How is that the 970 has so many capacitors on it and the 980 has even fewer than reference?
Both ASUS and MSI have huge rows of caps on their custom design 980 GPUs, what could it be about the Gigabyte G1 Gaming GTX 980 that causes it to have so many less capacitors on it yet still have a comparable amount of phases?
EVGA's Classified models also have a long row of cylindrical capacitors on the PCB, their ACX 2.0 models use reference as pictured above, their FTW model uses reference with the blank pair of phases on the reference design populated with components.
If anyone can explain to me what Gigabyte is doing so differently from the others I'd be fascinated to understand, as they use a comparable amount of components to the other OEMs on their GTX 970 card, but it is a very different picture with their GTX 980.
Some manufactures use some chokes that employ a doubler on each choke. This is how msi has some 36 phase motherboards. They have 18 chokes, but with doublers end with a 36 phase design.
It isn't necessarily cost saving, although it might be a little cheaper.The 900 series cards use so little power that it doesn't really require that many phases. With the 900 series cards your going to be limited by the silicon lottery before the power delivery limits you. The stock pcb's aren't really going to even hinder you on air cooling. On water cooling with bios hacks to allow more voltage, you might be limited by lesser power delivery but it is still unlikely.
Here is the real deal here.. Gigabyte has less power phases than the rest of the cards and I am unsure why. I think they are using chokes to double or aid the power phases.
In fact if you look at it.. the gigabyte cards actually just have the power phases distributed about the card a bit. but to my knowledge inherently gigabyte cards need bigger coolers because there are less VRMS for the same power distribution. It is not a bad thing. They also distribute it.. Count the caps and mosfets youll find that gigabyte has 2 less so its 16(+4 VRAM) Power phase vs 18(+6 VRAM) and even if they double it with chokes to make 32.. the difference is negligible. The reason the DISTRIBUTE the VRM here is because they have odd placement of the memory chips. Its just a different layout and it has little effect on the card. As to whether its more or less efficient I can not say. I would say less VRMS for same amount of power is less efficient BUT they maybe using higher quality chokes and inductors My question is what exactly do you want to know here?
Thanks for the response, what I am trying to do is nail down which OEM has built the best card, the Gigabyte is the most expensive where I am, the ASUS closely behind it, then MSI and EVGA are tied on price.
Basically I want to know if the Gigabyte card costs more because it has a better power design, or whether the reason for the premium is unrelated.
The most popular selling card here for 970 and 980 is by far the MSI model.
(the concept of doubling is new to me and is something I have learned today, I will learn more about it.)
So you are saying, that the EVGA card that has reference componentry is superior on a technical level to those custom designs of MSI and Gigabyte + others, or are you looking at it more from an actual chip quality perspective or meaning EVGA custom layout like Classified is superior to the other custom designs?
I have wondered why the custom circuits of other OEMs are necessary for air cooled setups, perhaps for longevity? But then EVGA offer longer warranty than any of them..
Its mostly about overclocking and a couple added features and software compatibility with the VGA bios stuff.. Theres not too much difference.. the EVGA custom layout is the best with NVIDIA
A lot of the Evga Acx 2.0 coolers have been kind of crappy. There have been a lot of people with Acx 2.0 cards with good airflow thermal throttling. I'm really starting to feel they're losing there top spot of Nvidia cards.
Its not because of the PCB design... its the cooler... I liquid cool my graphics cards so I dont bother with custom.. OEM is the way nvidia meant it.. so get a reference and liquid cool it.. if you dont go liquid then aftermarket is better.... Thats the gist of it
I wouldn't say that the reference pcb is the way Nvidia meant it be used. With the Titan's I would agree since they won't allow custom coolers on those. But with the normal stuff like the 980, 970, and 960, I think of it more like the reference pcb is the baseline pcb.
well honestly.. we are talking marginal difference with PCBs when liquid cooling. I have done so much that I can say its negligible... but honestly.. ASUS and EVGA SSC cards are the way to go with nvidia.. I know id choose ASUS but thats me
EVGA do a set of ACX 2.0 cooled cards with the reference PCB, even overclocked ones which have higher boosts than the custom offerings from other OEMs that even have custom PCBs in addition to the cooler, so clearly this is achievable with the Nvidia designed PCB.
Given this is the case, why do the other OEMs spend money bathing the PCB in extra components when they could just design their own cooler and use the reference componentry? They advertise that it provides a better environment for overclocking (stability etc) but really the actual chip itself is the more important part because the BIOS voltage lock is the same on all of them...
Just trying to seperate what is marketing from what is actually necessary hardware, using reference layouts with overclocked cards doesn't seem to be causing EVGA any problems.
Well thats actually a simple answer.. They do that because there customers and enthusiast have come to expect those features... to exclude them would mean losing customers.. In reality its all about How can i be unique and sell this card more than others
This is down to the silicon lottery, not due to the cooler being better or worse. There are cards with the stock Nvidia cooler that boost higher then many acx and dc2 coolers will. There are golden chips that will be amazing no matter the cooler.
If you really want to know the answer to this question, you will either have to find someone who has the knowledge of the various components of the VRMs, or learn it yourself.
The latter shouldn't be too hard, though it may be time-consuming. In this case, you quest may be finished just by looking up the capacitors.
This is something I am interested in, as I am in the market for a GPU upgrade as well.
So in days past, extended VRM circuitry may have been more beneficial than it is now? I had a very brief conversation with JJ from ASUS in the youtube comment section of the ASUS Illuminati channel about this, and that with Nvidia voltage locking GPUs and the power use being lower, it is actually no longer as necessary or beneficial to over build PCBs on GPUs, yet it is still happening because it had previously become the precedent.
What you have just said does make a lot of sense, for instance what is "MSI Military Class Component" that sounds like drivel to me in all honesty, but I can't objectively prove or say so because my knowledge isn't high enough.