-
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.
- Status
Looks like it might be Samsung. Raja mentioned they'll be dual sourcing, and the Vega GPU for Apple lists 400GB/s
Looks like they could be using slower and cheaper HBM2 on lower end Vega cards.
Also spotted on the iMac Pro tech specs is a 56 CU Vega card, and top one is 64 CUs.
https://www.apple.com/ie/imac-pro/specs/
https://www.apple.com/ie/imac-pro/specs/
Yeah it looks rather good actually; and it's impressive they're going to manage to shove an 18 core Skylake X in there, and 64CU Radeon Pro Vega, with up to 128GB DDR4.
That display takes the cake though! 5K, P3 Gamut support; even the new iPad Pro is impressive. That has a P3 Gamut display at 120Hz.
Plus AMD graphics in all their computers that have discrete GPUs.
It does suck that we've learnt a bit more about Vega from Apple than AMD though.
Also spotted on the iMac Pro tech specs is a 56 CU Vega card, and top one is 64 CUs.
https://www.apple.com/ie/imac-pro/specs/
Worth noting that 56CU model will be clocked at 1535 MHz, assuming the 11 TFLOP single precision figure is talking about the base model.
Pretty interesting in general you'll be able to get an iMac with a 1080Ti-class card (in compute numbers anyway). Especially since they're now supporting virtual reality.
Wonder if gaming on Mac could be making a comeback
EDIT: Just thought, have they said how thick the iMac Pro is? Or any pictures of it from the side?
It's one thing to get 18-core low-clocked Intel server CPUs in it, but another thing altogether to get a 64CU Vega card in it.
Does the iMac Pro have a pretty fat section, or is Vega more efficient than we're expecting?
iMacs of the past have always had mobile GPUs in them, due to thermal constraints.
Also running slower HBM2, most likely Samsung according to Videocardz as the model numbers don't match SK Hynix.
The Vega cards in the iMac run at 400GB/s, possibly the base model as well.
Where as the Frontier Edition has 480GB/s
Also true indeed. If it's running at slightly slower clockspeeds (the core that is, for thermal constraints) hopefully that won't matter. We've all yet to see how well HBCC works in practice across many examples.
I'm looking forward to the Frontier Edition's launch so we can see what's what.
Here's a Vega die shot in the meantime.
970 '4gb gddr5 memory'
Oh and the rops too dont forget them.
Nvidia do it too. Maybe if they said upto 4gb gddr5 memory they could have passed lol. However everyone knows that overclockers statement was a slip up. The 970 was not.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,831
Defensive much? Sheesh.
I'm looking forward to the Frontier Edition's launch so we can see what's what.
Here's a Vega die shot in the meantime.
Mmm, I'm looking forward to seeing what Frontier can do too.
Also that die shot is very nice, great photo in general.
And I've seen some people estimate the die size around 490mm^2 from it. Since the HBM2 stacks are meant to be 92mm^2 each.
For reference, GP102 (1080ti, Titan X/Xp) is 471mm^2 and GP104 (1070/1080) is 314mm^2.
The HBm2 memory SK hynix is advertising as soon available would give about 400GB/s so not obvious it is Samsung. Sure the code numbers don't match public catalog but there could be reasons for that.
You replied to some one in regards to what amd have stated. When amd said most improvments made since gcn was introduced. You replied with acording to amd. Then you replied to, well they know more than us with overclockers dream. Thought that was abit childish tbh. So i stated what nvidia have said before to enlighten you to the fact that its not just AMD that say stuff too. Nvidia have done it before and overclockers dream was more of a slip up than owt but AMD are ridiculed for it ever since.
Not realy defensive just pointing out you cant just knock AMDs credibility like that for overclcokers dream when it was just a slip up. See nvidia state the specs for the 970? They got away with that tbh. It was a non problem. I dont see D.P discredititng what nvidia say based on that. If it works one way it works both. Which is my point.
Do you need me to quote the exact words again? it categorically wasn't "just a slip up". When it comes to credibility you have to look at the overall picture. (Not that I'd be particularly given to believing anything that came out of nVidia's mouth at face value but they do kind of have slightly better form on actually delivering of late).
- Status