Radeon VII neck and neck with RTX 2080 in rumored 3DMark leak

[Velcade]

5 minutes ago, COTG said:

well you are part of the reason nvidia can now fuck their customers

Yeah it's my fault AMD can't get their shit together and release cards that out perform NVidia.

[COTG]

6 minutes ago, Velcade said:

Yeah it's my fault AMD can't get their shit together and release cards that out perform NVidia.

"I'd never buy an AMD card" is stating that you woulnd't buy an AMD card even if it had better performance at a lower price. 

AMD actually always had better performing cards at a lower pricepoint in the midrange, but mindsets like your's are what almost bankrupted the company.

[Tristerin]

15 minutes ago, Mira Yurizaki said:

Removing 8GB resulting in lost bandwidth is partially true, if they actually remove the memory layers themselves. The memory topology is that each chip is its own channel and the whole thing works like a massive n-channel memory system (or to spin it another way, a massive RAID-0 setup with memory chips).

 

However, you can also use less dense memory. The GTX 1060 3GB and 6GB launch models both had the same bandwidth. The RX 580 4GB and 8GB models both had the same bandiwdth.

 

I don't believe however that halving the amount of VRAM will also drastically lower the price, as in the case of those two GPUs above. In the case of the RX 580 specifically, the 8GB model was only $30 more than the 4GB model.

Wouldn't that be the difference in GDDR5 vs HBM2 though?  Huge cost difference compared (from what Ive been told/read)

[Mira Yurizaki]

1 minute ago, Tristerin said:

Wouldn't that be the difference in GDDR5 vs HBM2 though?  Huge cost difference compared (from what Ive been told/read)

The manufacturing would be the same, just with fewer things to put on the memory layer. So I don't think that would really save anything outside of material cost and it might not even be that much. They could also also make 16GB packages but make the tolerances looser, so they're still making the full shebang.

[mr moose]

 

I thinks it's awesome that AMD are putting up some competition, of course how good it is for the $$ only time will tell.

 

I also find it frustrating that when people legitimately have an AMD product to be enthusiastic about they still insist on tired old fallacies about RTX being a gimmick.    Let's ignore the facts (even the one about AMD producing a product worth talking about, heaven forbid it even be worth buying) and concentrate on shitting on it's competition.

 

 

[Dabombinable]

Still waiting to see if Vega VII will actually have full HW decoding of VP9 and HVEC

[mr moose]

11 minutes ago, COTG said:

 but mindsets like your's are what almost bankrupted the company.

No, what nearly bankrupted the company was years of poor decisions, lackluster products and a GPU architecture that required significantly more R+D to keep it competitive in the midrange department.

 

You can't blame consumers for not wanting to buy second best.

[Tristerin]

13 minutes ago, mr moose said:

No, what nearly bankrupted the company was years of poor decisions, lackluster products and a GPU architecture that required significantly more R+D to keep it competitive in the midrange department.

 

You can't blame consumers for not wanting to buy second best.

When the market evolves you either change with it...or get left behind.  They almost got left behind.  ALMOST fortunately.

[BuckGup]

I am more interested in benchmarks outside of gaming. Like rendering videos,, hashcat, mining, and so on. As that could easily justify not going with ray tracing as you would never use 2/3 of the GPU you overpaid for

[Mira Yurizaki]

4 minutes ago, BuckGup said:

I am more interested in benchmarks outside of gaming. Like rendering videos,, hashcat, mining, and so on. As that could easily justify not going with ray tracing as you would never use 2/3 of the GPU you overpaid for

That figure should be higher, because of all the SFUs those functions wouldn't be touching either.

 

Which also exist on AMD cards too.

[Madgemade]

1 hour ago, Dabombinable said:

Still waiting to see if Vega VII will actually have full HW decoding of VP9 and HVEC

As it's just a die shrink of Vega 10 with some compute feature added then you are probably going to be disappointed regarding this.

[Mihle]

1 hour ago, mr moose said:

No, what nearly bankrupted the company was years of poor decisions, lackluster products and a GPU architecture that required significantly more R+D to keep it competitive in the midrange department.

 

You can't blame consumers for not wanting to buy second best.

Well, at the time where and was more perf compared to cost in the GPU segment with a good margin, more people still bought Nvidia than AMD because mindshare.

CPU is another matter tho....

[Mira Yurizaki]

3 minutes ago, Mihle said:

Well, at the time where and was more perf compared to cost in the GPU segment with a good margin, more people still bought Nvidia than AMD because mindshare.

I keep hearing this but nobody provides any actual examples.

[Mira Yurizaki]

1 minute ago, valdyrgramr said:

Well, AMD themselves even blamed FX from what I remember.  GPUs might have caused a small impact, but it was mostly their choices with FX.  One engineer even said it was bad who helped make FX, iirc.   It makes sense since that's when people started buying Intel way more due to the single thread performance differences and bottlenecks.

I was asking more about the GPU side where people keep claiming about how you got a better value buying from Team Red in the midrange. And according to one person, it's always a better value.

[Mihle]

21 minutes ago, Mira Yurizaki said:

I keep hearing this but nobody provides any actual examples.

Around GTX 480 times for example if I remember right. Don't have time to check exact numbers or give sources rn.

[BuckGup]

55 minutes ago, valdyrgramr said:

For me, I wanna see how well it does in Maya, Blender, and a few others.  The Vega 64 did really well in Blender.

Yeah faster clocks, more VRAM, and a huge bandwidth bump makes it sound good

[Mira Yurizaki]

23 minutes ago, valdyrgramr said:

Well, that depends on the market. 

For the sake of argument, final retail price is irrelevant because it's not consistent enough. It's the same reason why I ignore any argument that uses the used market as ammo.

[yian88]

9 hours ago, exetras said:

RTX is like 3D movies, looks nice but add nothing much in terms of value. For now.

RTX is like 3d/hdr/virt reality/aug rality   sounds nice  but no one uses it, forever.

 

Fixed it for you.

[Mira Yurizaki]

21 minutes ago, valdyrgramr said:

I'm still waiting for real-time raytracing to mature more and hit games I actually care about, same with DLSS.  Until then Radeon VII is the ideal card for me.

DLSS should be a driver wide thing, like MFAA. I don't know what NVIDIA is doing making it a "must be explicitly supported" thing.

[leadeater]

9 hours ago, CarlBar said:

 

Seriously people need to go take a look at DX12 uptake. In the first 3 months after DX12 release there was exactly one title with DX12 support, and that was a microsoft title developed with in house knowledge. The first out of house game didn't arrive till 3 months after DX12 dropped. he next 4 took another 4 months, or 7 months after launch, to arrive.

 

Get back to me at the end of june and then we can talk weather uptake is good or not on RTX.

I know this is a quote from page 2 so don't slap me too hard if it's already been discussed. Most of the DX12 games that have come out are/have been just DX12 wrapped over DX11 with no DX12 features or optimizations. One of the tell tale signs of that is DX12 performance regression when you enable it.

 

At least RTX will also help with proper DX12 development, they are pretty much hand in hand. You might as well do a proper DX12 implementation if you have to use DX12 explicit features to use RTX in the first place.

[leadeater]

1 hour ago, Mira Yurizaki said:

I was asking more about the GPU side where people keep claiming about how you got a better value buying from Team Red in the midrange. And according to one person, it's always a better value.

There were times where that was true, or even when AMD did actually have the highest performance product. The problem is those are point in time situations in a time when GPU architecture and fab nodes were happening faster. In reality, how it has actually worked out, is that AMD won a few races but in no way won the championship over all. Supporters will always like to point one those few race wins, that's just how it is.

 

Personally I'd rather not go drag up history and list out the past GPUs, which ones were good, how they sold, when the competition introduced new cards etc etc because frankly it's all irrelevant now at this point in time. Do we actually care? I'd have the discussion if it was one about GPU development history or some similar topic but this topic isn't that.

[Mira Yurizaki]

2 minutes ago, leadeater said:

Personally I'd rather not go drag up history and list out the past GPUs, which ones were good, how they sold, when the competition introduced new cards etc etc because frankly it's all irrelevant now at this point in time. Do we actually care? I'd have the discussion if it was one about GPU development history or some similar topic but this topic isn't that.

I don't really care about talking about the history itself, but it's more that people aren't just spouting out random "facts." Especially when people seem to conveniently forget history for the sake of bolstering their argument.

 

26 minutes ago, leadeater said:

I know this is a quote from page 2 so don't slap me too hard if it's already been discussed. Most of the DX12 games that have come out are/have been just DX12 wrapped over DX11 with no DX12 features or optimizations. One of the tell tale signs of that is DX12 performance regression when you enable it.

Nitpicking here, but the overall feature set with the exception of DXR as far as producing the image itself is identical between DX11 and DX12. So speculating here, the only uplift you'd get from DX11 to DX12 is the creation of multiple queues so GPUs that can do asynchronous compute can do so. But if the developer didn't even bother with that, then yeah, that's more of a marketing move.

[CarlBar]

1 hour ago, Mira Yurizaki said:

DLSS should be a driver wide thing, like MFAA. I don't know what NVIDIA is doing making it a "must be explicitly supported" thing.

 

DLSS relies on running a training algorithm on a big powerful supercomputer. They need a set of reference images to work from to train it on. if developers won't provide that they can't do it.

 

Once they done this on a few thousand different games they'll probably be able to produce a general purpose algorithm from it. But right now they have to build a custom DLSS algorithm for each game as each game has slightly different requirements. I recommend looking up codebullet on youtube, he's done a whole bunch of stuff with training required type AI's playing games and it's the same broad principles.

 

46 minutes ago, leadeater said:

I know this is a quote from page 2 so don't slap me too hard if it's already been discussed. Most of the DX12 games that have come out are/have been just DX12 wrapped over DX11 with no DX12 features or optimizations. One of the tell tale signs of that is DX12 performance regression when you enable it.

 

At least RTX will also help with proper DX12 development, they are pretty much hand in hand. You might as well do a proper DX12 implementation if you have to use DX12 explicit features to use RTX in the first place.

 

Sure but BFV is a pretty janky RTX implementation by all accounts so how good a DX12 implementation those games where is less relevant than weather it was DX12 supporting for a comparison point. Everytime i hear people go on about how ray tracing will never take off i want to strangle a kitten it's so frustrating.

[leadeater]

27 minutes ago, Mira Yurizaki said:

Nitpicking here, but the overall feature set with the exception of DXR as far as producing the image itself is identical between DX11 and DX12. So speculating here, the only uplift you'd get from DX11 to DX12 is the creation of multiple queues so GPUs that can do asynchronous compute can do so. But if the developer didn't even bother with that, then yeah, that's more of a marketing move.

Well I wasn't really saying that there would be a performance uplift with DX12, just that if you wrap it it'll be slower and so not as well optimized anymore. Most of the DX12 games that have been released all started development before there were builds of DX12 that were ready enough to use so they are DX11 games either with a DX12 option but not actually DX12 or only have DX12 but also not actually DX12.

[Lathlaer]

7 minutes ago, CarlBar said:

if developers won't provide that they can't do it.

From what I can remember, NVIDIA said that for big budget AAA titles the devs provide it to them anyway (for driver optimization, hairworks and whatnot) so it will be just another thing they can add.