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

[Headfoot]

I mean who wants a powerful but power hungry card? And price be like $699 for the non water block model

Literally anyone who calls themselves an enthusiast. Who cares, at all, that a card uses up a couple dollars more electricity in a year? Completely 100% irrelevant for 95% of actual enthusiasts. What a completely asinine statement

 

[AnandThenMan]

It will be good if the 390X consumes less than 170 watts while gaming. I mean who wants a powerful but power hungry card? And price be like $699 for the non water block model

The 390X has to be much better, and less expensive, than Titan or it's a failure. Is that what you're saying here?

 

[exar333]

Literally anyone who calls themselves an enthusiast. Who cares, at all, that a card uses up a couple dollars more electricity in a year? Completely 100% irrelevant for 95% of actual enthusiasts. What a completely asinine statement

Totally agree.

Where efficiency (to me) comes into play is when you have relatively equal-performing options, the more efficient will be more desirable. What 'muddies' this water somewhat is also the perf/$ metric.

If you look at the GPU landscape pre-Titan X, you had (2) great performers in the GTX 980 and the 290X. Both competent and not far-off in performance. The 980 gave you efficiency + performance and the 290x gave you perf/$ + performance. Both are valid options, depending on what you value more.

In the end, would someone choose a less efficient, but much faster GPU? Absolutely.

 

[mohit9206]

The 390X has to be much better, and less expensive, than Titan or it's a failure. Is that what you're saying here?

Yes.
AMD's GPU market share is much less than Nvidia's so the only way people would take note of AMD would be if they release a gpu which is faster, cheaper and more power efficient than Nvidia.
I didn't mean 170 watts literally, just less than Titan and 980Ti while while being faster and cheaper as well.

 

[5150Joker]

Because someone like Joker doesn't care about money. $1000 or $600 makes no difference to him.

Yeah money grows on trees for me and I wipe with $100 bills. Plus NVIDIA pays me a lot to act as a shill on AT..I mean..oops! D)

Another graphics card from AMD/ATI just appeared in database --

ATI TECHNOLOGIES ULC

VIDEO CARD

C63431

MSIP-REM-ATI-102-C63431

ATI TECHNOLOGIES ULC

2015-03-24

I wish they hadn't retired the ATi brand. It has a fond memory for those of us that were around before AMD slapped their name on everything.

 

[Silverforce11]

@5150Joker
It was legit, in context with your declaration that 390X has to be 15% faster than Titan X, irrespective of the fact they are not even in the same price bracket/segment.

That's why Russian pointed out the obvious. Why should a much cheaper GPU be 15% faster than one that's $1000 to be considered good?

If you really feel that way for real, it suggests you don't care about money because price isn't a factor.

 

[5150Joker]

@5150Joker
It was legit, in context with your declaration that 390X has to be 15% faster than Titan X, irrespective of the fact they are not even in the same price bracket/segment.

That's why Russian pointed out the obvious. Why should a much cheaper GPU be 15% faster than one that's $1000 to be considered good?

If you really feel that way for real, it suggests you don't care about money because price isn't a factor.

Point of my post was that given the hype surrounding the R390x, especially with respect to HBM and the OP's quote where it says that the purported performance is "much higher" than leaked figures suggest, it creates a perception that this thing should be noticeably faster than Titan X. Originally we were told about 50-60% faster than 290X (which is where Titan X stands) and if the OP is to be believed, that was underestimating the true performance. Therefore, I said given the hype, it would be underwhelming if it doesn't at least achieve 15% over the Titan X. Price is not part of the discussion, I'm speaking strictly in terms of performance.

And for once I agree with RS about one small point: I wouldn't buy the R390X because I am locked into G-Sync + Titan X likely has a higher OC headroom than R390X would offer. If I hadn't purchased a G-Sync display and I had at least a 780 Ti in my system, I'd probably wait to see what R390x offers or the cut down GM200. But my situation is unique so I went with what is the best performance on the market. Plus this is the full GM200 which makes it better than the first crippled Titan.

 

[Silverforce11]

Point of my post was that given the hype surrounding the R390x, especially with respect to HBM and the OP's quote where it says that the purported performance is a "lot higher" than leaked figures suggest, it creates a perception that this thing should be noticeably faster than Titan X. Originally we were told about 50-60% faster than 290X and if the OP is to be believed, that was underestimating the true performance. Therefore, I said given the hype, it would be underwhelming if it doesn't at least achieve 15% over the Titan X. Price is not part of the discussion, I'm speaking strictly in terms of performance.

And for once I agree with RS about one small point: I wouldn't buy the R390X because I am locked into G-Sync + Titan X likely has a higher OC headroom than R390X would offer. If I hadn't purchased a G-Sync display and I had at least a 780 Ti in my system, I'd probably wait to see what R390x offers or the cut down GM200. But my situation is unique so I went with what is the best performance on the market. Plus this is the full GM200 which makes it better than the first crippled Titan.

Fair enough. There's a lot of hype, but also plenty of anti-hype. It depends where you fall as I said in the other post, whether you are pessimistic about AMD, realistic or optimistic. There's been leaks and such to support a wild range of performance increase, from 30% above R290X to much higher than 50%.

 

[5150Joker]

Fair enough. There's a lot of hype, but also plenty of anti-hype. It depends where you fall as I said in the other post, whether you are pessimistic about AMD, realistic or optimistic. There's been leaks and such to support a wild range of performance increase, from 30% above R290X to much higher than 50%.

I think AMD is aware of the pressure to perform and will do a better job this time around w/the R390x launch than they did the 290/290X. The real test will be how they position their other high volume products and of course marketing. Plus IMO to grab customers like me, they need to get back to regular driver release schedules every month and start pouring more dollars into their GE program.

 

[boozzer]

I think AMD is aware of the pressure to perform and will do a better this time around w/the R390x launch than they did the 290/290X. The real test will be how they position their other high volume products and of course marketing. Plus IMO to grab customers like me, they need to get back to regular driver release schedules every month and start pouring more dollars into their GE program.

it is just marketing bro, nothing but marketing. it is what amd needs.

 

[dacostafilipe]

I think AMD is aware of the pressure to perform and will do a better job this time around w/the R390x launch than they did the 290/290X. The real test will be how they position their other high volume products and of course marketing.

If they release a full GCN (64CU) we should expect at least a ~45% increase in performance if they don't hit other bottlenecks (power, cache, front-end, etc ...)

More performance can only be achieved with arch modifications (memory, cache, gating, etc... )

Plus IMO to grab customers like me, they need to get back to regular driver release schedules every month and start pouring more dollars into their GE program.

There's a lot going around the driver team at the moment that could explain the lack of regular updates.

The new AMDKFD drivers for HSA 1.0 support was released some weeks ago. This is not only a driver needed for AMD Hardware (ex: Carrizo) but it's also the template other vendors will use to create their own HSA drivers.

The "new" linux driver AMDGPU is around the corner. This Kernel driver will have a huge impact as it intends to have a unique codebase for Catalyst and the opensource driver.

I'm not trying to find excuses, if they are late because of this, then they certainly have a resources management issue. I just wanted to tell you that they are advancing in other, maybe less visible for most, areas.

PS: Just wanted to add that they are also late with the AMDGPU driver. They will certainly miss the 4.1 merge window that closes in a week or so.

 

[AtenRa]

WTF? If anything, Russian is the real "fair and balanced". He always advocates whichever card offers the best perf/$, unlike some fanboys in denial.

Agreed but this time he speculates on OC capabilities for a unreleased product (R9 390).

 

[raghu78]

Agreed but this time he speculates on OC capabilities for a unreleased product (R9 390).

There was a blanket statement that Titan-X OC was 90% faster than R9 290X (1 Ghz). I debunked that statement with accurate nos for both stock and OC across anandtech's entire game suite. On avg Titan-X was 43% faster than R9 290X at 4k. Overclocked it gained roughly 17% perf in all games. So I extended the scaling and got 67% higher perf from Titan-X OC vs R9 290X (1 Ghz). So yeah I proved him wrong and he did not argue back. So I think he understood that he exaggerated it wildly.

btw I am expecting a stock R9 390X WCE at 55% higher perf than R9 290X with roughly 15 - 20% perf gain from overclocking. There are a lot of factors at work which should help AMD achieve it - architectural improvements from an improved tiled GCN arch to improve perf/sp, massive bandwidth improvements due to HBM and bandwidth efficiency improvements from color compression, ROP and tesselation performance improvements.

 

[looncraz]

What difference is there from the drivers view in relation to GDDR and HBM? It was you claiming that HBM was the reason for driver delay because it was a new thing.

There is no inherent difference for the driver in regards to what type of memory is being used. The GPU creates an address space which the driver references.

HOWEVER, an 8GB version means that the memory space shouldn't be treated as uniform since the card can likely only address a 4GB section at a time as a result of HBM's physical connection to the GPU.

Multiplexed HBM for the 8GB variant means that the GPU will need to have its memory allocations organized to minimize access delays. That is the driver's job. Can you imagine the negative feedback if the 8GB version of the 390X was slower than the 4GB version?!

 

[Shehriazad]

Why does the 390X have to be faster, cheaper AND more power efficient?


If it fills in number 1 and 2...the most of the people will jump to it. Especially the AIO water cooled one.
It's not even gonna get to 70°C while gaming with watercooling if your ambient temps aren't insane. (22-25°C) I've seen 295x2 top out at 65°C with their water cooling blocks...and that's a dual GPU with 500W TDP.

The Titan X however easily reaches 80°C after some light overclocking and sits in the 70°C zone even without additional OCs.


So if the 390X can be comparable to a Titan X (or even stronger), cost less...and reach lower temps...then it won on all ends that actually matter as wattage itself is hardly of any importance to people that buy such cards.


I'm not even sure why...but people screaming "but muh wattage" annoy the crap out of me...especially if they claim to be enthusiasts or core gamers.

 

[ocre]

No, I looked at many games in this review and this one (Titan X OC vs. 970/980/290X OC).

Titan X's OC performance is flat out amazing if it can hold 1.4Ghz+ clocks. I would have been all over 1-2 of those cards had NV priced them at $550-600 but at $1K I am spending $0. I am not supporting $1K single-chip videocards out of principle. My point though is if someone can afford it and doesn't care about principles, there is no reason to wait for the R9 390X when Titan X OC is ~ R9 295X2. :thumbsup:

That was my point for early adopters of Titan X, it's likely irrelevant what R9 390X costs and what its performance is. If a hypothetical R9 390X is $599 and is 5% faster than the Titan X, they would still find the Titan X worth it because they would say they owned it for 3-4 months. Early adopters pay these huge premiums to have the best on day 1/week 1. Remember how I said before that I thought R9 390X had a chance to beat the Titan X at stock but I was a lot more skeptical about R9 390X overclocking as well as the Titan X.

Think about it, if you strap a water block and unlock the bios, the Titan X has the potential to hit 1.5-1.55Ghz which would be a 40-50%+ overclock. IMO, R9 390X has 0 chance of matching that on water if it's clocked at 1.05Ghz given how historically leaky/dense AMD's chips have been. Look at 4890/5870/6970/7970Ghz/290X overclocking headroom - once they reach a certain wall, it's very very hard to go much higher. 7950/7970 cards are an exception because AMD low-balled their clocks out of the factory. NV cards are a lot more flexibility in that regard and overclock better with a lower increase in voltage. GTX460/470/560Ti/670/680/780/780Ti/970/980 are all awesome overclockers. NV tends to have a better track record for average high overclocks and Maxwell GPUs are among the best ever for NV's overclocking potential.

AMD's last chance to capture the high-end gamers is to launch R9 390/390X before the consumer GM200 6GB launches with AIB coolers. If NV is able to launch EVGA Classified, Asus Matrix, Galax HOF, MSI Lightning consumer GM200 6GB chip before R9 390 series, oh boy, AMD is going to lose a lot of sales. Right now the Titan X at $1K still means that a lot of high end gamers are sitting on the sidelines and waiting. Had NV launched the consumer GM200 6GB AIB cards in March, this would have meant a devastating blow to AMD's strategy with R9 390 series.

That Overclockers club review was great. We seldom get to see overclocked vs overclocked. Not only that, it had a range of resolutions and a decent enough amount of games. Review sites are so limited these days. So many are so skimpy.

What I took away from that review is that an overclocked 980 gets really close to the titanX at stock. My 980 will run boost clocks of just over 1500mhz with very little variation.

You once said that 980 owners made a bad choice and was gloating once the titanX launched. I just wanted you to see that this is not the case at all.
I also wanted to address your statement in another post that says that the 390x is a path for people not running a 980,970, 290x, etc. I think it still very well could be. The resale value of the 980 has held up very well. We have people selling their used 980s and buying titanXs. They are actually getting most of the money they paid back, even after gaming all these months.

The 980 resale value should hold up longer. Only the 390x stands the chance to plummet the 980 value. But today, 980 owners are in a great position. And unless the 390x is a Titan X killer, an overclocked 980 will not look just terrible against it. If the 390x is slightly faster than a titanX stock, an overclocked 980 will still have a respectable showing.

I see 980 owners as having plenty of really great options. It is really all sun shine for them no matter what they do. If the 390x looks to be a beast and titanX killers, 980 owners can sell their cards and recoup a huge chuck of cash to put on a 390x, just like gamers are selling their 980s now and buying titanX's. if they wait too long to sell their 980s, there is always an option to pick up another 980 for SLI. 980 owners have plenty of options now, they are in a great position.

If any 980 owner thinks the 390x is gonna be flat out amassing, they should sell their cards just before the launch. I like this option a lot myself. But I want to wait till some more reliable sources and rumors hit. Cause if the 390x is looking like it will just barely notch out a titanX, I think I will be fine with an overclocked 980, just like I am now after the TitanX launched.

But, I wouldn't mind a 390x that was way faster than the titanX. I wouldn't mind selling my 980 and picking one up. it will have to be quite a bit faster than an overclocked 980. Another thing is that I have a completely jam packed summer planned. I kind of wish we would be seeing the 390x soon. I hope it is before June.

 

[Atreidin]

Why does the 390X have to be faster, cheaper AND more power efficient?


If it fills in number 1 and 2...the most of the people will jump to it. Especially the AIO water cooled one.
It's not even gonna get to 70°C while gaming with watercooling if your ambient temps aren't insane. (22-25°C) I've seen 295x2 top out at 65°C with their water cooling blocks...and that's a dual GPU with 500W TDP.

The Titan X however easily reaches 80°C after some light overclocking and sits in the 70°C zone even without additional OCs.


So if the 390X can be comparable to a Titan X (or even stronger), cost less...and reach lower temps...then it won on all ends that actually matter as wattage itself is hardly of any importance to people that buy such cards.


I'm not even sure why...but people screaming "but muh wattage" annoy the crap out of me...especially if they claim to be enthusiasts or core gamers.

It's just an excuse to validate purchasing one thing over another. If they bought a video card that was proven to be more expensive but worse in every single metric except it was smaller and weighed less, then they'd be talking about how their favorite company has superior performance/density, as if they were planning on filling a space ship with them and desperately needed to save on rocket fuel.

 

[DownTheSky]

this one (Titan X OC vs. 970/980/290X OC).

This might be, testing wise, the best review I've ever read. :thumbsup:

 

[gamervivek]

This might be, testing wise, the best review I've ever read. :thumbsup:

They always did both stock and overclocked in their reviews, though this one isn't in the hodge podge way the earlier reviews used to be.
Also confirms my view that the showdown with 390X will come down to how well they overclock. Titan X does quite well, while the 1050Mhz core clock of the WCE edition doesn't fill me confidence.

edit: Decided to check out the numbers against 290X, first one is stock, second is overclocked.

metro last night : 50.2% 83%
bioshock infi 49.6%
crysis3 45.2% 67.7%
far cry4 53.8% 85.6%
bf4 70.6% 95.1%
ac unity 62.8% 88.2%
batman ao 21.9% 41.1%

Will be a tall order for AMD to go against overclocked big daddy maxwell.

 

[nurturedhate]

It's just an excuse to validate purchasing one thing over another. If they bought a video card that was proven to be more expensive but worse in every single metric except it was smaller and weighed less, then they'd be talking about how their favorite company has superior performance/density, as if they were planning on filling a space ship with them and desperately needed to save on rocket fuel.

I can see it now... arguments over who can fit more cards in a transatlantic cargo ship.

 

[Cloudfire777]

A new day, a "new" leak. Never seen this before. Was posted today.
Whoever made this chart, was spot on with Titan X performance. And that a long time before any leaks was posted on the internet if the timestamp is correct...

And top top that, +65% over R9 390X is just what the AMD leaked slides showed.
Don`t bother look through 3DMark database. Ive been there. The score have not been uploaded. Probably tested on a dedicated bench without internet access.
To top the top again (lol), rumors started floating around about a GTX 980 Ti with higher clocks than Titan X. Guess this is why...

Interesting huh?

 

[Elfear]

They always did both stock and overclocked in their reviews, though this one isn't in the hodge podge way the earlier reviews used to be.
Also confirms my view that the showdown with 390X will come down to how well they overclock. Titan X does quite well, while the 1050Mhz core clock of the WCE edition doesn't fill me confidence.

edit: Decided to check out the numbers against 290X, first one is stock, second is overclocked.

metro last night : 50.2% 83%
bioshock infi 49.6%
crysis3 45.2% 67.7%
far cry4 53.8% 85.6%
bf4 70.6% 95.1%
ac unity 62.8% 88.2%
batman ao 21.9% 41.1%

Will be a tall order for AMD to go against overclocked big daddy maxwell.

It looks like OC's Titan X is a particularly good sample and their 290X is particularly bad. Averages from HWBot are the following:

Titan X (116 submissions)
Average Air Overclock: 1385/2083
Stock: 1079/1752

Increase: 28%/19%

290X (7,104 submissions)
Average Air Overclock: 1144/1553
Stock: 1000/1250

Increase: 14%/24%

Titan X is without doubt a great overclocker but, on average, the disparity won't be as big as OC's review. Still means the 390X better have some overclocking headroom though.

 

[SlowSpyder]

This might be, testing wise, the best review I've ever read. :thumbsup:

I've always thought that was a good site for reviews. I remember a while back they did a review of CPU's and had the Phenom I fairly close to the Intel CPU's of the day (I think they were testing the Q9550 maybe?). That review was posted and people bashed the site because it wasn't well known. But, in my opinion they've always done fair and solid reviews.

 

[5150Joker]

Something relatively new from VR-Zone: http://chinese.vr-zone.com/146560/a...inidad-and-tobago-for-computex-2015-03232015/

The fascinating bit:

According to some whisper, AMD Radeon R9 390X (tentative), which is Fiji XT performance will be stronger than the GeForce GTX 980 by 20%. If the actual performance really so attractive, for consumers, but also pieces seem good thing.

 

[Riceninja]

Something relatively new from VR-Zone: http://chinese.vr-zone.com/146560/a...inidad-and-tobago-for-computex-2015-03232015/

The fascinating bit:

According to some whisper, AMD Radeon R9 390X (tentative), which is Fiji XT performance will be stronger than the GeForce GTX 980 by 20%. If the actual performance really so attractive, for consumers, but also pieces seem good thing.

如果實際表現真有如此誘人，對於消費者而言，似乎也是件不錯的事情

"If actual performance is really this attractive, then from a customer's point of view it's still a pretty good thing"

doesn't sound that good to me. weren't we looking at 40-50% better than 980 previously?