• 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.

Poll: ** The AMD VEGA Thread **

On or off the hype train?

  • (off) Train has derailed

    Votes: 207 39.2%

  • (on) Overcrowding, standing room only

    Votes: 100 18.9%

  • (never ever got on) Chinese escalator

    Votes: 221 41.9%
  • Total voters
    528
Status
Not open for further replies.
TheRealDeal said:
No as his results line up decently with the 1750 result which has no memory overclock. The core difference would make little difference but the memory overclock probably would. They also said 3dmark reports the top core speed. His top may have been lower but more constant. There is no reason to doubt the reported top core speeds as from what i have seen with all the Vega results is nothing but believable clocks.
Click to expand...

The people reviewing those cards will be using the stock settings and clocks.

If your link is showing something different it is because the guy has overclocked both the core and memory and 3dmark has got it wrong, something that happens quite often.

I also notice from your link that the guy has overclocked his RZ1800X to 4.225 which is pretty good, if he is going to do that he is not going to underclock the GPU.

Judging from his Ultra GPU score he is probably running the card at something like 1820/1050 which is in line with his CPU performance.
 
lolz at that price .. see if my 290 will o/c some more £499 then yes as I want a ek block to go with it but at £700 not got a chance ..
 
flea.rider said:
lolz at that price .. see if my 290 will o/c some more £499 then yes as I want a ek block to go with it but at £700 not got a chance ..
Click to expand...
Never tried overclocking my RX 480 (well 12 MHz but that doesn't count) but I might be able to get some more out of it if needed in the future. I was a bit confused yesterday though cos temps were pushing 75 degrees (whereas normally they are sub-65) despite it being a cold evening. I thought Afterburner had failed to apply my custom fan configuration after updating it but it was running at 50% under load as normal. *strokes chin*
 
Kaapstad said:
A lot of the cards in the post I made earlier were running their review cards @1750mhz on the core so yes 3dmark is probably reporting the core wrong in your link.

0PHH1mT.jpg
Click to expand...
The 1080/ti are 3rd party factory overclocked cards? So when the 3rd party Vega's come out they will be ahead of 1080's of any flavour?

The liquid cooled Vega 64 runs at 1677mhz and the blower runs at 1546mhz, so why do we see clocks specifically at 1630 and 1750 with different CPU's?
Either we are seeing a reference blower boosting past its advertised clocks, no results are at stock clocks or they are liquid cooled cards underclocking with some manual overclocks thrown in, but why have the memory clocks not increased also?
Or 3dmark is not reporting the clocks correctly.

Anyway, the core clock seems to give a jump in performance when getting near 1700 and then diminishing returns. I'm sure that memory clocks will offer a bigger jump in performance and maybe feel the core better past 1700mhz.
 
B12B6 said:
You do know games have been using ram and hdd/ssd for texture caches for ages? :D You seem to be pretty clueless.
For saying frame I tought you would understand I ment that everything is loaded what is shown on frame, textures included....
Why would 2 gpu card need more memory or banwich when they are sharing the memory? Gpu 1 would not need to load then what gpu 2 needs to load. 2 gpu desing with shared memory would not need to load everything twice. I didnt say breaking the frame multiple parts would help memory, I said it would help utilize shaderz that are not utilized now because of AMD design. And I dindnt say see how cpu are build you can make gpus same way. Even Nvidia has already stated multiple chip cards is way to go forward. Yields making small dies are far better than using big dies.
Click to expand...

oh dear... I have already explained this - at current, to prevent texture pop-in, GPU's load far more textures in to vram than are needed for just the current frame and then don't un-load them again until its really needed to do so. this requirement doesn't go away by splitting the frame in to smaller chunks because say in one frame one GPU only needs sky textures, but then the user looks down and now needs all the ground textures, then looks up again and needs the sky textures again - trying to do that on a frame by frame basis would not be quick enough and you would get texture pop-in or frame hitching while the GPU pauses to load textures, neither of which is desirable and neither of which is solved by off board caches... shaders being better utilised doesn't help with either of these issues either

yes they are both working on it but neither have released any details on how they intend to solve this particular issue, so where as now with crossfire/SLI where we are now seeing utilisation/improvements of less than 50% and typically only 20-30% (or even negative scaling in many games as devs have given up on support), doing an off-die memory controller that gives a consistent 60% would be much better than current crossfire/sli solutions, but its still no where near 100% or even optimistically 90% that would be needed to make a multi-die board cost effective in terms of selling a multi-die card vs. a monolithic one that doesn't have these issues

you can try cramming more dies on a single interposer to overcome this to an extent but then you run in to issues with interposer sizes

an API like vulkan or DX12 is 3 steps too far down the stack to make any difference and would need game dev support so is a non-starter as well, just as the previously touted "vram can be shared" rumours never came to fruition

none of this is AMD bashing, its a really fundamental requirement to get multi-die cards working well and neither AMD nor Nvidia have released any details on how they intend to work around this issue
 
Kaapstad said:
The people reviewing those cards will be using the stock settings and clocks.

If your link is showing something different it is because the guy has overclocked both the core and memory and 3dmark has got it wrong, something that happens quite often.

I also notice from your link that the guy has overclocked his RZ1800X to 4.225 which is pretty good, if he is going to do that he is not going to underclock the GPU.

Judging from his Ultra GPU score he is probably running the card at something like 1820/1050 which is in line with his CPU performance.
Click to expand...

Its probably gibbo raiding the stock cupboard again so he can post his own review @ 2pm on monday.
 
Hotwired said:
Well this all depends on how well it sells.

If they sell out and sell out for an exceedingly long time then I'm pretty sure they couldn't care less.
Click to expand...

If they sell to miners and not gamers it'll make their market share even worse in gaming surveys which could lead to devs taking them less seriously and being unwilling to optimise for AMD.
 
Status
Not open for further replies.
Back
Top Bottom