[VC]AMD Radeon R9 390X WCE Speculation Thread

[wand3r3r]

Well, May is only a bit over a month away. Was the latest rumors stating a complete lineup in May, or was it June?

So many rumors. I can only hope they nail it with an 8GB HBM titan beating $550 390x. I'll believe it when I see it though.

 

[krumme]

What do we know that says 390x is not a dual gpu solution? - and perfect drivers is not just bringing cf to new level.

 

[AtenRa]

Well, May is only a bit over a month away. Was the latest rumors stating a complete lineup in May, or was it June?

So many rumors. I can only hope they nail it with an 8GB HBM titan beating $550 390x. I'll believe it when I see it though.

Dont expect a 8GB HBM WC 390X to release bellow what 290X was originally released. Especially if it will be faster than Titan.

 

[DownTheSky]

390X is single GPU, we know some basic specs, some guidelines for performance > 290X+50%, and price >700$.

 

[AnandThenMan]

I think the sweet spot price for the 390x would be $649. High but not too high, and more than low enough to be a punch in the gut to the Titan X.

 

[AtenRa]

I think the sweet spot price for the 390x would be $649. High but not too high, and more than low enough to be a punch in the gut to the Titan X.

Im expecting higher than that for the WC edition.

 

[xthetenth]

In this thread, people argue that a solution that is objectively proven to result in lower temperatures, lower noise, with a higher power consumption GPU is somehow worse.

What's next guys, how water cooling isn't a "premium design" like the Titan cooler because it doesn't use aluminum?

If you don't want a higher performing solution, you're not an enthusiast. Watercooling has long been very effective but expensive. Now it's just very effective.

Water cooling isn't a luxury option, clearly. This is of course because it delivers price for performance, while a luxury option delivers a feeling of satisfaction in return for price, with performance and price disconnected.

I'm starting to wonder how many people in this forum secretly have a case with only an 80mm exhaust and cover their case in a towel though. 120mm exhausts shouldn't be hard to come by, I've got a nice clear one despite a huge NH-D15 hanging off my socket, and if I switched that to a 200+mm rad for my CPU I'd have the back exhaust slot.

I can't wait, if current "high end" products hold their price, I think an upgrade might be in order.

 

[theeedude]

You guys are mean.
AMD just had to eat the R&D cost of HBM1, and as the primary HBM1 customer, it will have to eat much higher memory prices for sole supplier SK Hynix to amortize its HBM1 costs over very low volume, and it will have to eat the cost of an expensive cooling solution to deal with energy densities of a big die GPU in close proximity to stacked DRAM.
And after knowing all that, and how desperate AMD is to make some money, you just want to give it a measly $550 or even $700, knowing that a huge chunk of that will just go directly to third party suppliers, and not AMD's gross margin?
Is that fair? Do you expect AMD to not make any money at the top end? And if it doesn't make any money at the top end, how do you expect it to be around to invest in R&D and spend money on driver support for your product?
Be reasonable, folks.

 

[AnandThenMan]

nm

 

[Subyman]

The only issue I see with an AIO isn't with enthusiasts, it with OEM. If the reference design includes AIO, then that could cause problems with people buying these for workstations, SFF, older Mac Pros, or companies like Dell/Alienware/Origin. It may not be a large chunk of people, but it does make the card more niche. No doubt about it. Straying from industry standard design will exclude some sales. Curious to know how it will affect sales, especially if they start doing AIO on anything lower than the high end.

 

[garagisti]

The only issue I see with an AIO isn't with enthusiasts, it with OEM. If the reference design includes AIO, then that could cause problems with people buying these for workstations, SFF, older Mac Pros, or companies like Dell/Alienware/Origin. It may not be a large chunk of people, but it does make the card more niche. No doubt about it. Straying from industry standard design will exclude some sales. Curious to know how it will affect sales, especially if they start doing AIO on anything lower than the high end.

ATI/ AMD have set a fair few standards in GPU world, just consider it one such, why not?

 

[AtenRa]

The only issue I see with an AIO isn't with enthusiasts, it with OEM. If the reference design includes AIO, then that could cause problems with people buying these for workstations, SFF, older Mac Pros, or companies like Dell/Alienware/Origin. It may not be a large chunk of people, but it does make the card more niche. No doubt about it. Straying from industry standard design will exclude some sales. Curious to know how it will affect sales, especially if they start doing AIO on anything lower than the high end.

Most of those PC boutiques are already using CPU AIOs with big Cases. I dont believe they will have any trouble fitting another one for the GPU. The only problem i see is if they will like to install 3 or 4 GPUs. But up to 2x 390X AIOs will be easily installed in those big towers.

 

[AnandThenMan]

Unless AMD outright prohibits board makers from going air cooling which I highly doubt, we will see such versions.

 

[Kippa]

I was seriously thinking of getting the Titan X, basically for two reasons. The first was for gaming at 4K and more importantly the second was for CUDA in 3d rendering, specifically Octane. One interesting development with Octane is that they're going to release an OpenCl version in version 3 later on in the year which means that I'm not tied down to Nvidia anymore, and price wise AMD looks more appealing. I am wondering if there is a slow shift going from CUDA to OpenCl in general. Either way I am definetly NOT getting the Titan X now and will wait to see the price and performance of the AMD cards. Interesting times ahead. I wonder if other devs will start shifting to OpenCl now that it is becoming more mature.

 

[xthetenth]

Actually, a question for anyone who knows about the cooler on the 295X, do you think that I'd be able to swap out the fans on the rad for that cooler, and if so would that sort of thing give a noise/cooling boost like I've heard it does with things like the Corsair AIOs?

 

[Headfoot]

I was seriously thinking of getting the Titan X, basically for two reasons. The first was for gaming at 4K and more importantly the second was for CUDA in 3d rendering, specifically Octane. One interesting development with Octane is that they're going to release an OpenCl version in version 3 later on in the year which means that I'm not tied down to Nvidia anymore, and price wise AMD looks more appealing. I am wondering if there is a slow shift going from CUDA to OpenCl in general. Either way I am definetly NOT getting the Titan X now and will wait to see the price and performance of the AMD cards. Interesting times ahead. I wonder if other devs will start shifting to OpenCl now that it is becoming more mature.

It might be worth waiting to see what the actual performance of the OpenCL code path is. Many times it ends up being less performant than the CUDA path because the devs have less experience coding for OpenCL. But, its becoming more common now so it might end up being pretty good. I'd wait and see

 

[Headfoot]

Actually, a question for anyone who knows about the cooler on the 295X, do you think that I'd be able to swap out the fans on the rad for that cooler, and if so would that sort of thing give a noise/cooling boost like I've heard it does with things like the Corsair AIOs?

If the WCE is reasonably priced, and I decide to get one, I'd set it up push-pull for sure. You can get another 4-8c from the second fan at least on CPU AIO's

 

[Vesku]

I dont think anyone is questioning the performance of an AIO CLC. But time and time again, many people have already pointed out why its a bad idea to have one as a reference design especially from a business perspective unless its a niche product like the R295X where it actually makes sense.

Perhaps the GM200 is pushing the blower cooler to its limits.. It would have been wise for them to go with a Titan Z type open air cooler (e.g. 7900GTX - central fan) but perhaps theyll do that next time.

Titan X is pretty niche and most likely the same for 390X WCE @ $700+. Also if it is labeled WCE that implies there will be a non water cooled version. The CLC model gives AMD a better chance to claim fastest single GPU with decent temps and noise level which is a pretty valuable marketing crown to hold.

 

[boozzer]

390x would probably be around 700$ I consider that price to be just right.

 

[RussianSensation]

390x would probably be around 700$ I consider that price to be just right.

Same. NV rode the market with 8-15% faster 980 for 6+ months. When overclocked a 980 would beat Max overclocked 290X by about 25%. NV charged 80-90% more for this, but same VRAM. Let's face it, AMD needs money right now more than market share. If you start earning profits, your business is sustainable and you can gain market share in the future with subsequent generations. It's not as if NV's loyal customers will switch anyway at the high end because at the High end, they are extremely loyal to NV. Where AMD can get the most market share gains is mobile dGPUs and low/mid-range desktop. For brand agnostic gamers, even if $699 390X is 95% as fast as a Titan X, it's already a better buy. Moreso, imagine 390 non-X with 85-87% of Titan X's performance for $499-549. It would make a lot more sense to get 2 of those than a single Titan. I would mich rather lose 15% single GPU performance when CF doesn't work and gain 40-50% more performance when it works if spending $1000-1100.

Either way, I don't see why AMD needs to lowball 390X considering even when they delivered 4870/5870/6950 unlocked that all offered 80-85% of NV's flagship, NV users didn't switch. In that case might as well make money off brand agnostic gamers than lowball, and still not get NV loyalists to convert.

Where I think AMD should be very aggressive are $100-400 price segments. If AMD delivers price/performance at all of those price levels, then they do not need to sell a card with near Titan X performance at $549. I would love for it to be $549, but it sounds unrealistic.

 

[3DVagabond]

As far as pricing goes it's pure supply and demand. You charge as much as you can while still selling all that you make/buy.

 

[RussianSensation]

As far as pricing goes it's pure supply and demand. You charge as much as you can while still selling all that you make/buy.

Now that AMD has the benefit of seeing where Titan X sits, extrapolating a cut-down GM200, they can price 390 non-X as the high end "value play" for $499-549. More or less I expect NV to drop 980 to $399-449 after 390 non-X launches. AMD lost first mover market advantage to launch with very high prices - the strategy they used with HD7000. The reason I keep talking about non-X 390 is because 2nd best AMD cards (5850/6950/7950/290 non-X) tend to be knock outs with unbeatable price/performance. I know people are going to be naturally excited for the top 390X card but AMD tends to charge a very large premium for their top card which is often just 5-10% faster than their 2nd best when both are max overclocked.

Basically the non-full yielding 390X has to go somewhere, but AMD has to still use the 4096-bit HBM for it. That's what I like about AMD cards is the memory and ROP side is rarely cut down. That's what allows the 2nd fastest AMD card to overclock to within 90-95% of their flagship. The same doesn't happen for NV very often. In overclocked states 980 retains its 16-19% lead. Similarly you can't really overclock 470 or 780 to 95% of 480/780Ti's overclocked speeds.

That's why I think the most exciting high end card match up might be 390 non-X OC vs. GM200 6GB OC, and 390 non-X OC CF vs. Titan X OC.

 

[Silverforce11]

I haven't gotten AMD's top tier card since forever, since they've never given me good reason to. I up the clocks on the 2nd tier card to match the top tier card and its within 5% performance. That isn't worth the 20-25% premium.

Depending on how its priced, maybe the 8GB HBM is reserved for 390X early on, with 4GB HBM on 390.. at which case, the extra premium will be worth it along with the minor performance gains.

4GB has so little "future-proofing" these days.

 

[AtenRa]

The reason I keep talking about non-X 390 is because 2nd best AMD cards (5850/6950/7950/290 non-X) tend to be knock outs with unbeatable price/performance. I know people are going to be naturally excited for the top 390X card but AMD tends to charge a very large premium for their top card which is often just 5-10% faster than their 2nd best when both are max overclocked.

Yeap, 3850/4850/5850/6950/7950 all of those had the best perf/price in their market segment.
Im also expecting R9 390 (non X) to be close to $499-$549 and give you 85-90% of Titan-X performance.

Basically the non-full yielding 390X has to go somewhere, but AMD has to still use the 4096-bit HBM for it. That's what I like about AMD cards is the memory and ROP side is rarely cut down. That's what allows the 2nd fastest AMD card to overclock to within 90-95% of their flagship. The same doesn't happen for NV very often. In overclocked states 980 retains its 16-19% lead. Similarly you can't really overclock 470 or 780 to 95% of 480/780Ti's overclocked speeds.

That's why I think the most exciting high end card match up might be 390 non-X OC vs. GM200 6GB OC, and 390 non-X OC CF vs. Titan X OC.

Yeap, 3850/4850/5850/6950/7950 all of those had the best perf/price in their market segment.
Im also expecting R9 390 (non X) to be close to $499-$549 and give you 85-90% of Titan-X performance.
Also, R9 380 may be the next perf/price champion (im estimating perf to be between GTX970 and GTX980 at $299 ??)

 

[flopper]

you can buy old tech like titanx or the new tech with amd 390x with hbm and if one games with eyefinity and 4k it seems plausible the new tech are a better fit.