[BitsAndChips]390X ready for launch - AMD ironing out drivers - Computex launch

[2is]

of course it is not fast enough now compare to titan x.

but when it was first release oct 2013, it was the fastest gpu on the market, especially when it is the vapor x version with non of the problems of the reference models.

so 80 to 100% performance increase is not enough for you to upgrade? hahahahahahahaaaaaa, ok I am done

You realize SLI isn't a model number right? It means I have 2 680's. It seems like you aren't familiar with this technology based on your responses. I didn't say anything about Titan X.

EDIT: Just to clear it up... That's 2 680's in the same machine working together to increase performance over a single 680...

http://en.wikipedia.org/wiki/Scalable_Link_Interface

 

[Kippa]

Out of curiosity when are the 14nm based cards due out? Are the first ones roughly January of 2016 or more like mid to late 2016? I'd be happy to wait another 6 months but cba with 12 to 18 month wait.

 

[S.H.O.D.A.N.]

I wouldn't hold my breath for 14/16nm in January.

 

[x3sphere]

There is no way AMD ships a HBM card with more than 4GB in 2015 for anything but the very top SKU. It would be colossally stupidly expensive just to give the extra 512MB that some games need beyond 4GB. It is far more likely that games like GTA V will be optimized and patched to work with 4GB with no performance penalty. There is no reason any reputable game maker should be using more than 4GB at this time, unless you go beyond 4K.

That's going to go over well with NV having majority share in the GPU market. NV will probably encourage devs to include higher resolution textures and push up VRAM requirements as a result, knowing their competition is at a disadvantage.

 

[boozzer]

You realize SLI isn't a model number right? It means I have 2 680's. It seems like you aren't familiar with this technology based on your responses. I didn't say anything about Titan X.

EDIT: Just to clear it up... That's 2 680's in the same machine working together to increase performance over a single 680...

http://en.wikipedia.org/wiki/Scalable_Link_Interface

then get 2x 290x 8gb, what is your point?

ps, I know you didn't say anything about titan x. just pointing out the performance of 290x in oct 2013. and of course it isn't top dog now

pps: I get the feeling that your #1 go to defense is personal attacks, that is just sad bro. get a grip.

I wouldn't hold my breath for 14/16nm in January.

we will probably see leaks in q1/2 2016 and probably a gpu around q3 or 4, just in time for the holidays. but that is just my wishful guesstimate.

 

[Pinstripe]

Out of curiosity when are the 14nm based cards due out? Are the first ones roughly January of 2016 or more like mid to late 2016? I'd be happy to wait another 6 months but cba with 12 to 18 month wait.

Dec 31th 2016, because we can't have the nice things too early.

 

[2is]

then get 2x 290x 8gb, what is your point?

ps, I know you didn't say anything about titan x. just pointing out the performance of 290x in oct 2013. and of course it isn't top dog now

pps: I get the feeling that your #1 go to defense is personal attacks, that is just sad bro. get a grip.

It's not a personal attack. It really seemed like you didn't know what SLI was. We've been going back and forth for a while now, I've mentioned what I have several times it's also in my sig yet you keep spouting 80-100% performance advantage. What else would you have me surmise?

What's sad is letting brand loyalty dumbing down your arguments. There's a lack of comprehension on your part, perhaps it's intentional and you're trying to prove a point, but as with the last back and forth... It's not really working out.

To answer your question, I want to move away from multi-gpu setups. It was great when I first got it but a little less so now that the cards aren't as well supported. I find myself either waiting longer for patches or making manual adjustments in nVIdia control panel to get it working optimally. Not always the case, but it happens more often than it used to, which was practically never. Though I may still keep one of my 680's as a dedicated PhysX card.

 

[Tuna-Fish]

What about latency (timing)? DDR3 has much lower latency than GDDR5, but DDR3 doesn't have as much bandwidth. Does HBM bridge the gap, offering low latency along with the high bandwidth? Is it more like one than the other?

DDR3 memory does not have much lower latency than GDDR5. In fact, they are pretty damn close.

Most GDDR5-using memory subsystems have a much higher access latencies than typical DDR3-based ones. This is not a feature of the RAM, but of the memory controller. Coalescing accesses increases bus utilization, and in GPUs access somewhere in a texture is often followed by other access near it, so in a heavily bandwidth-optimized controller it makes sense to wait a while after receiving a request before issuing it just to see if you want more ram in one go.

 

[RussianSensation]

then get 2x 290x 8gb, what is your point?

ps, I know you didn't say anything about titan x. just pointing out the performance of 290x in oct 2013. and of course it isn't top dog now

pps: I get the feeling that your #1 go to defense is personal attacks, that is just sad bro. get a grip. .

His story has been inconsistent for a long time now but he won't admit it. He makes wild statements like 4GB of VRAM is probably going to be a bottleneck for 1080P soon, while he supposedly has been gaming on GTX680 2GB SLI at 1440P for the last 3 years but his 2GB bottleneck wasn't a problem that bothered him enough? 2GB has been a real bottleneck not some 'hypothetical' bottleneck sometime in the future that seems to be more of a case with 4GB vs. 8GB. Yet, gaming on 2GB with reduced settings is A-OK for years but buying a card with "only" 4GB that might become bottlenecked in 1-2 years is a waste of $? Seriously? Also, his entire GPU upgrading strategy to be frank is outdated. Sure, there will be gamers who buy flagship cards and hold on to them for 3-4 years, but you know what, this strategy itself has proven to be inferior than buying and reselling every 2-2.5 years. Let's face it, most enthusiast gamers who had 780Ti SLI or Titan X SLI owners today upgrade every generation. The types of guys buying $700 cards don't care to future-proof for 3-4 years. The minute a card 40-50% faster is out, they are dumping their old cards because these are an elite group of gamers who always buy the best. The reason someone buys Titan X SLI today is not to future-proof years down the line but to have awesome performance today. If you have a 4K monitor, you need all the performance you can get - that's why you want cards like Titan Xs. Since these gamers are on the cutting edge of tech, they won't hold on to these cards for too long once Pascal launches. In the group of guys who buy $700+ cards in pairs, he is actually a tiny minority who wants to keep his cards > 3 years. Most of them on our forum and OCN do not upgrade like that.

If someone is OK paying $150-200 more per card for minimal performance gains just to future-proof, it's that gamer's choice but he should at least admit that this strategy is inferior to start with. It's been proven mathematically and statistically that buying 2nd tier cards and upgrading more often is superior to buying dual flagships and holding on to them or 3-4 years. He won't admit this is a fact. From that perspective, unless I have a 4K gaming monitor, why the heck would I are if R9 390 or GM200s last me 4 years? If someone wants to waste hundreds of dollars extra to future-proof, let them do it. Plenty of people wasted $ on 680 SLI vs. 670 SLI for e-peen.

If his story added up and he couldn't bare running into VRAM bottlenecks, there is no way he'd be gaming on 680 2GB cards right now. He would have at least sold them a long time ago and gotten R9 290/290Xs or 970s/980s as an intermediary solution assuming his standards are SO high that he can't stand VRAM bottlenecks. Then, if VRAM requirements grew beyond 4GB, he would have sold those cards and gotten GM200s, 14nm, etc. That means for a 100% fact, he games reducing texture and other settings to minimize his own 2GB bottleneck or avoids all titles that require > 2GB. What he preaches in this thread is inconsistent with what he actually does himself. He must have been suffering 2GB VRAM bottlenecks for at least a a year now, but somehow it's not acceptable for someone to save $ and get the 4GB 390 non-X or even 390X version and also reduce some settings to save $? Woot?! (for example, someone might not care to spend $150-200 more just for 15% more performance and 6-8GB VRAM - but he never considers this and constantly evades this point).

If you go check 20-30 modern games tested in the last 15-18 months, ARES II (basically HD7970 1.1Ghz in pairs) crushes GTX690 by 25-70% at 1440P.

This has been consistent especially in AAA games. There have been a giant slew of games that cripple 680 2GB SLI setup specifically because of 2GB of VRAM:

http://gamegpu.ru/images/remote/htt...trategy-Homeworld_Remastered-test-hr_3840.jpg

- Evolve
- Wolfenstein TWO
- Mortal Kombat X
- AC Unity
- Dead Rising 3
- Watch Dogs
- FC4
- DAI
- GTA V
- Shadow of Mordor, etc.

See HD7970Ghz CF (Ares II) has been mopping the floor with GTX690 (aka GTX680 2GB SLI) but he didn't have much of problem with the 2GB VRAM bottleneck that was real and measurable over the course of the last 12 months, not in some "hypothetical future".

Look at this, it's ridiculous!
GTX690 = 45 fps, HD7990 = 63 fps, ARES II = 71 fps. He seems to have been gaming A-OK with 2GB for the last 12 months, but 4GB, oh boy it's A FAIL. What about someone who will only buy a single R9 390/390X card, 4GB is also a fail?

Surely he could have sold 680s and gotten a cheap R9 290s CF upgrade (considering as of December 2014 they were deals on those cards on black friday such as $200 Sapphire TriX 290 or $200 PowerColor PCS+ 290). It wasn't even that hard to see those deals on Slickdeals.net.

Yet today, he trashes an unreleased card that will be way slower than GTX980 SLI but never warned 970SLI/980SLI gamers from "wasting their money when 4GB is DOA starting the day GM200 6GB is out" (his view) only because in his mind it's 6-8GB of bust and "4GB of VRAM" is a fail without even knowing the price or having any concrete proof of where 980 SLI 4GB runs into VRAM bottlenecks at 1440P!!

He also ignores that a lot of gamers care even more about 2nd tier GM200/R9 390 cards due to their superior bang for the buck (this has been the case for both NV and AMD, say 6800GT vs. 6800U or GTX570 vs. 580 or R9 290 vs. R9 290X). As such, since he completely ignores price/performance in his statements, it shows he isn't open minded to the rest of the PC market. Other gamers just say it's good to have 4GB and 8GB options, understanding that both work for different markets.

Many other gamers in this thread seem to be oblivious to prices of GPUs around the world. Even Americans who are bordering Canada, seem to be clueless about this.

In Canada:

MSI Gaming 980 = $690-700 CDN
Asus STrix 980 = $700 CDN
EVGA Superclocked 980 = $700+ CDN
Gigabyte Windforce 980 = $700+ CDN

Asus Titan X = $1369 CDN

All of these prices are before tax, which means for say the province of Ontario, add 13% tax. That would make a 980 a $780-800 CDN after tax.

It's quite amusing to see people trash a card "only" because of its 4GB of VRAM without knowing its price or performance. All I know is if I am in Toronto, I am paying nearly $800 CDN for a 980 (WTH, that's a next mid-range card ffs). THAT is why dismissing R9 390 4GB non-X is pure ignorance on some posters' behalf who seem to have narrowed down this thread to R9 390X vs. GM200 6GB, forget all other R9 300 SKUs. :sneaky:

 

[Cloudfire777]

I have trouble believing AMD will skip 16nm from TSMC and do 14nm. This could just be an experiment from AMD, but also something they are shooting for with 400 series. Who knows

ASIC / Layout Design Engineer at AMD

Back-end dGPU chip design

Responsibility:
• Netlist to GDSII implementation / verification
• dGPU top level clocks RDL design
• Power noise IR/EM analysis / signoff

Projects experience:
• Currently work on dGPU chip 14nm project

 

[boozzer]

It's not a personal attack. It really seemed like you didn't know what SLI was. We've been going back and forth for a while now, I've mentioned what I have several times it's also in my sig yet you keep spouting 80-100% performance advantage. What else would you have me surmise?

What's sad is letting brand loyalty dumbing down your arguments. There's a lack of comprehension on your part, perhaps it's intentional and you're trying to prove a point, but as with the last back and forth... It's not really working out.

To answer your question, I want to move away from multi-gpu setups. It was great when I first got it but a little less so now that the cards aren't as well supported. I find myself either waiting longer for patches or making manual adjustments in nVIdia control panel to get it working optimally. Not always the case, but it happens more often than it used to, which was practically never. Though I may still keep one of my 680's as a dedicated PhysX card.

you realize you just brush aside my 290x cf and contradicted yourself in the same post to boot?

damn I feel sad. I thought your posts were the way they are on purpose, I guess not. damn, I was having fun too. daaamn, I feel so sad. :'( good luck with everything bro. :thumbsup)

@rs we should move on from his posts. Really don't want to continue in light of the recent post. I feel like a damn bully atm. I thought we were both having fun too

regarding 4gb 390x. it is what I prefer actually. I would save monies till my next upgrade. the perfect scenario would be for both 390x 4gb + 8gb models. who here hates saving monies? I would 100% go 4k the moment a single gpu can handle it at around 60 fps. so 2016/17? maybe? hopefully, both the gpus and monitors would have matured tech for 4k by then, if not i can always wait longer.

Infraction issued for calling out another member and trolling.
-Rvenger

 

[RampantAndroid]

you realize you just brush aside my 290x cf and contradicted yourself in the same post to boot?

damn I feel sad. I thought your posts were the way they are on purpose, I guess not. damn, I was having fun too. daaamn, I feel so sad. :'( good luck with everything bro. :thumbsup)

@rs we should move on from his posts. Really don't want to continue in light of the recent post. I feel like a damn bully atm. I thought we were both having fun too

I think he said it's not enough of an upgrade over having two 680s in SLI. Frankly, on the multi-GPU thing, I agree with him. I have two 670s in SLI. In the long run, it wasn't really worth it. I'll sooner just go on single GPU setups. Better power usage...who cares about Physx.

 

[Cloudfire777]

I wonder when the hell we will see 3DMark or anything benchmark from 390X. We are so close, yet nobody have seen any leaked benchies or GPU-z or anything from the 300 cards.

Crazy thinking we have R7 360, R7 360X, R9 370, R9 370X, R9 380, R9 380X, R9 390, R9 390X and R9 395X2 thats suppose to be released

 

[RussianSensation]

I have trouble believing AMD will skip 16nm from TSMC and do 14nm. This could just be an experiment from AMD, but also something they are shooting for with 400 series. Who knows

Isn't AMD switching from TSMC to GloFo for next gen? That's what I thought but perhaps I am wrong. I think GloFo's/Samsung's joint node is 14nm not 16nm.

I think he said it's not enough of an upgrade over having two 680s in SLI. Frankly, on the multi-GPU thing, I agree with him. I have two 670s in SLI. In the long run, it wasn't really worth it. I'll sooner just go on single GPU setups. Better power usage...who cares about Physx.

If we follow his consistent message that VRAM bottlenecks are critical for him at 1440P, then we would need to compare GTX670 2GB SLI or 680 2GB SLI to modern 3-4GB VRAM SLI/CF setups at 1440P. If you look at my last post, I provided this exact data. What we have is 670 2GB SLI or 680 2GB SLI or 690 falling apart! That means making statements like R9 290X CF/970 SLI wasn't a good enough upgrade is hilarious, when ARES II 7970Ghz CF cards are smashing 690 by 25-70% at 1440P. I don't own 680 SLI so I can't tell you how well it runs but just looking at the benchmarks, it's an awfully slow setup today even against the ancient 7970s. R9 290s CF or 970 SLI CF would have been miles faster than his 680s if he is consistent with his message that his goal is to run things maxed (hence the reason he wants 8GB > 4GB). Given the 2GB VRAM bottleneck at 1440P, even a single R9 290X or a single GTX980 is faster than his dual 680s.

regarding 4gb 390x. it is what I prefer actually. I would save monies till my next upgrade. the perfect scenario would be for both 390x 4gb + 8gb models. who here hates saving monies? I would 100% go 4k the moment a single gpu can handle it at around 60 fps. so 2016/17? maybe? hopefully, both the gpus and monitors would have matured tech for 4k by then, if not i can always wait longer.

That's been my point. Shoving 8GB of VRAM on 390 cards and not releasing a much more affordable R9 390 4GB is a big mistake too. AMD needs 4GB cards, arguably more so than 8GB because more gamers will want a fast $400-500 card than a $700-800 one.

Imagine a $499-549 card that ends up at 138 on this chart. Even if it has 4GB of VRAM, how is that not great?! I mean look at it - 280X CF > 980 and 280Xs only have 3GB of VRAM. His smack talking about 4GB of VRAM for gamers not particularly interested in buying dual R9 390/390Xs is not even aligned with reality.

As of April 1, 2015, R9 280X CF is 55% faster than GTX Titan 6GB and easily beat a 980. If 4GB VRAM bottleneck was so common at 1440P (at playable FPS) in modern titles, this would never happen chief!

 

[96Firebird]

I kind of picture GPU memory like a water reservoir. If you use a garden hose to keep it from overflowing vs a fire hose. The fire hose would allow more water volume to flow thru. GPU is the hose in this scenario.

Now I'm not expert on hardware and all that, but here is how I see it (working off your reservoir analogy):

We'll say the reservoir is the amount of VRAM available, we'll use 2GB for arguments sake. Now, a 4GB reservoir will hold twice as much water. There is a pump outside the reservoir, it's power is defined by the GPU power. The size of the fire hose leading to the pump is the memory bandwidth. From the pump, the water goes back into the reservoir, passing an impeller that generates power (the image, FPS). There is also a smaller garden hose (PCI-E lanes) that fills the reservoir with fresh water (new assets needed in VRAM) from a lake (system RAM), and another hose that transfers the dirty water (unneeded assets) back to the lake.

I hope I haven't confused anything yet. When a game loads, all of the VRAM assets are loaded from the system RAM to VRAM. This is the initial filling of the reservoir. The GPU pumps the water past the impeller, in a closed-circuit. As you move around in the game, new assets are needed, but are not in the reservoir because it could only fit so much to begin with. So the fresh water needs to travel through the garden hose, the dirty water travels back to the lake. If the game needs the fresh water soon (ie. it didn't cache it soon enough), the pump has to wait. This is what causes the stutters. If the pump has to wait, the impeller slows down (drop in FPS) until the reservoir gets all the new water.

Now, if the reservoir was larger, it would already have that fresh water available, and wouldn't need to wait on the garden hose from the lake.

That just is an example of how VRAM capacity =/= bandwidth.

Having a higher bandwidth (larger fire hose) is a good thing, but that only helps out if the GPU pump can flow more fluid than the hose can handle. If the pump is starved, increase the diameter of the pump (overclock the memory).

Like I said, I'm no expert at this stuff. That is just how I see it, after reading a bit. If anyone wants to clarify or criticize, I'm all ears.

 

[RussianSensation]

Like I said, I'm no expert at this stuff. That is just how I see it, after reading a bit. If anyone wants to clarify or criticize, I'm all ears.

OR not worry about any of that stuff and do what you are doing -- Buy a 560Ti/570, then buy a 670, then buy a 970, then buy a Pascal GP204, while investing the proceeds from the resale of your old card. No matter if one goes with R9 390CF or GM200 SLI, in 3-4 years a single $700 card will beat either setup. HD7950/7970 owners got "lucky" that there was no price premium for extra VRAM but paying extra for more VRAM would have been a mistake. It would have been better to instead save the $ and upgrade sooner later. All those gamers who paid $100-150 extra for 680 4GB, HD7970 6GB, GTX780 6GB, GTX770 4GB wasted $. That's what AIBs, AMD and NV count on - to get you to spend hundreds of dollars extra for "future-proofing." That's their marketing game and they want us to do that. We need to make an evaluation and see when it makes sense (8800 GT 512MB over 8800GT 256MB = yes make sense, HD4870 1GB vs. HD4870 512MB = yes makes sense, $450 770 4GB vs. $300 770 2GB = waste of $, $550 780 6GB vs. $400 R9 290 4GB = waste of $).

The strategy of spending $1400 on dual 390X 8GB or GM200 6GB and holding on to them for 4 years is going to fail, and I'll bet anything on it. The person buying dual $400-500 cards June 2015, reselling them in 2 years and buying another set of $400-500 cards, while re-investing the proceeds from the sale come June 2017 will be better off even if GM200/390X have 32GB of VRAM.

You can even go check NV's or AMD's history and you will see that 2nd tier and 1st tier cards become outdated at the same time. The flagship cards are intended for people who upgrade every generation or for those whose income ensures that $700 cards are easily affordable. For everyone else, flagship cards are a bad buy and upgrading more often is the right way to keep your PC up-to-date. There are examples where spending a bit more is worth it for performance such as $240 R9 290 vs. GTX960 but spending hundreds of dollars extra for the flagship card over the 2nd tier card is almost always a waste of $ for those who truly can't afford to upgrade every new gen.

 

[Enigmoid]

You are really trying aren't you? I may be wrong but this is what I get from his logic.

His story has been inconsistent for a long time now but he won't admit it. He makes wild statements like 4GB of VRAM is probably going to be a bottleneck for 1080P soon, while he supposedly has been gaming on GTX680 2GB SLI at 1440P for the last 3 years but his 2GB bottleneck wasn't a problem that bothered him enough?

I don't think he denies it bothers him. Its a problem but you can turn down settings (as he has been doing) to get the game into 2 GB. Its workable but not ideal. That said its not something he wants to go through again.

2GB has been a real bottleneck not some 'hypothetical' bottleneck sometime in the future that seems to be more of a case with 4GB vs. 8GB. Yet, gaming on 2GB with reduced settings is A-OK for years but buying a card with "only" 4GB that might become bottlenecked in 1-2 years is a waste of $?

Because he learns from his mistakes? Once bitten twice shy.

Seriously? Also, his entire GPU upgrading strategy to be frank is outdated. Sure, there will be gamers who buy flagship cards and hold on to them for 3-4 years, but you know what, this strategy itself has proven to be inferior than buying and reselling every 2-2.5 years. Let's face it, most enthusiast gamers who had 780Ti SLI or Titan X SLI owners today upgrade every generation. The types of guys buying $700 cards don't care to future-proof for 3-4 years. The minute a card 40-50% faster is out, they are dumping their old cards because these are an elite group of gamers who always buy the best. The reason someone buys Titan X SLI today is not to future-proof years down the line but to have awesome performance today. If you have a 4K monitor, you need all the performance you can get - that's why you want cards like Titan Xs. Since these gamers are on the cutting edge of tech, they won't hold on to these cards for too long once Pascal launches. In the group of guys who buy $700+ cards in pairs, he is actually a tiny minority who wants to keep his cards > 3 years. Most of them on our forum and OCN do not upgrade like that.

He bought 680 SLI a while back and is trying to make it stretch. Many people try out CF and SLI and decide they don't like it. He wants a single GPU card. However, any reasonably priced single GPU card isn't terribly more powerful than 680 SLI when not vram limited.

Also I don't believe he is even in the group of people who bought $700 cards.

If his story added up and he couldn't bare running into VRAM bottlenecks, there is no way he'd be gaming on 680 2GB cards right now. He would have at least sold them a long time ago and gotten R9 290/290Xs or 970s/980s as an intermediary solution assuming his standards are SO high that he can't stand VRAM bottlenecks. Then, if VRAM requirements grew beyond 4GB, he would have sold those cards and gotten GM200s, 14nm, etc. That means for a 100% fact, he games reducing texture and other settings to minimize his own 2GB bottleneck or avoid all titles that require > 2GB. He must been doing that for 1.5+ years but somehow it's not acceptable for someone to save $ and get the 4GB 390 nonX or even 390X version and also reduce some settings to save $? (for example, someone might not care to spend $150-200 more just for 15% more performance and 6-8GB VRAM - but he never considers this and constantly evades this point).

I don't doubt that you are right. But this is a costlier approach than sitting through mid/high texture settings and waiting for a large substantial increase. Again, no CF/SLI.

If you go check 20-30 modern games tested in the last 15-18 months, ARES II (nothing but an HD7970 1.1Ghz) crushes GTX690 by 20-40% at 1440P. This has been consistent especially in AAA games. There have been a giant slew of games that cripple 680 2GB SLI setup specifically because of 2GB of VRAM:

- Wolfenstein TWO
- AC Unity
- Dead Rising 3
- Watch Dogs
- FC4
- GTA V
- Shadow of Mordor, etc.

And you know what?

You are comparing an overclocked 7970 setup to an underclocked 680 SLI setup. If he has 680 SLI he gains ~+5% perf over the 690. If he is using aftermarket cards and has them overclocked then he can gain another 10-15% over the 690. If he reduces settings appropriately then the 690 tanking due to vram can be reduced further.

On launch the Ares II did almost 20% better than the 690 (removing games with no CF support) at 1440p/1600p.

Yet today, he trashes an unreleased card that will be way slower than GTX980 SLI but never warned 970SLI/980SLI gamers from "wasting their money when 4GB is DOA starting the day GM200 6GB is out" only because in his mind it's 6-8GB of bust and "4GB of VRAM" is a fail without even knowing the price or having any concrete proof of where 980 SLI 4GB runs into VRAM bottlenecks at 1440P!!

In 2012 no one could pull up instances where 3 GB on 7970 ghz CF was advantageous over 680 SLI 2 GB either.

390X is frankly in a different performance class. Would 4 GB be a problem with ~titan X performance?

390X only having 4 GB would be like AMD releasing the 290X series with only 3 GB.

He also ignores that a lot of gamers care even more about 2nd tier GM200/R9 390 cards due to their superior bang for the buck (this has been the case for both NV and AMD, say 6800GT vs. 6800U or GTX570 vs. 580 or R9 290 vs. R9 290X). As such, since he completely ignores price/performance in his statements, it shows he isn't open minded to the rest of the PC market. Other gamers just say it's good to have 4GB and 8GB options, understanding that both work for different markets.

Sure its good to have 4 GB and 8 GB versions. Having only a 4 Gb version is a little iffy though. That said, any card releasing > than the 980 level of performance should have more than 4 GB.

Many other gamers in this thread seem to be oblivious to prices of GPUs around the world. Even Americans who are bordering Canada, seem to be clueless about this.

In Canada:

MSI Gaming 980 = $690-700 CDN
Asus STrix 980 = $700 CDN
EVGA Superclocked 980 = $700+ CDN
Gigabyte Windforce 980 = $700+ CDN

Asus Titan X = $1369 CDN

All of these prices are before tax, which means for say the province of Ontario, add 13% tax. That would make a 980 a $780-800 CDN after tax.

Again with the inability to price anything. Price in canadian no tax with rebates.

Cheapest 980 $630 - http://www.vuugo.com/computer-hardw...ards-ZT-90204-10P.html?tracking=5108053939762

Windforce - $640 http://www.ncix.com/detail/gigabyte-geforce-gtx-980-wf-a6-103002-1334.htm?affiliateid=7474144

Strix - $ 685 http://www.ncix.com/detail/asus-geforce-gtx-980-strix-cc-102671-1334.htm?affiliateid=7474144

superclocked - $690 http://www.newegg.ca/Product/Produc..._-na&utm_medium=affiliates&utm_source=afc-%zn

Windforce looks the best

Titan X - $1212 http://www.vuugo.com/computer-hardw...d-GTXTITANX-12GD5.html?tracking=5108053939762

(Titan X is all reference so just crank up the speeds yourself - you will be cooler limited)

This is why I have to take anything you say about perf/price with a huge grain of salt... the prices are always wrong.

It's quite amusing to see people trash a card "only" because of its 4GB of VRAM without knowing its price or performance. All I know is if I am in Toronto, I am paying nearly $800 CDN for a 980 (WTH, that's a next mid-range card ffs). THAT is why dismissing R9 390 4GB non-X is pure ignorance on some posters' behalf who seem to have narrowed down this thread to R9 390X vs. GM200 6GB, forget all other R9 300 SKUs.

Why compare to the 980? The 2nd cut down chip (970) is half the price for 85% the performance?

 

[96Firebird]

OR not worry about any of that stuff and do what you are doing...

That's all good and well, but I tend to keep my cards for longer than one generation, except this last time. I only sold my 770 because I was paying close attention to the 970/980 launch, and knew the 970 would tank my resale value. Before that, I kept an 8800GT 512MB until I bought a GTX 460 1GB, which lasted me till my 770.

But my post was more a technical post about how VRAM capacity =/= bandwidth, and how they both effect the performance.

 

[3DVagabond]

so 980 n 970 is trash right? and please don't write that 980 is not top of the line. it was top of the line before titan x and it is a 550$ flagship even if it is only a mid range card. and since 5gb ram is all we need, what do you think of titan x and it's useless 12gb? because by the time we can use more than 4gb per you, it's gpu power would be the same as a mid range card by 2016.

so 980 is trash, titan x is a waste. looks like nv gpus got problems left and right.

@rail you should quote it, funnier that way, and would make a perfect point at the rediculousness of it.

I would just add that the console uses 5GB at often less than 1080 and very low quality compared to what we see for PC. So, if we really want to be safe we're going to need 12GB just to run 1080, and forget about 1440 or 4K, they are only pipe dreams because of these Uber consoles running around. Of course, you'd be a fool to buy anything but GCN too, if we want to continue this train of reasoning to it's logical conclusion.

 

[RampantAndroid]

That's going to go over well with NV having majority share in the GPU market. NV will probably encourage devs to include higher resolution textures and push up VRAM requirements as a result, knowing their competition is at a disadvantage.

While possibly true, isn't upping the textures kinda a good thing to be doing anyway? A few years ago, 2GB was OK for 1440p. Now no one recommends less than 4GB. Why not future proof yourself and just buy a card with more VRAM and be done with it? I'd rather be limited by the GPU itself and not VRAM quantity.

If AMD is releasing a top end card with only 4GB on it, I wonder how long lived it'll be - how quickly will people need to push up to a new card? (and this same question goes for GTX970 and 980 users); I want 1440p and going to 4GB would do it, but I don't upgrade very often anymore, so I'd rather make a slightly larger investment now if it'll last a lot longer.

 

[nsavop]

You are really trying aren't you? I may be wrong but this is what I get from his logic.



I don't think he denies it bothers him. Its a problem but you can turn down settings (as he has been doing) to get the game into 2 GB. Its workable but not ideal. That said its not something he wants to go through again.

Because he learns from his mistakes? Once bitten twice shy.

He bought 680 SLI a while back and is trying to make it stretch. Many people try out CF and SLI and decide they don't like it. He wants a single GPU card. However, any reasonably priced single GPU card isn't terribly more powerful than 680 SLI when not vram limited.

Also I don't believe he is even in the group of people who bought $700 cards.

I don't doubt that you are right. But this is a costlier approach than sitting through mid/high texture settings and waiting for a large substantial increase. Again, no CF/SLI.

And you know what?

You are comparing an overclocked 7970 setup to an underclocked 680 SLI setup. If he has 680 SLI he gains ~+5% perf over the 690. If he is using aftermarket cards and has them overclocked then he can gain another 10-15% over the 690. If he reduces settings appropriately then the 690 tanking due to vram can be reduced further.

On launch the Ares II did almost 20% better than the 690 (removing games with no CF support) at 1440p/1600p.

In 2012 no one could pull up instances where 3 GB on 7970 ghz CF was advantageous over 680 SLI 2 GB either.

390X is frankly in a different performance class. Would 4 GB be a problem with ~titan X performance?

390X only having 4 GB would be like AMD releasing the 290X series with only 3 GB.

Sure its good to have 4 GB and 8 GB versions. Having only a 4 Gb version is a little iffy though. That said, any card releasing > than the 980 level of performance should have more than 4 GB.

Again with the inability to price anything. Price in canadian no tax with rebates.

Cheapest 980 $630 - http://www.vuugo.com/computer-hardw...ards-ZT-90204-10P.html?tracking=5108053939762

Windforce - $640 http://www.ncix.com/detail/gigabyte-geforce-gtx-980-wf-a6-103002-1334.htm?affiliateid=7474144

Strix - $ 685 http://www.ncix.com/detail/asus-geforce-gtx-980-strix-cc-102671-1334.htm?affiliateid=7474144

superclocked - $690 http://www.newegg.ca/Product/Produc..._-na&utm_medium=affiliates&utm_source=afc-%zn

Windforce looks the best

Titan X - $1212 http://www.vuugo.com/computer-hardw...d-GTXTITANX-12GD5.html?tracking=5108053939762

(Titan X is all reference so just crank up the speeds yourself - you will be cooler limited)

This is why I have to take anything you say about perf/price with a huge grain of salt... the prices are always wrong.

Why compare to the 980? The 2nd cut down chip (970) is half the price for 85% the performance?

Spot on +1.

AMD releasing the 390X with 4GB would be like them releasing the 6970 with 1GB or the 7970 with 2GB. When is the last time AMD's flagship didn't ship with more memory then the previous gen? AMD has always been ahead of nvidia as far as vram is concerned and for good reason as most of Nvidia cards would vram bottleneck much quicker. Really hope the 390x is 8 GB.

 

[RampantAndroid]

Spot on +1.

AMD releasing the 390X with 4GB would be like them releasing the 6970 with 1GB or the 7970 with 2GB. When is the last time AMD's flagship didn't ship with more memory then the previous gen? AMD has always been ahead of nvidia as far as vram is concerned and for good reason as most of Nvidia cards would vram bottleneck much quicker. Really hope the 390x is 8 GB.

I hope it's 8GB, as it becomes a possible option for an upgrade for me. At 4GB, I'll be more likely to wait for a 980ti and consider my options.

 

[2is]

His story has been inconsistent for a long time now but he won't admit it. He makes wild statements like 4GB of VRAM is probably going to be a bottleneck for 1080P soon, while he supposedly has been gaming on GTX680 2GB SLI at 1440P for the last 3 years but his 2GB bottleneck wasn't a problem that bothered him enough? 2GB has been a real bottleneck not some 'hypothetical' bottleneck sometime in the future that seems to be more of a case with 4GB vs. 8GB. Yet, gaming on 2GB with reduced settings is A-OK for years but buying a card with "only" 4GB that might become bottlenecked in 1-2 years is a waste of $? Seriously? Also, his entire GPU upgrading strategy to be frank is outdated. Sure, there will be gamers who buy flagship cards and hold on to them for 3-4 years, but you know what, this strategy itself has proven to be inferior than buying and reselling every 2-2.5 years. Let's face it, most enthusiast gamers who had 780Ti SLI or Titan X SLI owners today upgrade every generation. The types of guys buying $700 cards don't care to future-proof for 3-4 years. The minute a card 40-50% faster is out, they are dumping their old cards because these are an elite group of gamers who always buy the best. The reason someone buys Titan X SLI today is not to future-proof years down the line but to have awesome performance today. If you have a 4K monitor, you need all the performance you can get - that's why you want cards like Titan Xs. Since these gamers are on the cutting edge of tech, they won't hold on to these cards for too long once Pascal launches. In the group of guys who buy $700+ cards in pairs, he is actually a tiny minority who wants to keep his cards > 3 years. Most of them on our forum and OCN do not upgrade like that.

If someone is OK paying $150-200 more per card for minimal performance gains just to future-proof, it's that gamer's choice but he should at least admit that this strategy is inferior to start with. It's been proven mathematically and statistically that buying 2nd tier cards and upgrading more often is superior to buying dual flagships and holding on to them or 3-4 years. He won't admit this is a fact. From that perspective, unless I have a 4K gaming monitor, why the heck would I are if R9 390 or GM200s last me 4 years? If someone wants to waste hundreds of dollars extra to future-proof, let them do it. Plenty of people wasted $ on 680 SLI vs. 670 SLI for e-peen.

If his story added up and he couldn't bare running into VRAM bottlenecks, there is no way he'd be gaming on 680 2GB cards right now. He would have at least sold them a long time ago and gotten R9 290/290Xs or 970s/980s as an intermediary solution assuming his standards are SO high that he can't stand VRAM bottlenecks. Then, if VRAM requirements grew beyond 4GB, he would have sold those cards and gotten GM200s, 14nm, etc. That means for a 100% fact, he games reducing texture and other settings to minimize his own 2GB bottleneck or avoids all titles that require > 2GB. What he preaches in this thread is inconsistent with what he actually does himself. He must have been suffering 2GB VRAM bottlenecks for at least a a year now, but somehow it's not acceptable for someone to save $ and get the 4GB 390 non-X or even 390X version and also reduce some settings to save $? Woot?! (for example, someone might not care to spend $150-200 more just for 15% more performance and 6-8GB VRAM - but he never considers this and constantly evades this point).

If you go check 20-30 modern games tested in the last 15-18 months, ARES II (basically HD7970 1.1Ghz in pairs) crushes GTX690 by 25-70% at 1440P.

This has been consistent especially in AAA games. There have been a giant slew of games that cripple 680 2GB SLI setup specifically because of 2GB of VRAM:

http://gamegpu.ru/images/remote/htt...trategy-Homeworld_Remastered-test-hr_3840.jpg

- Evolve
- Wolfenstein TWO
- Mortal Kombat X
- AC Unity
- Dead Rising 3
- Watch Dogs
- FC4
- DAI
- GTA V
- Shadow of Mordor, etc.

See HD7970Ghz CF (Ares II) has been mopping the floor with GTX690 (aka GTX680 2GB SLI) but he didn't have much of problem with the 2GB VRAM bottleneck that was real and measurable over the course of the last 12 months, not in some "hypothetical future".

Look at this, it's ridiculous!
GTX690 = 45 fps, HD7990 = 63 fps, ARES II = 71 fps. He seems to have been gaming A-OK with 2GB for the last 12 months, but 4GB, oh boy it's A FAIL. What about someone who will only buy a single R9 390/390X card, 4GB is also a fail?

Surely he could have sold 680s and gotten a cheap R9 290s CF upgrade (considering as of December 2014 they were deals on those cards on black friday such as $200 Sapphire TriX 290 or $200 PowerColor PCS+ 290). It wasn't even that hard to see those deals on Slickdeals.net.

Yet today, he trashes an unreleased card that will be way slower than GTX980 SLI but never warned 970SLI/980SLI gamers from "wasting their money when 4GB is DOA starting the day GM200 6GB is out" (his view) only because in his mind it's 6-8GB of bust and "4GB of VRAM" is a fail without even knowing the price or having any concrete proof of where 980 SLI 4GB runs into VRAM bottlenecks at 1440P!!

He also ignores that a lot of gamers care even more about 2nd tier GM200/R9 390 cards due to their superior bang for the buck (this has been the case for both NV and AMD, say 6800GT vs. 6800U or GTX570 vs. 580 or R9 290 vs. R9 290X). As such, since he completely ignores price/performance in his statements, it shows he isn't open minded to the rest of the PC market. Other gamers just say it's good to have 4GB and 8GB options, understanding that both work for different markets.

Many other gamers in this thread seem to be oblivious to prices of GPUs around the world. Even Americans who are bordering Canada, seem to be clueless about this.

In Canada:

MSI Gaming 980 = $690-700 CDN
Asus STrix 980 = $700 CDN
EVGA Superclocked 980 = $700+ CDN
Gigabyte Windforce 980 = $700+ CDN

Asus Titan X = $1369 CDN

All of these prices are before tax, which means for say the province of Ontario, add 13% tax. That would make a 980 a $780-800 CDN after tax.

It's quite amusing to see people trash a card "only" because of its 4GB of VRAM without knowing its price or performance. All I know is if I am in Toronto, I am paying nearly $800 CDN for a 980 (WTH, that's a next mid-range card ffs). THAT is why dismissing R9 390 4GB non-X is pure ignorance on some posters' behalf who seem to have narrowed down this thread to R9 390X vs. GM200 6GB, forget all other R9 300 SKUs. :sneaky:

You must have me mixed up with someon else. I don't game on a 1440p screen and I've acknowledged multiple times in this thread the limitations of 2gb on the 680s. Tell you what. You quote me where I claimed I'm gaming at 1440p and I'll quote myself about the limitations of 2gb on the 680s and we will see who's consistent and who's making things up. If I'm right you gift me a 980ti if your claim ends up being correct I'll gift you a 390x. Let's put the BS to bed and see who's the story teller.

 

[happy medium]

If 3.5gb of Vram is not enough for a gtx 970, how is 4gb of Vram enough for a 390x? I thought the 390x will be like 50% faster? Am I missing something here?

 

[Cloudfire777]

Its not enough for a 4K high end gaming card.
People have mentioned that GTX 680 did fine against 7970 despite less VRAM but neither cards had the pure power to drive very high res with maxed out settings skyrocket the VRAM usage like 390X/980Ti do.

Max FPS might do ok with 4GB HBM, but I dont think you will get a butter smooth experience with it overall.
More and more games will be unoptimized ports from the consoles too with memory hog like performance, so 4GB is NOT future proof