AMD RX Vega AIB Cards to Ship in Late July / Early August

[Captain_Tom]

Did I strike a nerve? Check my system specs to see where I am coming from

A "nerve"? I wasn't mad lol.

Why? - You mad bro?

 

[Captain_Tom]

The 1070 is about as fast a s a 980ti. We know an overclocked nano will match a Fury X which we know matches a 980ti. At 4k, a 1080 ti is substantially faster than a 980ti.

By Vegas release date, Volta will be looming so it's academic. I know you have been very pro AMD in almost, in fact every post you make but your unbridled enthusiasm has to meet reality at some point. Fury X had 8.5 T/flops of performance and matches a 980ti. That card only had 5.63T/flops. Just because the Frontier edition has 13 T/Flops doesn't mean much in the actual reality of performance. We'll all need to wait and see but going on past performance, AMD stomps Nvidia in compute (T/Flops performance) but it doesn't match it in real world gains. Yes, they have a new arch (they dont, they've added stuff to GCN, it's not new yet) but a new arch guarantees nothing.

I not trying to rain on any parades (why would I have bought Ryzen?) but I am genuinely amazed how much faster the 1080ti is to the 980ti. Considering all the misguided people that say the 10xx series is just a refresh and it's boring, well you know it is really freaking fast. Stock for stock a 1080ti is 82% faster than a 980ti. No matter how you spin it, it's very impressive. Vega needs to work absolute, R&D busting miracles to be better than it.

Yeah I am well aware of what TPU considers worth benchmarking. Call me crazy but I don't include Anno-nobody-plays-this, and broken Gamesworks games in my totals. You're cute though calling me a fanboy for simply reporting results. To clarify what I found:

1. Fury Nitro is clocked at 1100/518 (Performs almost completely identical, if not 1% stronger to an R9 Fury X). The Nano was clocked at 1055/655 (Not a typo). These are my 24/7 stable clocks, they can go higher, but I find "Benchmarking clocks" to be stupid. People should only bench stuff that's 100% stable.
2. Results:
   
   
   
3. This is an average over 1080p, 1440p, and 4K benches. Results do vary depending on res.
4. At 1080p, the Fury cards were ahead of the 1070 by 2-3%. This lead increased to a 5% lead in 1440p, and in 4K they were STILL ahead of the 1070 by ~1%.
5. The 1080 was always stronger than the Fury cards, but in 1440p the 1080 was only ~10% stronger.
6. In general the 1080 Ti was kind of a waste of a card imo. It was indeed ~20% stronger than the 1080/Nano in 1080p, but at 1080p every card had no trouble at all staying ahead of 120Hz. In 1440p the 1080 Ti definitely pulled ahead by a fair margin, and in 4K the 1080 Ti was a full 70% stronger than the Overclocked Nano. However the 1080 Ti only got an average 4K framerate of 57 FPS. That means minimums were in the 40's. and imo unnacceptable in 2017. So even if you pay $700 for the "mighty" 1080 Ti, it will still required lowered settings for 4K gaming.
7. An interesting final note - in Metro LL 4K the Overclocked Nano (With 671 GB/s of bandwidth lol) was 12% stronger than the 1080. Not playable though, I should have tested it without SSAA.

P.S. Not sure why you are mentioning the 980 Ti. That card is being matched by the 390X in a lot of recent releases. The Fury X is a full 15-25% Stronger than the 980 Ti. It is no longer 2015 buddy.

 

[Deleted member 172152]

AMD better be working on their next arch already due to all these delays, plus they only said 1080, not 1080Ti.

But AMD was already comparing engineering samples at much lower frequencies with fiji drivers to 1080's. Better drivers and it won't need vulkan, plus a probable boost in clockspeed, something like 1200-1600mhz would be my guess, really should get vega in 1080 ti territory.

 

[bogami]

So slightly above the GTX1080, I doubt it already for 500 €. To 1200 € of the cheapest model that will not reach these results, it's better to afford the 2x GTX 1080, and the SLI is here which is surely much stronger.
Not only crazy delays, little OC to double the door on the old chip new RAM and this is the new VEGA. The price will definitely kill VEGA. I mean what they praise. $ 1200 for a shady model that will not reach, I prefer to buy 2xGTX1080 11Gb / s, SLI .

 

[Deleted member 172152]

So slightly above the GTX1080, I doubt it already for 500 €. To 1200 € of the cheapest model that will not reach these results, it's better to afford the 2x GTX 1080, and the SLI is here which is surely much stronger.
Not only crazy delays, little OC to double the door on the old chip new RAM and this is the new VEGA. The price will definitely kill VEGA. I mean what they praise. $ 1200 for a shady model that will not reach, I prefer to buy 2xGTX1080 11Gb / s, SLI .

Guess you would have voted or have voted Trump.

1200 dollars is for a workstation gpu and is actually relatively cheap. Rx vega will be half or less than half the price with better cooled AIB's, higher clocks and more game optimizations. Rx vega 64 can perform anywhere between a 1080 and volta with a competitive price, so vega will be good, better or simply the best.

 

[bogami]

Guess you would have voted or have voted Trump.

1200 dollars is for a workstation gpu and is actually relatively cheap. Rx vega will be half or less than half the price with better cooled AIB's, higher clocks and more game optimizations. Rx vega 64 can perform anywhere between a 1080 and volta with a competitive price, so vega will be good, better or simply the best.

I do not need WS ,gamer , and even nVidia has CUDA support, the difference is only for one application, which can also be performed, only slower. Indeed, a cheap WS card is an excuse for an abnormal price. This will be sold very little !...

 

[Deleted member 172152]

I do not need WS ,gamer , and even nVidia has CUDA support, the difference is only for one application, which can also be performed, only slower. Indeed, a cheap WS card is an excuse for an abnormal price. This will be sold very little !...

What? Could you maybe use full sentences?

So rx vega will be sold for little, you don't need a workstation or gaming gpu, nvidia has CUDA cores and CUDA only matters for a very specific application, which non-CUDA cards can do, but relatively slow?! If that's what you meant, I completely agree with you,althoigh I'm not an expert on CUDA.

 

[Prima.Vera]

Too little too late. After 1 or 2 months nVidia will trash them with the 20xx series....

 

[Deleted member 172152]

Too little too late. After 1 or 2 months nVidia will trash them with the 20xx series....

Gddr5x or gddr6 though, vega might be great, volta seems to have a relatively small increase in performance, volta will probably be more expensive and geforce volta is likely only going to arrive next year.

 

[efikkan]

Why is nvidia card not lasting long ? Anyone own have experience owning nvidia high end gpu for more than 4 years ? Do share your views..

CPU I am adamant that I won't pay big fat blue. No matter what. But graphics card I want to go nvidia just for the fps reason but not able to convince myself after having bad experience with 7600 and later 8800GTX XFX.. both got burned and they never replaced it and both only lasted 2 and 1.6 years..

Nvidia cards do last. I have both GTX 580 and GTX 680, still in daily use with programming and gaming.
It usually comes down to the quality of the coolers. Asus and MSI are usually good, various other brands are more iffy.

Remember to clean them regularly though, dust is the enemy.

 

[I No]

Yeah I am well aware of what TPU considers worth benchmarking. Call me crazy but I don't include Anno-nobody-plays-this, and broken Gamesworks games in my totals. You're cute though calling me a fanboy for simply reporting results. To clarify what I found:

1. Fury Nitro is clocked at 1100/518 (Performs almost completely identical, if not 1% stronger to an R9 Fury X). The Nano was clocked at 1055/655 (Not a typo). These are my 24/7 stable clocks, they can go higher, but I find "Benchmarking clocks" to be stupid. People should only bench stuff that's 100% stable.
2. Results:View attachment 89277
   
3. This is an average over 1080p, 1440p, and 4K benches. Results do vary depending on res.
4. At 1080p, the Fury cards were ahead of the 1070 by 2-3%. This lead increased to a 5% lead in 1440p, and in 4K they were STILL ahead of the 1070 by ~1%.
5. The 1080 was always stronger than the Fury cards, but in 1440p the 1080 was only ~10% stronger.
6. In general the 1080 Ti was kind of a waste of a card imo. It was indeed ~20% stronger than the 1080/Nano in 1080p, but at 1080p every card had no trouble at all staying ahead of 120Hz. In 1440p the 1080 Ti definitely pulled ahead by a fair margin, and in 4K the 1080 Ti was a full 70% stronger than the Overclocked Nano. However the 1080 Ti only got an average 4K framerate of 57 FPS. That means minimums were in the 40's. and imo unnacceptable in 2017. So even if you pay $700 for the "mighty" 1080 Ti, it will still required lowered settings for 4K gaming.
7. An interesting final note - in Metro LL 4K the Overclocked Nano (With 671 GB/s of bandwidth lol) was 12% stronger than the 1080. Not playable though, I should have tested it without SSAA.

P.S. Not sure why you are mentioning the 980 Ti. That card is being matched by the 390X in a lot of recent releases. The Fury X is a full 15-25% Stronger than the 980 Ti. It is no longer 2015 buddy.

Dude seriously where in the world are you pulling those numbers from? Fury X 15-25% stronger than a 980 Ti? The 1080 Ti a waste of a card? So at this point what's option B ? You did forget that nVidia has the high-end on lockdown for the past year with AMD deciding to play around with the mainstream market (oh and they failed at this as well) .

By the numbers you got posted there Fury is a full 15-25% faster than a 980 Ti right? "This lead increased to a 5% lead in 1440p, and in 4K they were STILL ahead of the 1070 by ~1%." (Your numbers) vs "The Fury X is a full 15-25% Stronger than the 980 Ti" how's this possible ? Just asking out of shear curiosity. /s

"So even if you pay $700 for the "mighty" 1080 Ti, it will still required lowered settings for 4K gaming. " -> Easy. Because there is no alternative.

"Call me crazy but I don't include Anno-nobody-plays-this, and broken Gamesworks games in my totals."
Those "broken" Gameworks titles actually sell hence why they are important, people want to play those games. Now that would be a thing right? To bench a card through everything that's AAA and what's actually demanded by the people. The only thing you are doing is cherry picking the living daylights out of benchmarks. But ofc if your life mission is to play everything that doesn't run Gameworks by all means go nuts but please don't try to present this as "fair" benching.

 

[bogami]

What? Could you maybe use full sentences?

So rx vega will be sold for little, you don't need a workstation or gaming gpu, nvidia has CUDA cores and CUDA only matters for a very specific application, which non-CUDA cards can do, but relatively slow?! If that's what you meant, I completely agree with you,althoigh I'm not an expert on CUDA.

.
But you can read ? and iNvidia Use directCompute5.0 .

 

[Deleted member 172152]

.
But you can read ? and iNvidia Use directCompute5.0 .

Yes, I can read and I use proper english not "threadripper rubbish, threadripper cheap but hedt so warrant expensive, vega something, nvidia best". Just use full sentences and I might know what on earth you're trying to say without having to spend ages deciphering your comments.

Now, why is directcompute 5.0 so important?

 

[TheGuruStud]

You did forget that nVidia has the high-end on lockdown for the past year with AMD deciding to play around with the mainstream market (oh and they failed at this as well) .

Yeah...no one's gonna take you seriously with misinformation like that. Even before the fresh mining craze AMD was selling 30% of the market with JUST 460/470/480 LOL. That's a shit load of cards for so few SKUs and all on the lower end of the market (which we know the higher is selling very well, nearly as much as lower).

 

[xenocide]

Yes, I can read and I use proper english not "threadripper rubbish, threadripper cheap but hedt so warrant expensive, vega something, nvidia best". Just use full sentences and I might know what on earth you're trying to say without having to spend ages deciphering your comments.

He has said several times since joining that English isn't his first language. But please, berate him for trying.

 

[I No]

Yeah...no one's gonna take you seriously with misinformation like that. Even before the fresh mining craze AMD was selling 30% of the market with JUST 460/470/480 LOL. That's a shit load of cards for so few SKUs and all on the lower end of the market (which we know the higher is selling very well, nearly as much as lower).

https://jonpeddie.com/press-release...in-q117-from-last-quarter-with-nvidia-gaining


Last time I checked losing market-share while boasting that you have 1 segment locked-down as your niche is still failing. This is before the mining craze mind you. Also this current situation will tone down in a while like it did last time. This time the difference is nVidia is at it too. We can argue till the second ice age hits. Fact is AMD still doesn't get close to nVidia's earnings no matter the segment they attack. From my point of view targeting mid with competition and leaving the high-end to whomever wants to grab it it's just like handing them free money. Seeing that the 1060 3/6gb is giving them a run for their money well enough by now it doesn't even matter if the 460/470/480 outsell their counterparts by 10 to 1 nVidia would still rake in more money due to having 0 competition within a segment. This would be a fail in my books.

 

[Deleted member 172152]

He has said several times since joining that English isn't his first language. But please, berate him for trying.

I didn't know that, and asking me if I can read, even though it isn't his first language and my english, although not perfect, is at least understandable, isn't very nice either. So you shouldn't berate me either. I'm just not going to back down after a somewhat hurtful comment like that.

That's basically everyone's online english today anyway, so there was no way for me to know english isn't his first language and I just gave him friendly advice to use complete sentences, not just separate words so we could have a more fluent conversation.

 

[xenocide]

I didn't know that, and asking me if I can read, even though it isn't his first language and my english, although not perfect, is at least understandable, isn't very nice either. So you shouldn't berate me either. I'm just not going to back down after a somewhat hurtful comment like that.

That's basically everyone's online english today anyway, so there was no way for me to know english isn't his first language and I just gave him friendly advice to use complete sentences, not just separate words so we could have a more fluent conversation.

He asked "But you can read?" as a way to make sure you understood what he was getting at. You were approaching it as though it were a native speaker in which case that would be borderline insulting--"can you even read?"--but he was just verifying you understood what he was getting at. And if you want to get into insulting comments, maybe don't take the moral highground after negatively accusing someone online of being a Trump supporter--in a way that implies they are stupid. Kind of hinders your argument.

 

[Prima.Vera]

Anyways a lot of rumors are saying that nVidia will also launch the new generation by the same time, or 1-2 months late. AMD really needs to compete with those cards too, otherwise the future is very grim for them...

 

[Deleted member 172152]

Anyways a lot of rumors are saying that nVidia will also launch the new generation by the same time, or 1-2 months late. AMD really needs to compete with those cards too, otherwise the future is very grim for them...

Those rumours are likely rubbish. Don't expect volta to a) be as great a leap as maxwell to pascal (did the mathspeculation and even with dx12 optimizations, it's likely there will be e mere 15% increase in framerates) and b) to release before march 2018 or so.