• 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: Will you be buying a Radeon VII?

Will you be buying a Radeon VII?


  • Total voters
    352
Associate
Joined
24 Jun 2016
Posts
842
Location
Hartlepool
No, AMD never intended to give a consumer gaming card that much bandwidth. It all comes back to the fact that it is a salvaged part from failed Instinct Mi60 GPUs. Once the 4 stacks of HBM2 are mounted to the interpose you can't remove them. Therefore you get 16GB VRM with 1TB/S bandwidth, of the whole lot goes in the bin.


If AMD wanted to have a dedicated gaming part they would have stuck to 2 stacks of HBM for 8GB and 512GB/s

You're spot on with this. AMD would have been nuts to throw away all these 'failed' Instinct chips and HBM. An opportunity presented itself when Nvidia priced Turing so high where AMD could sell these very expensive parts.

Any money AMD can make (or save) by selling these is better than chucking it in the bin. The fact it is a single card and not a Vega56/64 replacement backs up the salvage part theory.
 
Caporegime
OP
Joined
24 Sep 2008
Posts
38,322
Location
Essex innit!
I totally agree. The 2080 was one of the most disappointing Nvidia products I can remember and made quite a few YouTubers' worst products of the year lists, for AMD to match this in terms of performance and price while having fewer features is not a good strategy.

Monitors are also a big bone of contention for me vs TVs, you get so much less for your money. Where are all the 1440P OLED HDR monitors for a reasonable price (£500-700)?
Can't argue with that. I do love my UW screen but paid £1400 for it and that is mental when you consider you can get a 55" OLED for a little more (last time I checked).

Also not helping me is gaming time and how little I do when I get it now. Nothing is sticking out for me and the only 2 games that have me a little excited are Anthem and TD2
 
Soldato
Joined
28 Oct 2011
Posts
8,354
How good the cards actually are has almost become irrelevant, red or green, the prices are just so crazy the old tiers have doubled in price from just a few years ago. As someone said earlier the AAA games market is fubar, pre-order culture, early access, PTW, micro transactions, season passes, games-as-a-service, always online, empty open worlds where rather than be entertained you're left in a sandbox paying for the privegae of doing 'busy work', loot boxes, gated content, unfinished games with game modes marked *coming soon, pub/devs unwilling to make SP games and so forth. I haven't bought a AAA for years and won't be doing so until things change.

I'll possibly upgrade once more and that will be it I think. I'll certainly be buying a PS5.
 
Caporegime
Joined
18 Oct 2002
Posts
32,615
60CUs, so less than Vega64, when we know the chip has 64CU shows it is salvage part, plus being able to sell it at all for $700 when it likely cost near enough $700 to make.



It would be interesting if Nvidia knocks $100 off the 2080 (and $150-200 off the 2080ti). Would AMD be able to justify a price match.
 
Soldato
Joined
30 Dec 2011
Posts
5,419
Location
Belfast
Some youtube "experts" are suggesting they could make it cheaper by removing half the ram. Not really sure how that's supposed to work considering it would half the memory bandwidth plus as they're supposedly instinct cards that didn't make the cut the ram would already be on the package, even if they disabled it the fact its on there costs them money and hurts the card overall.

Same as the gddr 6 argument, the core was setup for hbm as was the pcb design, changing to gddr6 would require a lot of work on both the core as well as having to design a new pcb with traces for the ram modules, PLUS having to alter the heatsink to make sure the ram modules had places to touch the heatsink material to help dissipate heat.

I have tried explaining this to some on other forums. It is a cut down Radeon Instict card and as such cannot by design take less VRAM or different VRAM to reduce consts. I belive AMD would never have released Vega 7nm to consumers if Nvidia had sensible pricing on 20x0 GPUs.
 
Soldato
Joined
26 Sep 2010
Posts
7,146
Location
Stoke-on-Trent
This is AMD's Volta equivalent, Navi will be their Turing equivalent. This is a good sign of what AMD can bring to the table later this year with Navi, I expect a 4096 core Navi to be firmly in between the VII and the 2080Ti, whilst having a smaller die size than the VII and GDDR6 over HBM2.

And if AMD manage to increase core count for Navi past the 4096 core limit, I'd say we'd be looking at 2080Ti levels of performance for a lot less die size, but there's still a lot of speculation over whether that will happen.

But that's not what Navi is about. Navi was always suggested to be a tiny chip focussed squarely and purely at the midrange market facing off everything RTX 2070 and down. That approach was given further credence by the AdoredTV leaks with naming convention, performance point and price.

And if roadmaps are still on track, everything is reset with Arcturus in about 20 months, so I can't see Navi getting the "big" treatment as it will just be a sidegrade from the Radeon 7, albeit at a lower price point.
 
Associate
Joined
27 Dec 2008
Posts
402
It's not the core count limiting the performance. With Navi they could go with fewer cores and have better performance still compared to Vega 7.

I don't know about that unless they increase perf/core quite a bit. Nvidia has better perf/core and they're still going with more cores than Vega for their high end parts. It's easier for them to increase cores than increase perf/core.

I think Navi will definitely bring more efficiency into GCN in terms of memory compression and better perf/core, but AMD will have to bring it past that current 64 CU/4096 core limit if they want to beat the 2080Ti in overall performance.
 
Associate
Joined
27 Dec 2008
Posts
402
But that's not what Navi is about. Navi was always suggested to be a tiny chip focussed squarely and purely at the midrange market facing off everything RTX 2070 and down. That approach was given further credence by the AdoredTV leaks with naming convention, performance point and price.

And if roadmaps are still on track, everything is reset with Arcturus in about 20 months, so I can't see Navi getting the "big" treatment as it will just be a sidegrade from the Radeon 7, albeit at a lower price point.

The mid range Navi that will launch first will definitely be a smaller chip, but they are also rumoured to launch a high end Navi towards the end of the year or this time next year. That should be a direct competitor to the likes of the RTX 2080 and 2080Ti and maybe even the Titan RTX.
 
Soldato
Joined
24 Aug 2013
Posts
4,549
Location
Lincolnshire
I’m interested in seeing how it turns out, to see if we have some competition. But with having a 2080Ti already it won’t be something I’ll be changing to so no won’t be buying one myself.
 
Man of Honour
Joined
21 May 2012
Posts
31,940
Location
Dalek flagship
This is exactly the point. 7-8% more FP64 performance for 4.29X the price.
I hope you are not using any of these, especially in CrossFire or SLi only to play games. :D

No

The point you tried to defend was the bandwidth was needed on the Radeon VII because of the FP64 performance.

I pointed out that the Titan V has better FP64 performance despite having a lot less memory bandwidth.

The argument has nothing to do with price and everything to do with you making a mistake and trying to cover it up.
 
Permabanned
Joined
2 Sep 2017
Posts
10,490
No

The point you tried to defend was the bandwidth was needed on the Radeon VII because of the FP64 performance.

I pointed out that the Titan V has better FP64 performance despite having a lot less memory bandwidth.

The argument has nothing to do with price and everything to do with you making a mistake and trying to cover it up.

I never argued about the memory bandwidth. It is a result of using four HBM2s. If they use two, the bandwidth will be half. If they use four but with half the capacity, the bandwidth will remain the same.
 
Man of Honour
Joined
21 May 2012
Posts
31,940
Location
Dalek flagship
I never argued about the memory bandwidth. It is a result of using four HBM2s. If they use two, the bandwidth will be half. If they use four but with half the capacity, the bandwidth will remain the same.

No

You posted screenshots of the Radeon VII and RTX Titan FP64 performance to defend the fact that the former card needed the very high memory bandwidth.

All I did was post a screenshot of the Titan V FP64 performance which is even higher despite using less memory bandwidth.

Man up you got one wrong.
 
Man of Honour
Joined
21 May 2012
Posts
31,940
Location
Dalek flagship
Having had time to think about it I may get a Radeon VII but it won't be on launch day. If I get one it will most likely be a Strix, Nitro or equivalent when they are available in good numbers.

The move to 7nm is interesting and trying the tech out first hand could be fun.
 
Soldato
Joined
26 Sep 2010
Posts
7,146
Location
Stoke-on-Trent
The mid range Navi that will launch first will definitely be a smaller chip, but they are also rumoured to launch a high end Navi towards the end of the year or this time next year. That should be a direct competitor to the likes of the RTX 2080 and 2080Ti and maybe even the Titan RTX.

The only thing I've ever heard about a "Big" Navi was in the Navi AdoredTV video when Jim said his source had "seen a Big Navi running and it was impressive". And there's been nothing else. And why would a Big Navi wait until mid 2020 when that's when Arcturus is expected?

It's possible the Radeon 7 is the start of a new pattern: AMD's current gen is built to the max to cover the top tier with the new architecture starting small to cover the midrange. So Radeon 7 covers off the 2080 now, Navi 10 covers 2070 and down. Then Navi 20 replaces the Radeon 7 at the top around this time next year with new Arcturus chips replacing Navi 10 later on. A year later there's a big Arcturus and Arcturus+ comes in.

Who knows.

We'll see how things pan out come Computex. If there's not talk of Navi in June then I'd be a bit worried about what's going on.
 
Man of Honour
Joined
21 May 2012
Posts
31,940
Location
Dalek flagship
The only thing I've ever heard about a "Big" Navi was in the Navi AdoredTV video when Jim said his source had "seen a Big Navi running and it was impressive". And there's been nothing else. And why would a Big Navi wait until mid 2020 when that's when Arcturus is expected?

It's possible the Radeon 7 is the start of a new pattern: AMD's current gen is built to the max to cover the top tier with the new architecture starting small to cover the midrange. So Radeon 7 covers off the 2080 now, Navi 10 covers 2070 and down. Then Navi 20 replaces the Radeon 7 at the top around this time next year with new Arcturus chips replacing Navi 10 later on. A year later there's a big Arcturus and Arcturus+ comes in.

Who knows.

We'll see how things pan out come Computex. If there's not talk of Navi in June then I'd be a bit worried about what's going on.

I would take anything on AdoredTV with a very big pinch of salt.

After listening to his predictions for Turing with a big bit of help from a source I think he made a bit of a fool of himself.
 
Back
Top Bottom