I was wondering when will we see the use of 12Gb GDDR6 chips, it can bring some VERY interesting memory configurations, like having a 192BIT card with 9GB of video memory.
At least later today we can get a glimpse into 16Gb size GDDR6 chips.

;)

would these req one 6-pin or one 8-pin?

dj-electricAt least later today we can get a glimpse into 16Gb size GDDR6 chips.

;)

I thought those debuted with the 24GB TITAN RTX?

So it's a cut down 5700

btarunrI thought those debuted with the 24GB TITAN RTX?

oh yeah, oh well at least now it comes in a decently priced product.

KV2DERPSo it's a cut down 5700

Isn't just about any gfx card that's not the top end a cut down version though

If they can get 1660 super performance with Consuming around 125W and pricing around the same. Then it can be a great product.

AlienIsGODIsn't just about any gfx card that's not the top end a cut down version though

Technically not, cut-down usually means the same chip with portions of it fused off. That being said I really doubt this would be Navi 10 cut down.

Keep waiting for AMD to release the Kraken, but they kept on releasing shrimp.

TurmaniaIf they can get 1660 super performance with Consuming around 125W and pricing around the same. Then it can be a great product.

the power doesn't matter that much.. cost does.

We don´t know if the memory-interface is now tied with the ROP-count.
If so, a 6G variant will also need to go down to 48ROP, wich is probably only possible when 1 whole array gets shut down, including the prim-unit etc.
If that chip is a cut-down Navi10, well you get 3 times 5 WGP (Dual-Compute-Unit) leading to 1920 Cores, maxed out.
We will se if thats true, next year.

TurmaniaIf they can get 1660 super performance with Consuming around 125W and pricing around the same. Then it can be a great product.

They can't. 5500 XT already has a 130W TBP according to leaks.

AlienIsGODIsn't just about any gfx card that's not the top end a cut down version though

That is one way to look at it. More accurately, to me, is when it uses the same GPU as others but is 'cut down' down there. For example, the same Navi 10 GPU with less CUs enabled. I don't call Navi 14 XTX cut down because it isn't Navi 10 but a full SKU different.

I really don't see many people buying these cut down cards unless the price is more affordable than other cards which are already available with similar specs, speed, features, ect....

Lets wait for reviews & an official launch though :p

bonehead123I really don't see many people buying these cut down cards unless the price is more affordable than other cards which are already available with similar specs, speed, features, ect....

Lets wait for reviews & an official launch though :p

Why not? Through generations of GPUs people have been buying cut down models.

e: And since it's AMD, it wouldn't be a surprise if the first 6GB models are 8GB physically, just artificially slowed down via bios.

Regarding efficiency, AMD is fighting more against Pascal then versus Turing.
They where braindead addicted to the mining-hype and did nearly nothing to make any big efficiency gains.
So you fight in 2019/2020 versus the efficiency of Nvidias chips of 2016, thats a pitty if you asked me.

_FlareThey where braindead addicted to the mining-hype

Yeah, and that's how their sales and stock price crashed! Oh, wait...

_FlareThey where braindead addicted to the mining-hype and did nearly nothing to make any big efficiency gains.

I think that the only braindead thing here is your assertion. Not only that the mining boom ended well before Navi was finalized but this architecture would eventually end up in the next generation consoles where power efficiency is critical, you gotta be really out of touch to think Sony or MS would accept anything but excellent power efficiency. AMD always had hardware around that was capable of achieving great efficiency under specific uses case, contrary to the popular belief most of you ill informed gamers or whatnot hold. The world isn't just desktop cards.

You gotta realize at some point that there is a difference between an architecture which is inefficient at it's core and a GPU which is inefficient. Fermi was a great example of an architecture that was very inefficient in all it's forms, Vega (the architecture), for instance, wasn't inefficient shown by the fact that you can have it into APUs that sip power while Vega 56 and 64 (the GPUs) were indeed very inefficient GPUs.

I hope this card will deliver and be a realistic if not better option than 1660 Ti/Super in both pricing and performance. 'one can hope' :)

I'm looking for an upgrade from my RX 570 in that price range but so far theres no alternative,and I would rather not buy used this time since I want to keep the new card for years.

Much disappointed with RX 5500 XT and also without hype for a RX 5600, My hype is only for RDNA 2.0, Ampere and Intel XE, this generation of video cards is terrible, worst generation in price per performance.

_FlareRegarding efficiency, AMD is fighting more against Pascal then versus Turing.
They where braindead addicted to the mining-hype and did nearly nothing to make any big efficiency gains.
So you fight in 2019/2020 versus the efficiency of Nvidias chips of 2016, thats a pitty if you asked me.

Yeah not great times, especially seeing as they have already played the 7nm card.

I think it's interesting that people don't seem to realize that AMD split the Radeon tree.

The Fury X which is the basis of Polaris, was aimed to do games and not maths. Navi is currently based on this approach and makes it meh at mining. It's also the Nvidia approach for basically everything but Volta really. NV gaming cards get strapped with brutal FP32 to FP64 dividers.

The R290 X which gave rise to the Vega, was more meant for professional stuff, and had great maths. Great for professional stuff, meh for games currently, unless you actually use the maths for RTRT or something. It's great for mining and maths.

R9 290X smacks a Fury X at FP64 and mining. The Fury X desperately needed 8GB Vram though.

Fluffmeisterespecially seeing as they have already played the 7nm card.

7nm DUV only.
7nm EUV brings more, at least, according to declared spec.

gamefoo21I think it's interesting that people don't seem to realize that AMD split the Radeon tree.

I think we get it, but most simply do not care. It was a curious move to bring so much compute when it is marketed as a gaming card. I don't mind a more clear separation of church and state.

medi017nm DUV only.
7nm EUV brings more, at least, according to declared spec.

Hopefully then will it be where the competition is at 12nm. Not that it is far behind now, but still behind even with the process node advantage, details be damned.

medi017nm DUV only.
7nm EUV brings more, at least, according to declared spec.

And let's just hope they don't charge more for it.

EarthDogI think we get it, but most simply do not care. It was a curious move to bring so much compute when it is marketed as a gaming card. I don't mind a more clear separation of church and state.

Yeah, AMD shouldn't have marketed the V2 so heavily on gaming. It's saving grace in my opinion is that the minimum frame rates are noticeably better, but for most people they only care about what it hits at max, and compared to a 5700XT it's not great.

I would love to get my hands on a dual 64CU V20 card, that uses the fancy interconnect to make the two GPUs act as one. Sadly that's huge money.

I mean even the Fury X2 or Radeon Pro Duo, still commands a hefty price tag used and it's crossfire.

Interestingly I think it's a bit telling that the Xbox One/S and PS4 variants use basically a Polaris based GPU and either can't do 4K or seriously struggle with it and need to use 'optimizations'. While the Xbox One X uses a GPU based on the R290/Vega line and as Sony whined 'brute forces 4K'.

I think having piles of compute power may actually be more future proof. Look at Crytek and their RTRT software demo, they used a Vega 56, and got decent results. The V20 core has significantly more computational power even in consumer dress.

FP64 perf: = Shaders/TMUs/ROPs/CUs
V10 - Vega 64: 0.786 TFlops = 4096/256/64/64
Fastest FP64: 0.854TFlops - Water cooled V64
V20 - V2/VII: 3.360 TFlops = 3840/240/64/60
Fastest FP64: 7.373TFlops- Instinct M60

The V2 gets stuck with less compute hardware, and doubled divider at 1:4 vs the pro cards getting 1:2 for FP64. Seems V10 had the FP64 divider couldn't go past 1:16.

For comparisons sake the fastest Navi GPU and the fastest 2080 Ti...

5700XT PC Liquid Devil: 0.662TFlops
2080 Ti Zo Amp Extreme: 0.494TFlops

So in summary the V20 stuff is meant to crush numbers very quickly. Too bad AMD locked the bios so can't try to flash unlock the V2 cores into fully functional ones like in the past.

I will say though the consumer air cooler for the V2 is probably the best stock air BBA Radeon cooler ever and the 50th AE one just looks sexy to me.