• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD VEGA confirmed for 2017 H1

Status
Not open for further replies.
Technically I would say they demonstrated "slowdown avoidance" (in the absence of enough RAM) as opposed to performance improvement. Like you said in the second paragraph the only reason this happened is that they ran the game with 2GB of memory so that game data does not fit in VRAM. This forced the card to move data back and forth from system RAM, which the HBCC powers.

In reality this is a non-issue as cards come with enough VRAM for games to work properly. Nowadays 4GB is more than enough for 1080p gaming and 8GB is more than enough for 1440p. Cards that are capable of 4K like the 1080ti have 11GB.

As long as you have enough to store all game data, you don't need the HBCC, the game will run at full speed and the HBCC cannot "improve" the speed. Only in cases where you don't have enough VRAM would the HBCC become handy and that would just prevent the slowdown.

On the other hand, it makes for future-proofing your card. If you buy a 4GB Vega that is fast enough for 4K and a game with 7GB of data sets comes along, you should be able to play it at full speed (without slowdown) because of the HBCC.

The HBCC is really a feature for professional use where cards doing data crunching need to process huge data sets.

The system with HBCC off was running 4GB and the system with HBCC On was limited too 2GB it still showed 50% better avg and 100% better min FPS
They is more too this than just saying we have 8GB Vram XX game only needs 4GB Vram you should be good.. We need more tests shown but here is the demo running Deus Ex You can easy see the difference on screen. On the left is 4GB the games recommended Vram spec vs 2GB HBCC

 
4Gb isn't enough for 1080p gaming. Dishonered 2 is using 5.5Gb at 1080p. (According to Guru3d)

My 970 which is more than powerful enough to run the game at this res, hitches badly due to vram. Currently down two levels to high to get it to pay fine.

I'm not sure using Dishonored 2 as any kind of representative example is a wise choice considering how universally panned the PC port is for being a pile of ****.
 
4Gb isn't enough for 1080p gaming. Dishonered 2 is using 5.5Gb at 1080p. (According to Guru3d)

My 970 which is more than powerful enough to run the game at this res, hitches badly due to vram. Currently down two levels to high to get it to pay fine.

Ok but still... my point is that the HBCC does nothing to IMPROVE speed. It can help AVOID SLOWDOWN in such cases. If you have an 8GB card you should be good for 1080p or even 1440p 'forever' (i.e. in practice you will within 3-4 years upgrade and it should hold you over until then). Cards like the 1080ti have 11GB in anticipation of 4K using more than even 8GB over the next 3-4 years. By then you will get a xx80ti or something like that with 16GB of VRAM and you will have never ran into the issue.

I'm not saying the HBCC is useless, I'm just saying it's really not a "gaming" feature but more of a "we built this for professional number crunching, but hey here's how we can use it in games: we'll make cards with little VRAM and get games to only keep their 'working set' in it so that there's no slowdown due to not having enough VRAM".

Most importantly: I'm trying to get people to realise that it is not something that IMPROVES performance: if a game runs at 80FPS with 8GB VRAM because all its data fits in that VRAM, then that game will not run with >80FPS in that same card with 8GB VRAM _plus the HBCC_ because there'd be nothing for the HBCC to do in that case: all the data is present in VRAM already!
 
I dunno, I'd bet money on Nvidia getting involved with a game soon after Vega releases and trying to get the memory usage pushed beyond 8gb's so the Ti and TXp's can be seen to stand above every other card just like they did with the Fury and ROTTR.

No need ROTTR will still do the job

Max settings
2160p

DX11
IHZveQD.jpg


DX12
NPmyFF4.jpg


Anyone know why the DX11 version runs faster, uses less memory, gives better performance and is bug free unlike the DX12 version.:D
 
No need ROTTR will still do the job

Max settings
2160p

DX11
IHZveQD.jpg


DX12
NPmyFF4.jpg


Anyone know why the DX11 version runs faster, uses less memory, gives better performance and is bug free unlike the DX12 version.:D


And like AMD saying how much of that Vram is really used? the game is just using up all the memory = bad performance
 
The system with HBCC off was running 4GB and the system with HBCC On was limited too 2GB it still showed 50% better avg and 100% better min FPS
They is more too this than just saying we have 8GB Vram XX game only needs 4GB Vram you should be good.. We need more tests shown but here is the demo running Deus Ex You can easy see the difference on screen. On the left is 4GB the games recommended Vram spec vs 2GB HBCC


Let's say:

- system A, has 4GB of VRAM and HBCC off: 25FPS minimum and 50FPS average
- system B, has 2GB of VRAM and HBCC on: 50FPS minimum (+100%) and 75FPS average (+50%)

What you need to understand is this: if you take Deus Ex and run it with an 6GB pack, then system A will SLOW DOWN because 6GB > 4GB and it will be pausing to move data back and forth. System B will SLOW DOWN LESS thanks to the HBCC. However, let's assume you have 8GB of VRAM meaning you never need to move data around. In that case the two systems will have identical performance:

- system A+, has 8GB of VRAM and HBCC off: 60FPS minimum and 80FPS average (best card can do when memory is no issue)
- system B+, has 8GB of VRAM and HBCC on: 60FPS minimum and 80FPS average (best card can do when memory is no issue; HBCC has no job here)

I hope that clarifies it for you.
 
Let's say:

- system A, has 4GB of VRAM and HBCC off: 25FPS minimum and 50FPS average
- system B, has 2GB of VRAM and HBCC on: 50FPS minimum (+100%) and 75FPS average (+50%)

What you need to understand is this: if you take Deus Ex and run it with an 6GB pack, then system A will SLOW DOWN because 6GB > 4GB and it will be pausing to move data back and forth. System B will SLOW DOWN LESS thanks to the HBCC. However, let's assume you have 8GB of VRAM meaning you never need to move data around. In that case the two systems will have identical performance:

- system A+, has 8GB of VRAM and HBCC off: 60FPS minimum and 80FPS average (best card can do when memory is no issue)
- system B+, has 8GB of VRAM and HBCC on: 60FPS minimum and 80FPS average (best card can do when memory is no issue; HBCC has no job here)

I hope that clarifies it for you.

Let's wait and see for more testing ;) I think you going to be quite surprised by the results.
 
If the AMD system actually works, theres no doubt it will be the direction to go. They can aim for quality not quantity of ram, higher speeds possible and higher performance. I dont know that how it'll turn out but it gives options if they can really employ the brain of the card to good effect. We should all hope they have cracked it

4Gb isn't enough for 1080p gaming. Dishonered 2 is using 5.5Gb at 1080p. (According to Guru3d)

My 970 which is more than powerful enough to run the game at this res, hitches badly due to vram. Currently down two levels to high to get it to pay fine.
I'm not sure using Dishonored 2 as any kind of representative example is a wise choice considering how universally panned the PC port is for being a pile of ****.

Sounds a good first test to remember for release. Hand picked examples of performance is one thing but excelling with horrible code that doesnt even want to work with AMD tech is another and it still matters because maybe this game is my favourite and its gameplay is excellent however bad the port was
 
Something else, is that Vega could end up working better on a Zen based system due to it using lower latency CCIX to communicate between the CPU and GPU.
 
Developer dump all the data into VRAM, HBCC only puts what is needed into VRAM. This allows more VRAM for other things and reduces the amount of data that needs to be moved about leading to smoother framerates and better minimums.
 
If the AMD system actually works, theres no doubt it will be the direction to go. They can aim for quality not quantity of ram, higher speeds possible and higher performance. I dont know that how it'll turn out but it gives options if they can really employ the brain of the card to good effect. We should all hope they have cracked it

There is a very good chance it could work well.

I used to run a pair of GTX 690s in a PC with a very fast RevoDrive. The cards only have 2gb of VRAM per GPU but often got away with going over that by offloading to the RevoDrive.
 
Whether it is good or bad is not the point, games like this exist and are not going to go away.
And we should all just rush to the checkout and give these developers of ******* games our money and then use them as a yard stick. When similar things happen in the professional word companies tend to go bust, well unless your apple or adobe :P..
 
And we should all just rush to the checkout and give these developers of ******* games our money and then use them as a yard stick. When similar things happen in the professional word companies tend to go bust, well unless your apple or adobe :p..

Bloody Adobe! I can't be arsed with them for much longer; buggers still cant use many cores effectively, nor GPUs.

DaVinci Resolve is becoming an NLE, and their early versions are already faster; might look into it more.
In that software an RX 480 was barely slower than Titan X ( Maxwell ), so hopefully Vega is a massive jump up.

Just needs to give me some decent gaming performance as well.
 
Status
Not open for further replies.
Back
Top Bottom