GPU Buying Guide - (Read first post)

Since Fury was about 15% slower than the Titan at launch, I guess they've managed to close that gap a bit through driver updates.

crimsonstag84 said:It’s just like how Fury was about 15% slower than Titan at launch; they've managed to close that gap a bit through driver updates.

The reality is that Fury actually boasts significantly higher GFLOPS than Titan (8,602 vs 6,144), which completely contradicts the logic used by electricharbor30—his whole "GFLOPS matters" argument. 😬
Nobody actually looks at those numbers except him (and maybe a few console gamers).

EDIT

If we use the HD7770 as a benchmark, the GTX 750 OC is actually 30% more powerful than the GTX 460, according to this link, and this link. Obviously, the games used in both summaries aren't identical, but you get the point. Someone who doesn't know any better would be seriously misled if they relied solely on that site. Newer titles can increasingly penalize the aging GTX 460 due to driver support—in fact, even the newer Kepler cards have been left behind.

Timothy Kim60 said:I mean, it's basically about 20% faster in actual gaming scenarios (+/-), but according to those links you sent over, it's supposed to be weaker, even though its "gflops" rating is like 5% higher. So yeah, that's why

So, what you're trying to tell me is that the older chip is actually faster? I guess I'd need to see an actual link to some test to believe that.

I just went ahead and did a quick Google search and stumbled onto this: http://www.tomshardware.com/reviews/...rchy,4388.html

And based on this, even the 256-bit version is ranked lower...

🤦
what's even being argued about here

electricharbor30 said:So, you're actually claiming an older chip outperforms the newer one? Do you have a benchmark link I can check out?

Which older one? I have no idea what you're talking about. 🤣

BTW, I updated my previous post. It turns out the GTX 750 is even more capable than the GTX 460 than I initially thought.

Timothy Kim60 said:The whole thing is that Fury has way more GFLOPS than Titan (8,602 vs 6,144), which totally flies in the face of electricharbor30's little "GFLOPS matters" theory. 😬
I mean, nobody actually cares about that besides him (and maybe some console gamers here and there).

EDIT

Even if we look at the GTX 750 OC being about 30% faster than the GTX 460—using the HD7770 as a benchmark click, click—it’s pretty clear. Sure, they didn't use the exact same games in both summaries, but you get the point. I guess someone who isn't clued in and just follows that one site would be seriously screwed. Newer games might even end up hurting the old GTX 460 even more because of driver support (I think even the newer Kepler cards have been kind of neglected lately).

http://www.gpuzoo.com/Compare/ASUS_G...4260183361190/

It basically works out that the OC version is roughly 20-30% faster....so, what's your point?

If you've got 1164 GFLOPS on a 750 OC versus 940 GFLOPS on a 460....how much of a difference is that really?

Man, just go take a nap or something, please...

electricharbor30 said:That's 1164 gflops on an overclocked 750 compared to 940 gflops on a GTX 460.... what percentage is that?

Nice job. You found one instance where theoretical numbers actually line up with real-world performance... and you still don't have any actual game benchmarks. 👏

crimsonstag84 said:Nice job, you actually found one single instance where theoretical numbers might actually line up with real performance... and yet, you still haven't provided any actual, concrete gaming results. 👏

Look, nobody has perfectly precise performance numbers when it comes to gaming, because honestly, there are like a million different variables involved... I mean, I never said anyone should follow these specs like they're gospel, but it’d be pretty helpful to have some idea of what kind of hardware you're actually buying and what it's supposed to do... specifically, with the GPU, at least you know something, but everything else is just a total guessing game... you've got drivers, your motherboard, your RAM, your CPU, your monitor's refresh rate, how much background stuff the system is running, the OS itself, OS settings, even the game can behave differently on every single card... that's why I don't really bother looking at those random online benchmarks that some guy ran under who-knows-what conditions, probably not even professionally and maybe even biased... usually, I think you just gotta go straight to the manufacturer's website and read the data for the specific card you're getting, and normally, a ton of that info is buried, so this GPuzoo thing was actually super useful for me... I used to rely on Wiki, but that's public and some idiots out there just change the data whenever they feel like messing around...

Reviewers test hardware under standardized conditions. That’s the absolute foundation of everything they do. They explicitly list the specific hardware used, the software version, the drivers, the settings, and the resolution. If someone's CPU is bottlenecking the system, or if their power supply is too weak for the build—that isn't the reviewer's fault.

You mentioned recently that you don't trust a single website or reviewer—wtf? As if everyone is out there specifically plotting to screw people over. And just as a side note: if you want to switch your definition of performance from texel rate to pixel rate one day, or GFLOPS the next—be my guest, but stop confusing everyone else with those half-baked arguments.

Timothy Kim60 said:Reviewers test hardware under identical conditions. That’s basically the whole foundation of the job. They list the exact hardware used, software versions, drivers, settings, resolution... If one guy’s CPU is bottlenecking his rig, or if someone else’s power supply is way too weak, that’s not on the reviewers.

You mentioned recently that you don't trust a single website or reviewer—wtf. Like, everyone does this just to screw people over. And the way your criteria for performance shifts from texture fillrate yesterday to pixel fillrate tomorrow, or GFLOPS today—be my guest, but maybe don't confuse everyone else with all that.

well, I didn't even really get what all that stuff was actually for myself, I mean, based on analyzing data in my head, I somehow concluded that GFLOPS were the most important thing... and I was right... then I started trying to calculate how you even get to that number and it turned out to be pixel rate times 48 clock per cycle... but then some idiots on Wikipedia changed the GFLOPS info, so I didn't know what to follow anymore, and then I stumbled upon an official document stating that clocks per cycle aren't even the same for all cards... and then I eventually found this site that cleared up all the murky stuff in my head regarding graphics... it explains everything perfectly and now I know exactly what does what... the Titan X will be faster at 4K than the Fury X because it has a higher pixel fillrate, but at lower resolutions, the Fury X is gonna eat it for breakfast thanks to its GFLOPS... obviously, assuming they're in the same conditions and both have decent drivers that can squeeze the max out of both GPUs in some unbiased game... it's all laid out clearly... and things will look better on the Fury X because of the higher texture fillrate... which has always been the hallmark of ATI, anyway, images look better there compared to NVIDIA... and as for reviewers, hey, believe them if you want, but I won't...

electricharbor30 said:Look, I didn't even know what any of that was for myself... and honestly, if you want to trust those reviewers, go ahead. I won't.

Fine then, run all the numbers yourself, write up a whole blog post about it, and let us know how it goes.😉

So, I’ve got a bit of a dilemma regarding the R9 380. I'm looking to pick one up, but I’m stuck choosing between the 2GB and the 4GB models. I've been digging through countless forums and reviews, and honestly, everyone seems to have a different opinion on the matter. It's getting a little confusing. My main question is: is this card actually powerful enough to make use of more than 2GB of VRAM—say, at least 3GB? I really don't want to drop an extra $30 or $40 on the 4GB version if the hardware isn't going to actually utilize that extra headroom. I'm planning on playing everything at 1080p.

swiftfalcon89 said:So, I'm kind of stuck in this dilemma regarding the R9 380 card. I'm planning on picking one up, but I seriously can't decide if I should go for the 2GB or the 4GB version. I feel like I've read a million different debates and reviews online, and honestly, everyone seems to say something completely different. So, is that card actually beefy enough to make use of more than 2GB—like, maybe at least 3GB? I really don't want to waste an extra $30 or whatever on the 4GB model if it's just not going to utilize anything beyond 2GB. I'd be playing at 1080p resolution.

It’s pretty much the same logic as having RAM in your PC, I guess... like, someone might have 4GB, and once that fills up, the system starts hitting the hard drive, whereas someone else has 16GB, but if they're doing heavy photo editing where those RAW files are absolutely massive, even that might feel too small because drives are slow, so you're just sitting there waiting forever just to load a single shot... and usually, you've got a ton of photos to get through...

Following that same idea, games basically stash their textures in the VRAM, and typically every time you enter a new area in a game, it gets loaded into the VRAM... so, if you're playing these massive open-world games, or if you're cranking up the texture settings to high res—where you might need 4GB of textures ready to go—then yeah, you'll definitely see the benefit; otherwise, the system will just keep pulling everything over 2GB straight from the disk...

IMHO, like 99% of games nowadays are designed to use smaller areas, mostly because it isn't just about the textures, it's also about the server load when you're playing online, so you probably don't have a super realistic need for more than 2GB right now... though, you might stumble upon maybe three or four games, if not immediately, then maybe in a year or so, that actually could use it... I guess you just have to judge for yourself based on what you play... just check the system requirements for your specific games...

electricharbor30 said:Look, games stash textures in VRAM. Usually, every time you hit a new area, it loads more stuff up. If you're playing massive open worlds or running everything at max texture settings—where you might need 4GB just to hold it all—you'll see the benefit. Otherwise, if you don't have the headroom, the system ends up constantly pulling over 2GB straight from the disk...

You didn't actually give him any real facts.

People have run tests on this already, so check it out and see what those extra 2GB of RAM actually do.
http://www.eurogamer.net/articles/di...2gb-4gb-review

The 380 isn't some beastly card. I doubt he’ll even touch that 2GB limit because, honestly, with non-high-end titles at 1080p, he'd be better off just turning down the AA or something. That’ll boost his FPS and lower the VRAM usage anyway
(let's be real, if he actually wants decent FPS)

By the way, the GPU isn't going to fry the hard drive; it'll just borrow from the System RAM
Just look at the attached (I'm running a 2GB 7850)

There are all sorts of games out there, but here is one absolute beast that pushes both RAM and VRAM to the limit: click

VRAM Put to the Test: 2GB vs. 4GB GTX 960 and R9 380, 4GB R9 290 vs. 8GB R9 390

In a handful of the tested titles, you can see a slight difference in FPS—though they didn't measure frametimes, where I suspect the gap would be even more significant in those specific games. It’s quite interesting to note that the extra 4GB provides a much larger boost for the R9 380 than it does for the GTX 960 (likely due to the lower bandwidth on that card)

wiredotter14 said:anyway, the GPU isn't gonna fry your hard drive or anything, it'll just borrow some System RAM, which you can see right here (I'm running a 7850 with 2GB)

And I guess it'll just borrow system RAM from the hard drive too...

Quick question for you all—I’m thinking about calling this guy who’s selling a used graphics card. He claims it still has two years left on the warranty, but he originally bought it over in Germany.

When I brought up the warranty, he suggested that if the card actually fails while it's still covered, I could just take it to a local service center here in the States. I guess that sounds plausible, but does anyone actually know how that works? Is there any way that's legitimate?

silvermoose14 said:Quick question—I'm planning to call someone selling a used graphics card that still has two years of warranty left, but it was originally bought over in Germany.

When I asked about the warranty, he mentioned that if it breaks during the coverage period, it could just be taken to a local service center here? What kind of setup would even allow that? Does anyone know how this works?

If there's a manufacturer's warranty, it should work! For example, I once bought a Samsung TV at MediaWorld, and after just two weeks, the motherboard died. I took it back to MediaWorld, but they told me they only handle returns within the first seven days; after that, I had to find an authorized Samsung service center myself. I found one, brought my receipt and warranty, and they swapped it out in just two days. It’s likely the same deal with products from Germany, but don't take my word as absolute law.

So, if it's an NVIDIA GPU, I suppose you could just take it to a repair shop that specializes in ASUS hardware?

It's one of several options out there—I see plenty of them popping up on Google searches...
Just a heads-up, I'm not entirely sure how things are handled over in Germany. For instance, the second link lists some specific retailers, but that might not be the full picture. It’s always best to shoot them an email just to be absolutely certain!