GTX 970 and GTX 980 is officially launched

[jpiniero]

I just recently bought a ASUS R9 280 GPU for $220. I'm wondering if this was a mistake and if I should have waited for a GTX 960 (assuming the performance will scale as similar as the 770 to 970 did).

It's a different die... if it ends up being 8 SMM (exactly half of the GM206), no. If it's 10 (2x 750 Ti), then maybe. Power draw should continue to be stellar though.

 

[dmoney1980]

has anyone seen a review for the EVGA 970 "blower" card? I'm wondering how it would compare to the stock Titan cooler that's used on the 980's

 

[toyota]

has anyone seen a review for the EVGA 970 "blower" card? I'm wondering how it would compare to the stock Titan cooler that's used on the 980's

980 cooler only looks like a Titan cpoler but it's not one as there is no vapor chamber being used.

 

[Fastx]

Toyota, I am curious if you are going with a 980 or 970?

 

[x3sphere]

980 cooler only looks like a Titan cpoler but it's not one as there is no vapor chamber being used.

Huh, one of the reviews I read (Guru3D) mentioned it uses the same vapor chamber design, where did you hear that?

http://www.guru3d.com/articles_pages/nvidia_geforce_gtx_970_and_980_reference_review,3.html

 

[toyota]

Huh, one of the reviews I read (Guru3D) mentioned it uses the same vapor chamber design, where did you hear that?

http://www.guru3d.com/articles_pages/nvidia_geforce_gtx_970_and_980_reference_review,3.html

I have seen it plenty of places but here you go. http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980/5.html


The heatsink no longer uses a vapor-chamber base, which, while it helps NVIDIA reduce cost, also reduces thermal capacity slightly.

 

[Termie]

has anyone seen a review for the EVGA 970 "blower" card? I'm wondering how it would compare to the stock Titan cooler that's used on the 980's

I'm going to guess that based on the short PCB used, this is probably a variant of the original GTX 670 cooler. It is not equivalent to the Titan cooler.

 

[toyota]

Toyota, I am curious if you are going with a 980 or 970?

I have this MSI 970 on the way. http://www.newegg.com/Product/Produc...-833-_-Product

the Gaming one was out of stock but I think this one will be just fine.

 

[boozzer]

It is much more than the clocks though. The right card depends on the setup and the performance of the card, sometimes these two are interdependent. The blower cards are much superior for small, cramped cases and/or SLI configurations. The custom, multi-fan cards are generally better in a more open case, where heat may not be as much of an issue. I would much rather have a blower-type card in a setup where I already am hitting temps on my CPU, for example. Obviously, because of these issues, the 'same' card could hit different performance targets in the same machine, based on how the card is exhausting the heat.

isn't it an accepted fact that if you need a blower type gpu, nv is the way to go? there is no contest what so ever. who the hell wants a monster 60 dba amd gpu in their computer case? I stand by my comment about comparing the best of the best from both sides. a review "only" containing reference 290/290x would be dismiss as PR. 980 so far seems to be great(price/performance wise comparing to 290x), there is no need for any kind of sneaky marketing.

 

[Fastx]

I have this MSI 970 on the way. http://www.newegg.com/Product/Produc...-833-_-Product

the Gaming one was out of stock but I think this one will be just fine.

Congrat's on your new card, I did not think the 970 would be released at this price. I am so tempted to go with a 970 for my own reasons, but can't decide/not sure. I am going to put my Tri X 290 on e-bay as a buy it now for $350.00 (there are some completed Tri X 290 sold for $350 & is what I paid) and if it should sell at that price I will go with a 970. If it doesn't sell I think I am just going to keep it then.

 

[Silverforce11]

has anyone seen a review for the EVGA 970 "blower" card? I'm wondering how it would compare to the stock Titan cooler that's used on the 980's

Thats the short PCB design that we saw on the gtx670 (I had one), its really [inferior] to be honest. If you live somewhere with cold ambient, its ok, but in the Australian summer, it was throttling and running at >80% fanspeed to prevent overheating. It's damn loud, R290X loud.

Profanity isn't allowed in the technical forums.
-- stahlhart

 

[tidy]

.

 

[x3sphere]

I have seen it plenty of places but here you go. http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980/5.html


The heatsink no longer uses a vapor-chamber base, which, while it helps NVIDIA reduce cost, also reduces thermal capacity slightly.

Guru3D seems to be wrong then, thanks.

According to ComputerBase, the true Titan vapor chamber heatsink and the 980's heatpipe heatsink should be physically interchangeable. I'd be really interested to see what would happen if one put a Titan heatsink on the 980. It seems like it could be the holy grail of quiet performance, or do you think the vapor chamber might actually lose efficiency because it's been designed for the much higher TDP?

It's possible the vapor chamber wouldn't be as effective at a low TDP. I doubt it'd hurt cooling performance though. Would've rather them included it, as TDP can get up to around 250W when overclocking.

 

[Cookie Monster]

The bolded part -> I find it just the opposite. Once O/C'd differences in efficiency and heat tend to diminish. O/C'd CPU/GPU is not as efficient and power usage can climb pretty dramatically.

Im not quite sure about that but given the reviews, we see that 20% overclock on GM204s increases given power consumption by 20%. They may at some point converge to the same efficiency i.e. heat but because the starting point is different, I doubt you'll come across that scenario.

A GPU that consumes on average 100W less will show a difference unlike comparable cards with similar power characteristics like the GTX780Ti and R290X (20~40W).

 

[tidy]

.

 

[Attic]

isn't it an accepted fact that if you need a blower type gpu, nv is the way to go? there is no contest what so ever. who the hell wants a monster 60 dba amd gpu in their computer case? I stand by my comment about comparing the best of the best from both sides. a review "only" containing reference 290/290x would be dismiss as PR. 980 so far seems to be great(price/performance wise comparing to 290x), there is no need for any kind of sneaky marketing.

The nVidia blowers are much better than AMD blowers. And I agree, 970/980 should be benched against non reference 290/290x, as well as available OC 780 cards, clearly. That's what folks are going to be deciding against. [H] is pretty good at reviews including established and available cards. In their test they used an XFX non reference. Stock clocks only but the cooler should have kept it from throttling. I'd expect the nVidia review guide "suggested" reference card competition from AMD and non OC'd 780 stuff. Sometimes those review guides leak.

 

[Cookie Monster]

Tom's indepth power measurements using the oscilloscope and current clamps should be more accurate but the execution side of things might be flawed. Heres why:

Current clamps first of all have precision/noise issues. Not sure if its calibrated to give you the right measurements especially for a current load that is always changing. And Im not sure why one would measure power consumption in 1ms division because that is not accurate.

Most components can handle alot higher load when subjected to a transient e.g. in current or voltage in those time periods (us~ms). Most don't know that the maximum ratings often are during the continuous state.

Also he forgets that there are usually noise in those rails. Very high frequency noise is present in those rails and these spikes can usually be very high but not damaging and should be filtered out during the measurement. Im not sure how he measured them but looks like there is no ground reference point with the measurement?? Even if he did, is the ground loop minimized?

These are some basic flaws on the top of my head that I do think renders his results largely inaccurate although I still give him A for the concept and effort. It would have been far more accurate to just directly measure all the input current/voltage using an accurate DMM if it can be accessed (PCI-e measurement might be a little difficult and require something else).

P.S. I think its logical enough to assume that TDP is based on the average power consumption of a product give the only thing a GPU can do with that energy is convert it back to heat. Even with the so called 300W peaks, your temperature doesn't suddenly change in 1ms nor tripping the PSU's current limiter..

Maxwell may have the better power saving techniques but I do think that their underlying architecture does indeed improve power efficiency which I disagree with Tom's conclusion.

Bonus - The power measurements probably aren't what the GPU is actually consuming either because the DC-DC converter that converts the 12V rail to the GPU supply voltage has its own efficiency (and in return produce extra heat along with the GPU!). Say they have a 90% efficiency and the measurements read 160W average. The GPU is likely to consume an average of 144W. So its not as simple as most seem to think. The efficiency of these power converters also come as a cost and from my memory, AMD usually uses more efficient yet expensive designs than nVIDIA.

 

[BFG10K]

The 980 is a fantastic part - especially in terms of perf/watt and perf/mm2 - but I think I’ll wait for the 980Ti. When AMD's 3xx comes out I expect a faster 980, like last time.

The Titan has been a great card but in hindsight it would've been better to get a 780Ti as it was much cheaper and non-trivially faster in many cases.

Based on the 980's specs there are obvious holes with texturing and bandwidth that can easily be plugged, plus they have a lot of power consumption to play with before then hit 250W TDP, which even then is very reasonable for a high-end part. So it could be as easy as raising the core clock speeds by 20%-30%.

Also I play a lot of older games and they could well be affected by the reduction in bandwidth/texturing.

 

[ksheets]

Well, just got back from Microcenter. Unfortunately, they didn't have any of the EVGA in stock so I had to grab the Gigabyte which was $30 more, but Oh well- what can ya do??

Now, the REALLY interesting thing came when I decided to check out monitors. They had a 1080p AOC Gsync that I had never seen before for $449. While I was checking out that monitor, I decided to ask the sale's guy if they had and PG278Q swifts in stock and I'll be damned- They showed ONE in stock and it was, so I bought it.

It's like Christmas in September, Kevin is a happy boy

 

[Attic]

Well, just got back from Microcenter. Unfortunately, they didn't have any of the EVGA in stock so I had to grab the Gigabyte which was $30 more, but Oh well- what can ya do??

Now, the REALLY interesting thing came when I decided to check out monitors. They had a 1080p AOC Gsync that I had never seen before for $449. While I was checking out that monitor, I decided to ask the sale's guy if they had and PG278Q swifts in stock and I'll be damned- They showed ONE in stock and it was, so I bought it.

It's like Christmas in September, Kevin is a happy boy

Very nice.

My local frys had one 970 EVGA acx for 339 (right next to it was 2gb 770 for 319) and one 980 for 569 (with 780ti next to it for 659).

I had to grab my right arm with my left arm to put the GTX 970 back on the shelf.

 

[tidy]

.

 

[Abwx]

Im not sure how he measured them but looks like there is no ground reference point with the measurement?? Even if he did, is the ground loop minimized?

There is no ground loops in galvanicaly isolated probes....

P.S. I think its logical enough to assume that TDP is based on the average power consumption of a product give the only thing a GPU can do with that energy is convert it back to heat. Even with the so called 300W peaks, your temperature doesn't suddenly change in 1ms nor tripping the PSU's current limiter..

To measure rapidly varying currents you need to increase the sampling frequency, it works the same in audio...

Maxwell may have the better power saving techniques but I do think that their underlying architecture does indeed improve power efficiency which I disagree with Tom's conclusion.

At full throughput it has no better efficency than previous gen, to summarize the heavier the GPU loading the lower its efficency.

Bonus - The power measurements probably aren't what the GPU is actually consuming either because the DC-DC converter that converts the 12V rail to the GPU supply voltage has its own efficiency (and in return produce extra heat along with the GPU!). Say they have a 90% efficiency and the measurements read 160W average. The GPU is likely to consume an average of 144W. So its not as simple as most seem to think. The efficiency of these power converters also come as a cost and from my memory, AMD usually uses more efficient yet expensive designs than nVIDIA.

This is the comsumption of the card, if it s VRMs are badly chosen it will show in the result, indeed.

 

[ksheets]

Well, just got back from Microcenter. Unfortunately, they didn't have any of the EVGA in stock so I had to grab the Gigabyte which was $30 more, but Oh well- what can ya do??

Now, the REALLY interesting thing came when I decided to check out monitors. They had a 1080p AOC Gsync that I had never seen before for $449. While I was checking out that monitor, I decided to ask the sale's guy if they had and PG278Q swifts in stock and I'll be damned- They showed ONE in stock and it was, so I bought it.

It's like Christmas in September, Kevin is a happy boy

Well that happiness was short lived- No post from the video card so I'm pretty much left with my 560ti/gd235....but I will have to say the PG278Q looks REALLY nice behind my monitor...I KNEW this was going too good...

 

[Cookie Monster]

There is no ground loops in galvanicaly isolated probes....

Thats if he used those probes with them which are EXPENSIVE. Perhaps he used a differential probe but it doesn't look like it from the pictures. ALL oscilloscopes with their basic probes have a ground reference.. or else you won't be able to measure anything because theres no reference to be measured against!!

To measure rapidly varying currents you need to increase the sampling frequency, it works the same in audio...

It doesn't matter if you increase the sampling frequency if your measurement has noise.. And Im not sure what point your trying to make here because like i said, those supposed peaks that happen in micro-seconds could very well be meaningless.

At full throughput it has no better efficency than previous gen, to summarize the heavier the GPU loading the lower its efficency.

Not sure what you mean here. And you also forget that these GPUs boost so it will utilize all the additional headroom it has under the default settings.

This is the comsumption of the card, if it s VRMs are badly chosen it will show in the result, indeed.

I never said they were bad. All im saying is that there is indeed a loss through the cards VRM stage and it can effectively be between 1~30W depending on load. You won't know its its bad or not til you measure the actual current/voltage through all the VDD pins found in the GPU.

 

[desprado]

Dam GTX 970 is killer for every type of High end gpu competition