• 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.

@Gibbo - Vega Preorders?

Status
Not open for further replies.
CAT-THE-FIFTH said:
Its why I think memory bandwidth might be holding back the card. After all the Fury X has more memory bandwidth.
Click to expand...

Thats hynix's fault for only getting hbm2 upto 480mb/s/pin, maybe amd should have designed for gddr5/5x but they didnt and found themselves stuck with what they could get.

NV can afford to have two chips - one for HBM2 (P100) and ones for GDDR5/5X but AMD cannot.

But will still be buying one tomorrow.
 
Last edited:
MDPlatts said:
We don't know - probably some announcement on the forum (to be kept secret and not blabbed all over the web where miners may see it) will be posted by Gibbo tomorrow saying how it will work and what time to set alarms for - possibly in the :) thread but maybe in a new one. Maybe tomorrow am maybe lunchtime about the 2pm go-live.

It was only one possible suggestion he was considering but he may decide to do it a different way.

Basically nobody but Gibbo knows anything other than 2pm tomorrow (ish) :D
Click to expand...

I've been watching so many threads, I do hope I can nab one; as much as my Fury X is serving me well I feel the Vega 64 is the best card for my 4K freesync. I'm not expecting a discount, has Gibbo confirmed that anywhere?

I sense tomorrow may involve less work being done than a typical monday for me; as I refresh the forums and check the site. :D
 
CAT-THE-FIFTH said:
Its why I think memory bandwidth might be holding back the card. After all the Fury X has more memory bandwidth.
Click to expand...

They had to add near 4 billion transistors to let the clocks scale, apparently to the detriment of a longer pipeline (which is another reason why people think this is not much more than a fiji thats been heavily modified). Its very possible this could be pentium 4 territory where the core can show good numbers in terms of mhz but not necessarily performance. So 1900mhz might only get a 2-3 fps bump over 1600mhz, but its possible also it could scale well with performance.
 
Gerard said:
They had to add near 4 billion transistors to let the clocks scale, apparently to the detriment of a longer pipeline (which is another reason why people think this is not much more than a fiji thats been heavily modified). Its very possible this could be pentium 4 territory where the core can show good numbers in terms of mhz but not necessarily performance. So 1900mhz might only get a 2-3 fps bump over 1600mhz, but its possible also it could scale well with performance.
Click to expand...

Memory bandwidth is less than a Fury X whilst being a faster core - look at the TFLOPS rating for the core for example. Its obviously based on Fuji has it has the same basic number of shaders,ROPs,etc.

The thing is if you look at the main reasons for it being delayed it seems to match the timeframe of Hynix trying to get their HBM2 out.

I suspect AMD only went with a two stack solution since they probably expected a much higher clockspeed than then they actually got,and it would be released earlier. AMD also has now dual sourced HBM2 with Samsung and IIRC people suspect the Vega FE uses Samsung HBM2.
 
CAT-THE-FIFTH said:
Memory bandwidth is less than a Fury X whilst being a faster core - look at the TFLOPS rating for the core for example. Its obviously based on Fuji has it has the same basic number of shaders,ROPs,etc.

The thing is if you look at the main reasons for it being delayed it seems to match the timeframe of Hynix trying to get their HBM2 out.

I suspect AMD only went with a two stack solution since they probably expected a much higher clockspeed than then they actually got,and it would be released earlier. AMD also has now dual sourced HBM2 with Samsung and IIRC people suspect the Vega FE uses Samsung HBM2.
Click to expand...

The hbm2 delay and slower speeds is definetely one snag. The thing is in comparison of Vega's vs Fiji, you can't just base the gpu performance on memory raw bandwidth figures.
Vega has a 4mb l2 cache vs 2mb in Fiji, and then there's the modifications of interconnection and pathways of the l2 cache (primitive discard, Hbcc, Culling, Draw Rasterizer, deferred shading l2).
Vega is just a tweaked Fiji with front end improvements to improve load balancing, geometry throughout, shader utilization and power consumption whilst targeting a 1.7ghz clock speed.

The mistakes Amd have made are not transitioning to a 6 gemotry engine architecture, and not targeting a 4608 or 5120 cu size chip with the current 4 wide engine.
The battle Amd now have is pricing these big chips against the smaller cheaper Gp104.
It's all well and good people saying oh it needs to be cheaper than a 1070 or a 1080, but if Amd cannot price match or compete then it's effectively putting a noose around their necks.
 
Davedree said:
The mistakes Amd have made are not transitioning to a 6 gemotry engine architecture, and not targeting a 4608 or 5120 cu size chip with the current 4 wide engine.
The battle Amd now have is pricing these big chips against the smaller cheaper Gp104.
It's all well and good people saying oh it needs to be cheaper than a 1070 or a 1080, but if Amd cannot price match or compete then it's effectively putting a noose around their necks.
Click to expand...

I feel it's less of a mistake and more of resource issue, looking at the leaks and talk around Vega it was really only ready a few months ago and that's without having to deal with upping the CU count and making the die bigger (I actually think the GPUs would be almost to expensive at that point due to the size of the dies and the "hit rate" of defects).

AMD (I know people hate this argument on the web) have limited resources in comparison to Nvidia and intel, the last few years have obviously been on the CPU and APU division and the GPU side has taken a back seat (probably focusing on APUs). Yes Vega is underwhelming and that's mainly down to marketing pumping it up to be something it wasn't, I'm however guessing that Vega will scale well down to APU scenarios.


100% agree, price / price Vega cards need to be at the same price level (actually sale price not recommended retail prices) as the nearest Nvidia card. That looks like 1070 for Vega 56 and 1080 for 64.
 
mining cards with proper warranty terms and lower price will encourage miners to buy them, even with just an HDMI or cheaper port. If more mining cards are produced - enough for the miners - gamers will have enough cards. Only the smaller miners buy gaming cards with intention to resell after some time. The big miners buy to use them till they drop
 
ZombieGames said:
mining cards with proper warranty terms and lower price will encourage miners to buy them, even with just an HDMI or cheaper port. If more mining cards are produced - enough for the miners - gamers will have enough cards. Only the smaller miners buy gaming cards with intention to resell after some time. The big miners buy to use them till they drop
Click to expand...

Still boils down to producing much more cards.
 
I really hope Vega 64 air is not marked up to £599 based on unproven mining performance(both PCPER and Tweakdown say identical to Vega FE so more power hungry and less power efficient than polaris/1070). Cards performance in gaming and mining is simply not worth that amount of money.
 
Status
Not open for further replies.
Back
Top Bottom