The NVIDIA GeForce GTX 980 Review: Maxwell Mark 2
by Ryan Smith on September 18, 2014 10:30 PM ESTSynthetics
As always we’ll also take a quick look at synthetic performance. These tests mainly serve as a canary for finding important architectural changes, and with the exception of pixel throughput we are not expecting any major changes for GTX 980 and GM204.
GM204 is designed to have an ever-so-slightly higher triangle throughput rate than GK110 – 16 tris/clock versus 15 tris/clock, and sure enough the GTX 980 comes out on top in TessMark, slightly edging out the GTX 780 Ti. The difference is only very slight here, and though GM204 should be a bit more powerful than GK110 in practice it’s a dead heat.
Moving on, we have our 3DMark Vantage texture and pixel fillrate tests, which present our cards with massive amounts of texturing and color blending work. These aren’t results we suggest comparing across different vendors, but they’re good for tracking improvements and changes within a single product family.
Beginning with Maxwell NVIDIA reduced their texture-to-compute ratio from 12:1 to 16:1. As a result of this change Maxwell GPUs have fewer texture units than comparable Kepler GPUs. Compounding this effect is the fact that Maxwell CUDA cores are more efficient than Kepler CUDA cores, leading to NVIDIA placing fewer cores overall and further reducing the texture fill rate.
As a result the GTX 980 is not texture fillrate competitive with any of the GK110 cards. It is competitive with the GK104 cards, but only because these cards had the same number of texture units at 128. NVIDIA has told us that they believe this new ratio is a better fit for modern workloads, and judging from the performance we’re seeing elsewhere it would appear that NVIDIA is right.
On the other hand, thanks to NVIDIA’s newer 3rd generation delta color compression technology, our 3DMark pixel fillrate performance is through the roof. GTX 980 comes very close to doubling the throughput of our GK110 cards and more than doubles the throughput of the GK104 cards, reflecting the fact that it has 64 ROPs and more importantly has the available bandwidth to put them to good use.
This benchmark in a nutshell is why NVIDIA can deliver chart-topping performance despite having only 2/3rds the memory bandwidth of GTX 780 Ti. By improving their color compression to this point, NVIDIA can significantly reduce their memory bandwidth requirements Maxwell 2, allowing them to do more with less. In real games the result won’t be anywhere near this remarkable since this is a pure pixel fillrate test, but it goes to show that NVIDIA has been able to expand their effective memory bandwidth in concert with their ROP and shader performance improvements.
Facebook
Twitter
RSS
274 Comments
View All Comments
ppi - Saturday, September 20, 2014 - link
AMD will not beat 980 (they probably could put some fight, but nVidia could always defend it easily, so why do that - it would just dilute prices). What is more important for them, that *on desktop*, AMD can still stay relevant in lower price buckets by offering more performance per $ (while relying on partners for custom open-air cooling and ignoring the power draw disadvantage).Kjella - Sunday, September 21, 2014 - link
You do realize what you said pretty much exactly mirrors what people said about AMD and CPUs a few years back? Just trying to offer value while your competitor is making more efficient chips is a dead end where you're soon so far behind in technology that it's not enough. Nobody wants a 220W CPU (FX-9370/9590) and if AMD needs to pull a 300+W GPU to compete with GTX 980 it'll be equally dead on arrival.ppi - Sunday, September 21, 2014 - link
Not really. When Core2 was released, pretty much entire AMD's lineup was made irrelevant (I still use my 7 years old mid-range Core2Duo and I know that AMD chips were not even for consideration back then). Now the fastest AMD's card is faster than 2nd fastest nVidia offering. Look at TR 2014 HW survey where 80% clearly enthusiasts buy stuff for less than $400. Die sizes are similar. Both companies are fabless and thus have access to the same processes (unlike competition with Intel).AMD of course HAS TO come up with something better than what they have now. And soon. My point was mainly that they should be able to survive this holiday season sort of okayish.
I expect that AMD is focusing their limited resources on 20nm part, but it apparently did not work as well as it did in times of HD-5000 and 7000 series. And Maxwell improvements are greater than what is achievable just with die shrink. So there's some hard work for AMD ahead. Given necessary lead time for such products, I doubt 300-series will be good enough (unless they were going nuts with efficiency after seeing 680).
I admire nVidia for a long time always covering weak spots in their products. It could be seen from times when they went against 3dfx, though FX-5000 and now of course they show how they learned from 480 era.
Silma - Saturday, September 20, 2014 - link
I fully agree.As long as Intel does not succeed better in smartphones & tablets, it probably doesn't fully utilize its manufacturing capacities.
It could begin with opening 22 nm to NVIDIA and 14nm in 2015.
Seriously though, I'm not sure why Intel still hasn't bought NVIDIA, except if it foresees troubles getting the deal accepted with regulators.
This would not Mirror the AMD's ATI acquisition. crap + crap = crap.
Outstanding + outstanding = awesome.
Notmyusualid - Saturday, September 20, 2014 - link
+1SanX - Sunday, September 21, 2014 - link
Intel should buy NVIDIA long ago but they are in lethargy all last dacadeRyan Smith - Thursday, September 18, 2014 - link
BTW, before anyone asks: we're still working to get images and charts in. 4 days is very little time for a 20K word article. So please hold on for a bit.boot318 - Thursday, September 18, 2014 - link
Where is the Overclocking results? Not done yet? I see the page but it is blank.RaistlinZ - Thursday, September 18, 2014 - link
Ditto. I can't see the overclocking page.chizow - Thursday, September 18, 2014 - link
And no 970 results?