cough $hitworks, gimped image/color quality (Nvidia drivers using compression and using DLSS to improve FPS at 8-bit color 4:2:0)(thank goodness for madVR that allow me to be able to bypass Nvidia image damaging drivers for the best animu viewing experience on my GTX1080), as oppose to new consoles who's new CPU/GPU/RAM/Flash architecture and compilers will work hand and hand to provide a next-gen leap in gaming at a great perf/$

Blame AMD for not support compression coverage over certain operations https://www.reddit.com/r/Amd/comments/9du2w4/navi_update_gfx10_in_gpu_open/

currently DCC don't work for Unordered Access View (UAV), https://gpuopen.com/dcc-overview/

Navi seems to allow it