****Official OcUK Fury X Review Thread****

[pmc25]

I'm investing in second year 16nm, the first shrink is usually a little buggy, would like to do the same with CPU, Zen looks to be 2017 now and intel's 10nm is delayed also, I'm thinking this is the end of regular die shrinks. I may buckle and get a 2011, but I want to see how DX12 changes the landscape.

Zen is definitely H1 '16. Zen APU is (probably) '17.

 

[GoogalyMoogaly]

We going by information we have ready for us now.. mantle is still going to be around its out of Beta and now running with Vulkan api.. Again something you should already know.

Mantle set out to try and be an open standard that was AMDs idea and they have succeed in doing so.

Vulkan =/= Mantle. That's what I know. Vulkan may be based on Mantle specs and use some of the code, but they're not 100% identical. The 390X probably shares a lot of it's architecture (GCN) with an R7 250, but I'm not going to use the 2 interchangeably.

 

[pmc25]

The entire point of the APIs is to use features that only AMD support?

They have a vastly superior uarch.

 

[setter]

We going by information we have ready for us now.. mantle is still going to be around its out of Beta and now running with Vulkan api.. Again something you should already know.

Mantle set out to try and be an open standard that was AMDs idea and they have succeed in doing so.

Are you still running xfire mate, 290 iirc. We compared fps and frame times on bf4 a lot. At that time i was running sli 780's. I did try mantle on a single 290 that a relative owns. And i was very impressed, particularly as he had a lower cpu than mine, intel 3570k. Will you be purchasing the fury x yourself?

 

[GoogalyMoogaly]

They have a vastly superior uarch.

Things like Mantle weren't supported by too many games, neither was PhysX, why all of a sudden will developers use features that only AMD supports?

Do we know there's nothing that Nvidia cards do better than AMD cads that would give them a performance edge? Or things that Nvidia can do that AMD can't? (again it'd seem odd for developers to support these things)

 

[Boomstick777]

I'm investing in second year 16nm, the first shrink is usually a little buggy, would like to do the same with CPU, Zen looks to be 2017 now and intel's 10nm is delayed also, I'm thinking this is the end of regular die shrinks. I may buckle and get a 2011, but I want to see how DX12 changes the landscape.

Yeah if you can wait then go for it. If only I could do the same

Zen 'APU' is 2017. Zen 'CPU' is 2016, I think on AM4 with DDR4. The focus is on desktop CPU first then roll it out to APU the year after.

If you went X99 you wouldn't regret it, some good price on the mobo's. Get a good deal on a 5820K and some DDR4 and you are set.

 

[shankly1985]

Are you still running xfire mate, 290 iirc. We compared fps and frame times on bf4 a lot. At that time i was running sli 780's. I did try mantle on a single 290 that a relative owns. And i was very impressed, particularly as he had a lower cpu than mine, intel 3570k. Will you be purchasing the fury x yourself?

Yeah but just tuck one out to put stock cooler back on. GPUs going to family and I'll be getting furyx soon as stock is available.

 

[setter]

Mantle worked good, when bf4 worked that is. It was a terrible choice of game to show off any new tech. Bf3 had no backing from nv or amd and it was a buggy pos. Thats down to dice and ea pushing deadlines too much.

 

[Nutella33]

Argh thank you . I'll probably wait for the MSI ones never had issues with their RMA on the one card that died.

Does anyone know why the asus one is so much more than the rest? As far as I can tell it's exactly the same as the others with 3 year warranty .

Yeah. I ordered a cheapy one at 12.13 so hopefully I'm near the top of the line when they come in! (Sapphire)

Standard Asus pricing. They price themselves a bit like evga do but without any of the bonuses evga have. Gibbo did say their RMA was very good now but can't say I'd pay the extra

 

[setter]

Yeah but just tuck one out to put stock cooler back on. GPUs going to family and I'll be getting furyx soon as stock is available.

Ahh, so you went aio cooling on them mate, heck i had to sell my old 780's quick because family would have stole them lol.

Ive got an msi 980ti coming in the next few days, will be weird having one card again tbh. But it will be benched, and gamed on pretty hard. Tbh its a good time for us all as gamers, new cards, new games, (albeit buggy lol). I think the fury x will improve overa short period of time, it's a new card with new very unknown memory, but perhaps a good indication of what the future will bring. Good too see a bit of competition in the gpu market.

 

[D.P.]

Then there are number of other issues:

No HDMO 2.0 for 4K TVs - true
No DVI for older monitors - true
IDel noise is higher than TX/980TI - change the fan setting, btw its quiet, and more quiet at load as the TX/Ti
Not much sign of overclocking head room - wait for OC software updates
only 4GB VRAM - which is enough as it seems
The card get pretty damn hot even if the die is cooled by the AIO - the card is cool, the only thing gettig hot is the VRM, and that is only in Furmark stress test..in gaming its 60C as Tomshardware measured it
Performance is all over the place which means it is hard to predict how it will play future games. - which predicts it needs some driver ironing..new tech
Stock is very limited and supposedly will be for some time so prices have been inflated. - it could be ordered cheaper than everyone thought £509
AIO can be a pain for multi GPU use - in a halfly decent case like my 690II i could install 3..together with my H110
Memory clocks locked - no need to OC enough bandwith on stock

The idle noise is not from the fan but from the pump which you can't change.

If the FuryX was a great locker it would do OK on stock volts IMO, and would be drawing less power at stock.
I have yet to see good benchmarks of games that pull more than 4GB but I have seens very poor GTAV results at ultra settings which is indicative.


No, the whole card gets really friggin hot because there is no fan blowing over it, one of the hottest cards ever made!

The erratic performance is not a sign of bad drivers, AMD had a log time to write drivers and the architecture is very similar to hawaii and tonga. The performances differences are mostly due to architectural reason,e .g. they bumped up tessellation performance quite a lot so games like Witche3 with hairworks on show a great boost in performance, other games that are more texture limited for example show much less performance, because nothing changed there (same umber of TMUs). Besides which, how many drivers have AMD released in the last 6 months?

 

[D.P.]

Mantle (and by extension DX12 & Vulkan) was entirely about massive parallelisation (GCN is great at, pre-Pascal is poor at), independent workload for individual shaders (GCN is good at, pre-Pascal is completely unable to do), making much better use of max throughput (something GCN has a lot more of) and increasing draw calls, and reducing CPU overhead hugely (AMD had more - though apparently new drivers change that).

NVIDIA can't compete with AMD on the new APIs until Pascal rolls around .. except in NVIDIA backed titles where the GW gimping ****-show will continue.

Is this some kind of epic troll post or OCUk Joke I wasn't told about

All Graphics APIs are based on massive parallelism and the hardware reflects that, which is why all GPUS are nothing more than massive parallelized processors, and why we got things like CUDA and openCL. DX is based on parrlelism, it is just based on a legacy command queue.

Pre-Pascal is exceptional good at parallel teleprocessing, which is exactly why the graphics cards perform so well and why Nvidia have been dominating recently. Sure, AMD have more stream units but more doesn't make them better if they do less work, are less efficient or cannot be utilised entirely. There is a lot more to making a good graphics cards than simply shoving in thousands of cores, there is a very delicate balancing act of memory controllers, L2 cache, geometry and tessellation, texture units, ROPS, and a host of other things you need to get right. Nvidia could have simply shoved a lot more CUDA cores in there chips and blown their transistor budget on silicon that wont process games any faster.



You are completely ignorant and hopelessly wrong if you think NVidia can't compete successfully at DX12. The only benchmark we have for DX12 shows a convincing win for for Maxwell over GCN. Further proof comes form the available Mantle games that simply don't show the performance increase you are dreaming about.


The main concepts of DX12 have been pushed by both Nvidia and AMD since 2005 but likely for political and technical reasons nothing was done about it until now. Nvidia had a huge input into the DX12 design, just like AMD did, MS, Intel and a load of other parties. To day that DX12 is Mantle is extremely foolish.

 

[setter]

Id hazard a good guess that by the time we see DX12 games, the tx/ti and fury will be distant memorys. Rumour that star wars battlefront will be DX12, a game built on quite possibly the buggiest engine ever designed. Dice cant even get bf3/4 to run properly on frostbite never mind a new game. I play a lot of bf4, well i used to. Play cte mainly theese days as the base game is a joke.

 

[shankly1985]

Ahh, so you went aio cooling on them mate, heck i had to sell my old 780's quick because family would have stole them lol.

Ive got an msi 980ti coming in the next few days, will be weird having one card again tbh. But it will be benched, and gamed on pretty hard. Tbh its a good time for us all as gamers, new cards, new games, (albeit buggy lol). I think the fury x will improve overa short period of time, it's a new card with new very unknown memory, but perhaps a good indication of what the future will bring. Good too see a bit of competition in the gpu market.

No still running one water just the other is back air cool. That going to me little brother amd water cool will be second pc. Furyx is for me

Yeah competition is great for us all. We should all each very happy amd and nvidia is neck and neck.

 

[TheMorningStar]

Zen is definitely H1 '16. Zen APU is (probably) '17.

Oops my bad, getting confused with all the McShizzle to read at the moment. That reminds me though isn't there also something in DX12 that allows using APUs alongside dedicated GFX? Not crossfire just some kind of multi GPU/APU balancing. A little reminiscent of Lucid (was it? been a while)

Yeah if you can wait then go for it. If only I could do the same

Zen 'APU' is 2017. Zen 'CPU' is 2016, I think on AM4 with DDR4. The focus is on desktop CPU first then roll it out to APU the year after.

If you went X99 you wouldn't regret it, some good price on the mobo's. Get a good deal on a 5820K and some DDR4 and you are set.

Awesome, 2016 was a stretch, I couldn't see myself waiting till 2017. This old girl randomly lost a core a while back, came back after a few days and is oddly running better than it did before. AMD is currently a bit poo on the CPU front (will re-evaluate existing tech after DX12 and Vulcan hit) and Intel 1150 just doesn't have the damn PCIe lanes, 1151 looks the same as well, 2 GFX cards and an x4 M.2 drive gives me nothing to play with over the next 5 years, which is about what I want out of the base system. Also not really that many quality mobos for AMD offerings (never touching ASUS again) MSI's my new bezzy mate.

 

[triss]

We going by information we have ready for us now.. mantle is still going to be around its out of Beta and now running with Vulkan api.. Again something you should already know.

Mantle set out to try and be an open standard that was AMDs idea and they have succeed in doing so.

IIRC AMD themselves said mantles not out of beta and is back to being something they can use again if needed to do things a dev wants that the current API's cant
So its not dead and its not out as such its just lurking if required

Mantle must take on new capabilities and evolve beyond mastery of the draw call. It will continue to serve AMD as a graphics innovation platform available to select partners with custom needs

.
https://community.amd.com/community/gaming/blog/2015/05/12/on-apis-and-the-future-of-mantle

 

[Orangey]

Zen is definitely H1 '16. Zen APU is (probably) '17.

Zen is Oct 2016, if that.

 

[TheMorningStar]

IIRC AMD themselves said mantles not out of beta and is back to being something they can use again if needed to do things a dev wants that the current API's cant
So its not dead and its not out as such its just lurking if required


.
https://community.amd.com/community/gaming/blog/2015/05/12/on-apis-and-the-future-of-mantle

Mantle has better CPU scaling, DX12 only seems to use the 6 cores, I'd like to see AMD bring out a chip with 6 or 8 big cores rather than 16 multithreaded ones if that's going to be the case, it may be scaled up when it's out of beta mind. Vulcan apparently has a close relation to Mantle so things may be different there, I never hear much about Vulcan though everyone has DX12 excitement at the moment, sourse2 uses it apparently so more important to my gaming than DX12 really.

 

[setter]

No still running one water just the other is back air cool. That going to me little brother amd water cool will be second pc. Furyx is for me

Yeah competition is great for us all. We should all each very happy amd and nvidia is neck and neck.

Best attitude mate, I don't get the whole fanboi thing. Buy what suits your budget, has the features you need. Tbh on games there's so little no matter which competing card you own.

 

[TheMorningStar]

Has anyone looked into the thermal margin for HBM yet? Just been having a look for operating temps to see if this could restrict an overclock, gpu vcore etc. other than some info about thermal dummy bumps that help to dissipate heat in the stack I can't find anything on the actual thermal limit.