AMD Radeon VII Retested With Latest Drivers

[arbiter]

AMD has been doing the same thing since 7970. It was Nvidia who removed compute performance from their GTX cards.Now they have added it back, charging you double calling it RTX.

So then explain as to why AMD card is same price then? How would nvidia adding compute performance back make it so AMD cards are same price?

 

[londiste]

AMD has been doing the same thing since 7970. It was Nvidia who removed compute performance from their GTX cards.Now they have added it back, charging you double calling it RTX.

What exactly do you mean by compute? The discussion here has been about FP64 performance. RTX is an entirely different type of computations and a very specialized one at that - BVH traversal.

FP64 is almost completely useless when it comes to gaming. It is useful in certain types of compute scenarios. Both manufacturers have struggled to find a balance between workstation/server/GPGPU cards and consumer cards in terms of compute features. If you look at the history, both have also settled to the balance points they decided upon - AMD at 1:16 and Nvidia at 1:32, with both trying to have a compute GPU at the top of their lineups that can do 1:2 or thereabouts.

When it comes to FP64, AMD history looks like this (a little messy due to reuse of GPUs over generations):
- HD4000/5000 high and midrange cards have FP64 at 1:5 FP32 (4870/4850/4770/4750, 5870/5850/5830). Lowend does not do FP64.
- Some of (higher end) HD6000/7000 have 1:4 (Tahiti, 7950/7970). HD7000 midrange has 1:16 (Pitcairn, 7870/7850), lowend has 1:16. Some really lowend things do not do FP64.
- High end R* 200 series (Hawaii, R9 290/290X) has 1:8, midrange has 1:4 (Tahiti, R9 280/280X) or 1:16 (Tonga, R9 285/285X) and lowend has 1:16. Some really lowend things do not do FP64.
- Fiji (Fury/FuryX) has 1:16
- RX400/500 has 1:16
- Vega10 (Vega56/Vega64) has 1:16
- Vega20 (Radeon VII) has 1:4

FP64 situation on the NVidia side looks like this:
- GTX200 series high end (GTX280/260) has 1:8.
- GTX400 series (Fermi) high end (GTX480/470) has 1:8, midrange and lowend (GTX460/450/440/430) has 1:12 and lowest end does not do FP64.
- GTX600 series (Kepler) has 1:24, except Titans at 1:3 and some lowend cards that are Fermi and have 1:12.
- GTX900 series (Maxwell) has 1:32.
- GTX1000 series (Pascal) has 1:32.
- Volta (Titan V) has 1:2.
- RTX2000/GTX1600 series (Turing) has 1:32.

 

[bug]

What exactly do you mean by compute?

I believe he was referring to this: https://en.wikipedia.org/wiki/Maxwell_(microarchitecture)#Performance
However, that's Nvidia cutting back on double precision computing power, which is really different from what RTX and tensor cores do.

 

[Redwoodz]

What exactly do you mean by compute? The discussion here has been about FP64 performance. RTX is an entirely different type of computations and a very specialized one at that - BVH traversal.

FP64 is almost completely useless when it comes to gaming. It is useful in certain types of compute scenarios. Both manufacturers have struggled to find a balance between workstation/server/GPGPU cards and consumer cards in terms of compute features. If you look at the history, both have also settled to the balance points they decided upon - AMD at 1:16 and Nvidia at 1:32, with both trying to have a compute GPU at the top of their lineups that can do 1:2 or thereabouts.

When it comes to FP64, AMD history looks like this (a little messy due to reuse of GPUs over generations):
- HD4000/5000 high and midrange cards have FP64 at 1:5 FP32 (4870/4850/4770/4750, 5870/5850/5830). Lowend does not do FP64.
- Some of (higher end) HD6000/7000 have 1:4 (Tahiti, 7950/7970). HD7000 midrange has 1:16 (Pitcairn, 7870/7850), lowend has 1:16. Some really lowend things do not do FP64.
- High end R* 200 series (Hawaii, R9 290/290X) has 1:8, midrange has 1:4 (Tahiti, R9 280/280X) or 1:16 (Tonga, R9 285/285X) and lowend has 1:16. Some really lowend things do not do FP64.
- Fiji (Fury/FuryX) has 1:16
- RX400/500 has 1:16
- Vega10 (Vega56/Vega64) has 1:16
- Vega20 (Radeon VII) has 1:4

FP64 situation on the NVidia side looks like this:
- GTX200 series high end (GTX280/260) has 1:8.
- GTX400 series (Fermi) high end (GTX480/470) has 1:8, midrange and lowend (GTX460/450/440/430) has 1:12 and lowest end does not do FP64.
- GTX600 series (Kepler) has 1:24, except Titans at 1:3 and some lowend cards that are Fermi and have 1:12.
- GTX900 series (Maxwell) has 1:32.
- GTX1000 series (Pascal) has 1:32.
- Volta (Titan V) has 1:2.
- RTX2000/GTX1600 series (Turing) has 1:32.

Well you stated the basis there, Nvidia using 1:32 vs AMD using 1:16 for the past couple generations.

 

[notb]

Well you stated the basis there, Nvidia using 1:32 vs AMD using 1:16 for the past couple generations.

But you said that Nvidia "removed compute", so it's doesn't really matter how it relates to AMD. It's important to check what happened over the years, because that's what you're referring to.
And @londiste collected the data.
Over the last 10 years Nvidia moved from 1:8 / 1:12 to 1:32.
In the same period AMD moved from 1:4 / 1:5 to 1:16.

So it does seem that they both "removed compute". Do you agree?

 

[hat]

Which is really ironic since their compute accelerators aren't really a hit (that's why they're rebranding them for gaming).
They wanted to unify workstation and high-end gaming. This whole business strategy turned out to be a failure.

Everything could change if AMD focused on making purpose-built gaming chips and unify consoles and PC gaming, which would make sense for a change...

We'll see what happens with their datacenter products. IMO they'll give up and Intel takes over.

Eh, maybe. While it's true a console chip's primary purpose would be, well, gaming, console chips are also custom designed (to some extent) and last years. This latest generation (PS4/Xbone) being the exception, with bigger and badder variants of the same console (a la PS4 Pro), previous consoles use more or less the same hardware for a long, long time compared to desktop parts. Every year or so a new generation of graphics cards come out... meanwhile, the same hardware lasted 7 years with the PS3. 5 years and counting with the PS4, if you toss out the PS4 pro...

So, it's not quite the same animal. Making desktop chips requires constant improvements, where as you make a decent chip for a console and it's expected to last 5 years or more. AMD already has Navi in the bag, primarily for the PS5, and we'll see that too on desktops, but it remains to be seen how well it will do, or what comes after that.

 

[efikkan]

I think some need to look up what "compute" actually means. fp16, fp32, fp64, int32, FMA and tensor operations are all referring to compute, compute operations should not be confused with what people call "compute" workloads, which is typical CUDA or OpenCL simulations etc.

The thing about fp64 is that it's not normally used during rendering, in fact fp32 is overkill for parts of the rendering in games. E.g. after rasterization, when processing fragments("pixels"), only a tiny fraction of the precision of fp32 is actually used.

 

[Raven Rampkin]

Yes, RX570 wipes the floor with GTX1050Ti. They are cards from different segments.
RX570 is the intended competitor of GTX1060 3GB.
GTX1050Ti is the intended competitor of RX560.

Pricing in the lowend and midrage is FUBAR.

Looks like it's only FUBAR when the AMD option is cheaper