Anyone else sitting this round out?

[fleshconsumed]

I know, right? AMD whiffed on that market segment. Maybe they'll get it right next time around. Heh.

If they're releasing big NAVI in 2020, then the next midrange iteration is going to come in 2021. That's another two years, that's a long time to wait.

 

[darkswordsman17]

If they're releasing big NAVI in 2020, then the next midrange iteration is going to come in 2021. That's another two years, that's a long time to wait.

Navi 20 could be early 2020 (like Radeon VII was), leaving them plenty of time for new mid-range GPU rollout in the 2nd half of the year. AMD has Arcterus slated for 2020 on their roadmaps (although I think they just call it Next Gen, I believe it was referred to as Arcterus on their roadmaps at one point, but there's probably been a lot of shifting of things happened in the last couple of years).

I personally think there's a decent chance it might get pushed to early 2021, but it could still easily beat 2 year time by announcing at CES 2021 and launching not long after (if it launches in Jan 2021 it'd be 1.5 years from Navi) or in the Spring (March or April and it'd be like 1.75 years since Navi). And it could be part of a big change to AMD's GPU setup. By that I mean I think their pro/enterprise markets go mGPU with chiplets and I/O, while their consumer stuff goes to like a 3 tiered traditional setup (where they have 3 main GPU die that they have produced, with like 2-3 versions of each via binning and fusing).

I have very little idea what AMD is planning GPU wise. In many ways it seems like they're still figuring it out, but they also seem to have some plan/idea in place that with the release of Navi they're now implementing, and will be iterating on. Dr Su said that "it must be so" with regards to splitting consumer/gaming GPU development from the pro-enterprise/compute development, but what she meant (i.e. very different base GPU designs, or did she mean one going chiplets) is far from clear. She doesn't seem high on mGPU for gaming right now though, which makes sense as pro/compute stuff scales a lot better with mGPU than gaming does. I thought Navi 20 would be the big compute heavy version of Navi, but it kinda sounds like they're gonna roll with Vega for compute stuff for some time. It seems like Navi cut back on compute performance quite a lot, although I'm not sure if that's at the chip level or is more superficial lockout (much like how AMD and Nvidia have limited other than single precision performance of their consumer GPUs), meaning that a bigger Navi chip could be a compute monster as well. I think its the former (although I think they kept a lot of the compute ability that games utilize).

I don't know if they could make up that compute by doing non-GPU chiplets (so say they integrate Tensor processor; the way Lisa talked almost makes me think they'll let companies roll their own AI chips and then maybe integrate them on package alongside AMD GPUs). Which makes me wonder about implications for gaming as well (could they do a dedicated ray-tracing processor separate from the GPU?).

Something I'm interested in finding out, is AMD taking all the input from all the different companies and trying to develop the fewest GPU designs to try and meet them as best they can by optimizing the base GPU design and then scaling up and down as the need calls for, or are they going further than even before and specializing even more (so Google will slap their TPU alongside an AMD GPU, Sony and Microsoft have unique ray-tracing acceleration bits from each other, they have a significantly different GPU design in store for their Samsung deal, etc). I can almost see it being a mix of both, where they focus on making an efficient base GPU that can scale, while then pairing that with specialized bits for the bigger customers that can afford it - i.e. Google, Microsoft, Sony, etc).

 

[darkswordsman17]

I picked up a 2060 super. Navi still has some maturation to do before I buy another one. It will be the red devil limited. The reference was a bad time aka too much of a time sink and hassles. Super cost me $400 shipped, 2 games, good overclocker, cool and quiet. People obsess about value from PC parts in a way I rarely see outside of the industry.

The 2060 super has already provided me with 50hrs+ of fun. Opportunity cost FTW, as that is 50hrs I will never get back and life is too short to sit around waiting for the next BBD. Not including the hours wasted trying to get Navi going. As tech jesus opined, it has quality of life issues that need addressing before buying it again. I used to be on that merry-go-round of over analyzing price performance power etc, but I finally jumped off some time back. That does not mean I throw a bunch of cash at hardware, but my spending philosophy has definitely changed, for the better I think.

What issues is Navi having? I figure we'd be getting deluge of stuff about that if it were having even Polaris level issues?

I think people have been obsessing over it more, but its because the pricing has made them want to vote with their wallets mostly. Its less that they'd have a real issue, but the gains have been diminishing while prices have been going up. People have bought $400 graphics cards over the years without as much consternation, but generally its because that was doubling performance from like the year before while bringing pretty big graphics changes relatively quickly. Its harder to discern the graphics improvements as the baseline has improved so much (and many games, lots of indie stuff and then Minecraft, have shown graphics aren't everything and many people find things other than the most graphically impressive shooter).

At the same time, we don't have awesome 4K+ VR headsets readily available to make people want to upgrade and go "I'm willing to spend a couple hundred extra dollars as I'll be getting more than my money's worth from everyday play." We're kinda in a stagnant period, where things are good enough, but we also see there's big things just over there but knowing the current stuff won't get us there.

 

[VirtualLarry]

What issues is Navi having? I figure we'd be getting deluge of stuff about that if it were having even Polaris level issues?

OpenCL / mining / DC issues. May be wise to "steer clear" at this time. RX 5700 cards coming up with different results than other, older, prior cards, but consistently. Projects with a quorum of 2, are getting RX 5700 WUs accepted, if their wingman is also running an RX 5700, otherwise, completed but invalid WU, for some reason. Very odd. Hope it's not like the Pentium FDIV bug. It's also not like AMD is promoting NAVI for compute applications, either, which may be disturbing on another level.

I know that personally, I'm looking at getting an RTX 2070 Super myself now.

 

[DrMrLordX]

@VirtualLarry

Looks like Navi isn't a compute card. Hardly disturbing. You know what to get for compute, though . . . Radeon VII! Good luck finding one.

 

[DAPUNISHER]

What issues is Navi having? .

In addition to what Larry noted, there are quality of life issues too. I sent mine back because of media playback issues, wattman and aferburner stopped working so I could not control voltage and clocks to address overheating and crashing issues. Not too mention the hair dryer level noise.

Some have already been tackled, others remain. I bought launch day, so not surprising to have problems. But the overheating and crashing, with no way to stop it, were game stoppers. I was unaware of the OCN tool for the Radeon VII at the time. Which appears to provide the control I would have needed. Being forced to manually tweak simple things like Youtube and othe media playback was annoying, and had to be done every time I did a fresh win10 trying to solve the other problems.

 

[Fanatical Meat]

What issues is Navi having? I figure we'd be getting deluge of stuff about that if it were having even Polaris level issues?

I think people have been obsessing over it more, but its because the pricing has made them want to vote with their wallets mostly. Its less that they'd have a real issue, but the gains have been diminishing while prices have been going up. People have bought $400 graphics cards over the years without as much consternation, but generally its because that was doubling performance from like the year before while bringing pretty big graphics changes relatively quickly. Its harder to discern the graphics improvements as the baseline has improved so much (and many games, lots of indie stuff and then Minecraft, have shown graphics aren't everything and many people find things other than the most graphically impressive shooter).

At the same time, we don't have awesome 4K+ VR headsets readily available to make people want to upgrade and go "I'm willing to spend a couple hundred extra dollars as I'll be getting more than my money's worth from everyday play." We're kinda in a stagnant period, where things are good enough, but we also see there's big things just over there but knowing the current stuff won't get us there.

Well said

 

[Ranulf]

I've learned from the issues surrounding Polaris' launch to wait 6-12 months on AMD video cards. If the prices were $50 cheaper, I could see taking a risk on what is a polaris replacement card but like nvidia turing cards wants to pretend to be something more and thus worth the extra money.

 

[DrBoss]

Yup, waiting on the next generation of Navi. If they can realize gains with their new architecture similar to how they've done with Ryzen, the next iteration of Navi is going to be solid.

 

[fleshconsumed]

Yup, waiting on the next generation of Navi. If they can realize gains with their new architecture similar to how they've done with Ryzen, the next iteration of Navi is going to be solid.

People said same thing when Vega came out - going to wait till next generation (Navi). Now it's next generation after Navi?

I used to upgrade videocard semi-regularly just for the fun of it because I was getting solid 50-75% performance increase every few years at same affordable price points. This is no longer the case. Given the video card market right now I think the most reasonable choice is to stick with what you have for as long as you have so long as you can play whatever games you have without any issues. That's where I am at least.

And since I can't have fun with videocards I'm having fun upgrading my desktops with Ryzen instead. Just upgraded my HTPC from Intel i3-4170 to AMD 2400G this weekend. That was my last intel rig, after this upgrade all of my 5 desktops are officially running AMD chips. Exciting times on CPU front. Not so much on video card side.

 

[Fanatical Meat]

People said same thing when Vega came out - going to wait till next generation (Navi). Now it's next generation after Navi?

I used to upgrade videocard semi-regularly just for the fun of it because I was getting solid 50-75% performance increase every few years at same affordable price points. This is no longer the case. Given the video card market right now I think the most reasonable choice is to stick with what you have for as long as you have so long as you can play whatever games you have without any issues. That's where I am at least.

And since I can't have fun with videocards I'm having fun upgrading my desktops with Ryzen instead. Just upgraded my HTPC from Intel i3-4170 to AMD 2400G this weekend. That was my last intel rig, after this upgrade all of my 5 desktops are officially running AMD chips. Exciting times on CPU front. Not so much on video card side.

So true on the middle part.
I used to look forward to a card upgrade, $250 give or take would typically yield a solid upgrade. Now appears the entry for gaming is $250 and mid range is $350 give or take.
Not so exciting, as I said earlier I’m seriously thinking of sitting on my 280x because spending $250-$300 to basically double performance just isn’t something I’m looking to do. I expect more from a card that is around 6 years more current.

 

[JustMe21]

I'm guessingthat Nvidia will release 7nm GPUs in 2020, so why buy this year. I would expect better RayTracing performance in the next gen as long it's not a die shrink/clock speed increase refresh of the current gen.

 

[RadiclDreamer]

I'm guessingthat Nvidia will release 7nm GPUs in 2020, so why buy this year. I would expect better RayTracing performance in the next gen as long it's not a die shrink/clock speed increase refresh of the current gen.

Im really concerned that Nvidia is going to pretty much sit idle until they have a real competitor in the higher end graphics market. I really want AMD to put the fear back into them.

 

[DrMrLordX]

Im really concerned that Nvidia is going to pretty much sit idle until they have a real competitor in the higher end graphics market. I really want AMD to put the fear back into them.

That's what Big Navi(tm) is supposed to be: something to beat the 2080Ti in raw performance. It might even support RT but we don't know.

 

[Glo.]

That's what Big Navi(tm) is supposed to be: something to beat the 2080Ti in raw performance. It might even support RT but we don't know.

Im sure it will also dance and sing, at this point of the hype about it.

 

[DrMrLordX]

Im sure it will also dance and sing, at this point of the hype about it.

All nonsense aside, the 2080 Ti is not exactly a moving target. Maybe NV will update the product with a Super version of the card, or maybe they won't. But AMD has had plenty of time to take shots at it and the 1080 Ti (they haven't officially beaten either card even if Radeon VII sometimes manages to eclipse the 1080 Ti. Sometimes). It looks like they'll have another shot between now and at least Q2 2020 to beat the 2080Ti. Simply shipping Navi with a higher CU count may do the trick. Moving to 7nm+ would also help.

 

[railven]

As long as I can work and afford a generation, I'll never sit a round out! What I might sit out is top enthusiast cards if Nvidia continues to think $1300 is sustainable.

I'd happy throw my money at AMD if they can give me 90% of NV's performance at $1,000 (assuming next NV halo card sans secret-sauce cards is still $1300).