Nvidia GeForce GTX 1080 Ti is "world's fastest gaming GPU"

[Tunnah]

As best I can gather from leaks available, R7 1700X is a few percentage points (~5%?) behind 6800K in games specifically. That makes it an upgrade from 2600K with cores available for future proofing, if we don't factor in overclocks. Besides, just as fewer, faster cores work out better in current games, Vulkan and plausibly well-written DX12 games may yet turn that on its head. Speculative, yes, but don't poo-poo yet. Unless you prefer GTA V to any other game available.

That's at stock. From what I've read, they don't OC all that well (compared to Intel). From what I've seen you're looking at a 4.2ghz max, if you get a great bin. Typically with a many core chip you are limited with OC capability, and games don't take advantage of many cores (practically all games are console first, and even if not, the GPU does most the work load nowadays), so whilst the high scores might go to Ryzen, the actual best performance will go to Intel.

However, what we're talking about is..nothing. We're talking about 5, 10 frames, tops. Once you have a high end CPU (Core i5-2500K would be high end in this case), any CPU upgrade is just wasted money. You have to spend About £500 to get a 5% performance boost, which is just straight up window-licking-retarded. Put the money into a GPU!

I was actually going to get a Ryzen, but there is nothing that I do what will benefit from the cores. I recently moved from a 2500 to a 2700K, and the performance gain was MASSIVE. It was the biggest performance boost I've seen in a decade. And it was mostly because I had a none-K, I went from 3.5ghz to 5ghz. For example, Civ VI on the i5 I had to listen to the whole character intro, and then an extra 5s. With the 2700K it doesn't even begin the character-specific words. For anyone who hasn't played Civ VI, when loading you get 10s of generic speech before starting the character specific words. I went from 60s to less than 10s. Ryzen isn't going to hit 5ghz, so I reckon it'll be a step back.

[Tunnah]

Not really like a math co-processor at all: it's an extension of the memory controller than can address, and therefore cache, data from a much wider address space. It appears to be part of the GPU, rather than a separate device*.

From AMD's slides it looks like it can also drop data directly from system memory or non-volatile storage directly into L2 cache for use by the compute engines, which could in some circumstances increase performance by reducing latency if data doesn't have to go into VRAM (although that's speculative on my part, tbh ).

*This is based off the images in AMD's slides, although those images also appear to show the High Bandwidth Cache as being part of the chip, while most sources assert that the HBC is simply a fancy name for the on-interposer HBM2. If the cache is something different to the on-package HBM2 that would make the card even more interesting...

You need to stop being so technical. I feel like I'm reading el reg you big brained bugger

[nobodyspecial]

I don't like the way Nvidia are soaking up the early adopter cash by themselves and then allowing it's partners to release their own cards. If i was a partner of Nvidia this would heavily damage my relations with them everytime they do it

Cards sales are up, so they are just taking the top bit of excess for the rich who want to be first on the block (1st month) to own one. Who cares. If that pays for Volta to be the best it can be, bring it. I can wait. Just like I'll wait for the RIGHT motherboard (after reviews, seems they'll be coming by the dozen this next few weeks) and see if chip prices shakedown a bit. We already see competition bringing Intel down some, and obviously it looks like Nvidia is about to give more for either less, or more for more. I'll take it all, just have to see where my budget makes the most sense to spend. Looks like AMD (1800x) but I'll wait for them to respond possibly to Intel cuts (I wouldn't mind $400-450 for whatever I buy) depending on benchmarks no doubt. I also want to see Vega and if Nvidia does more than just better cards (maybe a cut too if Vega is really GOOD). AMD probably has me by default though, for 2 reasons. 1st, I have wanted to buy a chip from them for ages (I was a reseller for 8yrs and pissed about how Intel screwed them for 60B over 10yrs and still hasn't paid the 1.5B yet AFAIK, still appealing ruling), and 2nd they actually win quite a bit in handbrake that I'm using to rip all blurays to now (my collection). I hope they get me for vid card too, but that will depend on the money I see them make on the cpus (GPU drivers must have enough $$$ to improve like Nvidia) so I can afford to take what I'd call a "CHANCE" on gpu drivers. I know the chips from both sides are good for gpu, but I'm interested in DRIVER R&D. I have no problem buying cpu since it doesn't depend on monthly drivers etc, but gpu is different and Gsync is better (even though my ideal monitor isn't out yet, here comes gsync 2 though...LOL). That said I hope I'm ALL AMD in 3-6 months or by xmas.

EXCITED! Competition is awesome, but I'm willing to pay for whatever. Not asking for price cuts that kill AMD profits. Hopefully AMD comes out with a billion PROFIT NET, in the next year and again for 3 more. I'll gladly eat more or adjust my budget rather than ask AMD to kill their new products begging for price cuts. I'd pay it to Intel or NV anyway, so ASK WHAT YOUR PRODUCT IS WORTH! I'll adjust accordingly my own budget! NO FAVORS. Make whatever you can AMD!

[scaryjim]

As a side note, are they really going to be $500 USD ? If so, I'm going to try to find me a friend in the US, because that's £100+ cheaper than here

They're going to be $500 before sales tax. Once you add on the 20% VAT we have to pay and convert at today's horrendous exchange rates they come out to around £480, so we're actually paying about the right price. US states have different sales taxes so the US price always gets quote without taxes, and then you've got to think about shipping costs and import duties.

In other words, just buy one from a reputable UK retailer

EDIT: I assumed you were talking about Ryzen here, but this is the nvidia thread ... bit of cross posting going on?!

[Andy14]

EDIT: I assumed you were talking about Ryzen here, but this is the nvidia thread ... bit of cross posting going on?!

I think the Nvidia thread ended a long time ago. Somewhere on page 2 I believe.

[Ozaron]

That's at stock. From what I've read, they don't OC all that well (compared to Intel). From what I've seen you're looking at a 4.2ghz max, if you get a great bin. Typically with a many core chip you are limited with OC capability, and games don't take advantage of many cores (practically all games are console first, and even if not, the GPU does most the work load nowadays), so whilst the high scores might go to Ryzen, the actual best performance will go to Intel.

However, what we're talking about is..nothing. We're talking about 5, 10 frames, tops. Once you have a high end CPU (Core i5-2500K would be high end in this case), any CPU upgrade is just wasted money. You have to spend About £500 to get a 5% performance boost, which is just straight up window-licking-retarded. Put the money into a GPU!

I was actually going to get a Ryzen, but there is nothing that I do what will benefit from the cores. I recently moved from a 2500 to a 2700K, and the performance gain was MASSIVE. It was the biggest performance boost I've seen in a decade. And it was mostly because I had a none-K, I went from 3.5ghz to 5ghz. For example, Civ VI on the i5 I had to listen to the whole character intro, and then an extra 5s. With the 2700K it doesn't even begin the character-specific words. For anyone who hasn't played Civ VI, when loading you get 10s of generic speech before starting the character specific words. I went from 60s to less than 10s. Ryzen isn't going to hit 5ghz, so I reckon it'll be a step back.

Before I start, sorry Jim, if crossposting is a crime then I'm about to vadalize this thread...

Compared to 2600K and 2700K, Ryzen should feature at least a 15% up to plausibly ~23%ish IPC advantage. So, you can eliminate a fair proportion of the extra clock speed available to 2600K as null in performance terms. Then we've got real world scenarios - "high scores go to Ryzen, real world goes to Intel" - this more than likely won't be true on a hexacore / octacore Ryzen chip. In the real world, you aren't *just* running your game, but also OS, background apps, probably a web browser, etc. Would you gain performance further loading the 4 cores you have tapped on the game, or gain it by tasking these applications to the spare 2-4 you have kicking about idle, IF the game in fact doesn't use them?

Which brings us to the next point: XBone and PS4 both have, as best I can surmise, 8 core APUs. It's fair to assume that Scorpio and the Pro probably have the same or greater. That also means devs will be designing games for more threads, starting... a while ago, actually. I'm sure 8 thread Intel chips have always benefited from "right place right time" but 8 genuine cores will always perform better than 8 threads. Along with Bethesda agreeing to use Vulkan, Doom being a posterchild of high framerates on inexpensive GPUs and DX12 being forced onto everybody I think it's fair to assume the jig is up for lower core parts fairly soon.

Going completely off topic, is the AI in Civ VI any good yet? Last I checked they were all about zerg rushes and naturally hating you because you existed as a human being.

[Andy14]

Going completely off topic, is the AI in Civ VI any good yet? Last I checked they were all about zerg rushes and naturally hating you because you existed as a human being.

You are in a 1080Ti Thread discussing Ryzen. Wouldn't worry about going off topic ;-)

[Ozaron]

You are in a 1080Ti Thread discussing Ryzen. Wouldn't worry about going off topic ;-)

Well... I mean, at least it was hardware on a hardware thread.... ;-)

[scaryjim]

Before I start, sorry Jim, if crossposting is a crime then I'm about to vadalize this thread...

It's not a crime, but we probably need to get this thread back on track at some point - there's plenty of Zen threads out there already

But, having said that ...

... Which brings us to the next point: XBone and PS4 both have, as best I can surmise, 8 core APUs. It's fair to assume that Scorpio and the Pro probably have the same or greater. That also means devs will be designing games for more threads ...

Hmmm, yes and no. The consoles you mention above - and as far as we know their immediate successors coming this year - use 8 of AMD's small "cat" cores - these are lowish IPC (roughly equivalent to Piledriver) and also fairly low clocked. I also believe that developers can't necessarily use all of those cores for game logic - some are reserved for system use. Finally, programming for consoles is nothing like programming for PC, as you have extremely low level access to the underlying hardware which allow much tighter optimisation.

So while developers will become more adept at parallelising their games and engines, it's not as simple as saying "8 core consoles will make games use at least 8 threads". And it'll have less impact directly on PCs, as modern CPUs are pushing 3x as fast per core than the consoles, so can time slice on a lower number of cores and still achieve the same performance. Pull up the "thread" column on Task Manager some time ... you might be surprised at just how many threads some of your common programs are using...

[Ozaron]

Hmmm, yes and no. The consoles you mention above - and as far as we know their immediate successors coming this year - use 8 of AMD's small "cat" cores - these are lowish IPC (roughly equivalent to Piledriver) and also fairly low clocked. I also believe that developers can't necessarily use all of those cores for game logic - some are reserved for system use.

All true, but it's all about the shift in focus of developers, not the actual tech specs of those consoles. We don't really know what will be in the new generations either, beyond that it will be "better than last time".

So while developers will become more adept at parallelising their games and engines, it's not as simple as saying "8 core consoles will make games use at least 8 threads". And it'll have less impact directly on PCs, as modern CPUs are pushing 3x as fast per core than the consoles, so can time slice on a lower number of cores and still achieve the same performance. Pull up the "thread" column on Task Manager some time ... you might be surprised at just how many threads some of your common programs are using...

Never said it would be this simple. (Though that would be pretty nice, right?) Have to admit, as someone who occasionally looks into emulating old games I feel very sad that my 6600K has a fight on its hands to recreate True Crime NYC for PS2 at correct 60FPS framerate. Difference between direct-to-metal and top level amateur software in utilization efficiency is mind boggling.

[DanceswithUnix]

and Gsync is better

Really? What have I missed?

When FreeSync came out there were obvious flaws, notably lack of low frame rate compensation when you dropped off the bottom of the variable sync range and the lack of full screen windowed support. They were fixed.

The one bit of G-sync that totally put me off also got fixed, the way that early monitors were only allowed to have the single G-Sync port and any HDMI etc ports of non G-sync designs were removed on the G-sync version. The one nice feature of early Freesync, having a choice of what happens when you go faster than the frame rate upper limit, was supposed to be added to G-sync so I presume that has happened by now?

Now if you buy one of the £90 hdmi FreeSync monitors then you won't get all the features, but if you are comparing like for like on similar monitors the two systems seem like feature identical to me.