8GB 290x or 4GB Fiji?

[Noctifer616]

And before someone says that my current card is only 4GB, remember that under DX12 my card will be a full 8GB since all the memory will be accessible (or so we been told).

If the shared memory in DX12 will work in a similar way as in Mantle then memory is not actually shared in the sense that you now have 8GB or VRAM. What it allows is to create objects and images on one GPU, flag them as shared memory objects and other GPU's can now access that object.

In other words to implement this feature won't be that easy and will require quite some work. The end result will be saved memory space rather than a larger memory pool (at least that's how I understand it).

 

[h4rm0ny]

Can't properly answer this until we see benchmarks and costs for the new card. But if offered the choice between the two right now, I wouldn't hesitate to grab the new one. It's still going to be at least a couple of years before you suffer in any meaningful way, imo, from not having 8GB. Meanwhile the 390 is going to have new features and be significantly more powerful.

I don't think the question is between the two cards you list - definitely the new one, I think. The question is whether the new one is good enough to justify the cost if it is as pricey as the rumours are starting to suggest.

 

[swilli89]

Holy.. WAY too early for this thread. Neither have even been released!?

 

[maddie]

You forgot the rest of the story. Each card needs duplicate textures etc. So you are not getting 8GB at all in any way. Unless you somehow imagine that one card pulls memory from the other at a staggering 8GB/sec.

Multiadapter is mainly for post processing. And it adds lag as well.

http://www.pcper.com/reviews/Graphics-Cards/BUILD-2015-Final-DirectX-12-Reveal

I am quite aware that AMD tries to upsell the "4GB experience" after they shoot themselves in the foot going with premature HBM.

I see you are misleading again. Why do you do this?

To quote from your link:

"DirectX 12 also provides an alternative, called Explicit Multiadapter. This is a new concept for DirectX. Like OpenCL, individual GPUs can be separately addressed, send unique commands, and store unique data in memory. They do not even need to be similar in performance."

 

[ShintaiDK]

I see you are misleading again. Why do you do this?

To quote from your link:

"DirectX 12 also provides an alternative, called Explicit Multiadapter. This is a new concept for DirectX. Like OpenCL, individual GPUs can be separately addressed, send unique commands, and store unique data in memory. They do not even need to be similar in performance."

Its not me being misleading and claiming 4+4GB on 2 GPUs is suddenly like a single 8GB buffer in games. Its just not going to happen.

And you forgot to quote the rest:

DirectX 12 also provides an alternative, called Explicit Multiadapter. This is a new concept for DirectX. Like OpenCL, individual GPUs can be separately addressed, send unique commands, and store unique data in memory. They do not even need to be similar in performance. One possible application is for integrated GPUs to draw a layer of objects, such as a cockpit or a 3D HUD, over what the main graphics card draws. Max McMullen, Principle Development Lead for Direct3D and DXGI at Microsoft, specifically mentioned calculating VR/AR perspective warp on integrated graphics. He also showed the Unreal Engine 4 Elemental Demo with an integrated GPU drawing some of the post-processing effects while the primary GPU worked on the next frame.

 

[frozentundra123456]

Well, 980Ti leaks are up on another thread. Seems about 50% faster than 290x. Not sure Fiji will be any faster than this. Any of these gpus are way out of my league price wise, but if I were buying, i would err on the side of caution and go with the 980Ti unless Fiji offers a significant (20%+?) performance advantage. Less vram used with HBM, is still sort of an unproven technology, so I would not go with it unless there is a considerable performance advantage. If the rumors of 650.00 for 980Ti are correct, I think Fiji will have a hard time competing, at least at the prices they were hoping to ask for it.

 

[maddie]

Its not me being misleading and claiming 4+4GB on 2 GPUs is suddenly like a single 8GB buffer in games. Its just not going to happen.

And you forgot to quote the rest:

I see your confusion.

You are misunderstanding what was written. The case for the integrated GPU in conjunction with a discrete GPU is meant to show how a weaker integrated GPU can be used and not wasted. They are not claiming that you need duplicate textures, as you claim.

quote:
"One possible application is for integrated GPUs to draw a layer of objects, such as a cockpit or a 3D HUD, over what the main graphics card draws."

"He also showed the Unreal Engine 4 Elemental Demo with an integrated GPU drawing some of the post-processing effects while the primary GPU worked on the next frame."

Two discrete GPU will not need duplicated data. This was an outright claim in the article. No guessing needed.

 

[Spanners]

Cool, I hope something along those lines gets implemented I've had three generations of Intel iGPU and all three have never rendered a single frame.

 

[Carnage1986]

I think 4 GB is not enough for a card like 390X. Because Fiji aims 4K and ultra graphic details. If you buy a flagship card, you can't compromise texture detail or any effect that consumes VRAM. 7th generation consoles are out of time now and the developers will tend to use all of 8 GB unified memory of the new consoles. (More detailed textures, VRAM-eater technologies like Order Independent Transparency, very high shadows etc etc.)

 

[flopper]

I think 4 GB is not enough for a card like 390X. Because Fiji aims 4K and ultra graphic details. If you buy a flagship card, you can't compromise texture detail or any effect that consumes VRAM. 7th generation consoles are out of time now and the developers will tend to use all of 8 GB unified memory of the new consoles. (More detailed textures, VRAM-eater technologies like Order Independent Transparency, very high shadows etc etc.)

There be 8gb cards also.
question is at launch or a bit later.

 

[3DVagabond]

We need to see 4GB HBM vs 8GB GDDR5. We don't know yet. People can guess or assume, but none of us have any idea how HBM works in the real world.