Yeah, don't trust pclab benchmark. They biased.
AMD Radeon™ Discussion V12, Latest - 14.12 | WHQL - 14.12
AMD Radeon™ Discussion V12, Latest - 14.12 | WHQL - 14.12
|
|
Apr 15 2015, 04:39 PM
Return to original view | Post
#21
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
Yeah, don't trust pclab benchmark. They biased.
|
|
|
|
|
|
Apr 29 2015, 08:09 PM
Return to original view | Post
#22
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(S4PH @ Apr 29 2015, 06:40 PM) QUOTE(Minecrafter @ Apr 29 2015, 06:25 PM) I agree.Long gone the days of low vRAM. But they still make 1GB-2GB mid-range cards.. |
|
|
May 3 2015, 12:46 PM
Return to original view | Post
#23
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
ASIC is meaningless for hawaii.
|
|
|
May 11 2015, 10:57 AM
Return to original view | Post
#24
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
Lunch break reading.
http://forums.anandtech.com/showthread.php?t=2430693 This post has been edited by kizwan: May 11 2015, 11:07 AM |
|
|
May 15 2015, 08:57 PM
Return to original view | Post
#25
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(Human10 @ May 15 2015, 06:00 PM) But what I find more interesting is sometimes a stable OC for unigine is still unstable for games, despite stress tested with unigine. That's because different games, different stress level (to gpu), even between the games with the same engines. Unigine definitely not stress enough for absolute stability test. If I'm not mistaken Final Fantasy IV benchmark is more demanding than unigine. |
|
|
May 23 2015, 12:14 AM
Return to original view | Post
#26
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
Nah, that doesn't look short since the pic only show half of the card.
|
|
|
|
|
|
Jun 2 2015, 02:31 AM
Return to original view | Post
#27
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
|
|
|
Jun 2 2015, 03:12 PM
Return to original view | Post
#28
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(Archaven @ Jun 2 2015, 03:02 PM) hmm so for best SLI / xfire setup it's best to get 1000W instead? CPU will take up alot of power? i5 3570k but i foresee to upgrade in future. thanks If you're going to overclock, yes 1000W & above. If not, 850W is enough for 3570k with moderate overclock & a pair 290X's at stock clocks.EDIT: If you have the budget to upgrade PSU too, I recommend no less than 1200W. This post has been edited by kizwan: Jun 2 2015, 03:21 PM |
|
|
Jun 2 2015, 03:41 PM
Return to original view | Post
#29
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(Archaven @ Jun 2 2015, 03:29 PM) 1000W should be safe, otherwise i think 1200W is just overkilled and waste of money. correct me if i'm wrong. thanks Sure, 1000W should be enough too. And no, higher wattage PSU is not overkill because PSU only supply the amount of power the devices in your computer require. If 1200W PSU going to supply 1200W all the time even though the devices in your computer doesn't require that high wattage, sure it's overkill but this is not the case. |
|
|
Jun 2 2015, 08:34 PM
Return to original view | Post
#30
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
Not really an issue. You can disable Crossfire when playing certain games & enable Crossfire when running games that supported it.
|
|
|
Jun 4 2015, 02:07 AM
Return to original view | Post
#31
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(JohnLai @ Jun 4 2015, 12:39 AM) ...........so.....the rebranded hawaii 390x.....water cooling + 8pins and 6pins PCIE power connector? For GTA V, did you try 15.4 beta or the latest beta? GTA V work with crossfire here with no weird black checker board pattern.In addition to extra VRAM and higher GDDR5 clock speed (6Gbps instead of 290x original 5Gbps).......I think AMD also 'overclock' the gpu core to the maximum possible amount..... Power consumption is going high for sure, sigh.....lack of delta color compression(not like hawaii gonna need this because its memory width is quite large) -.- Feature wise, I still think maxwell GTX 980 or 970 (even with 3.5gb fiasco) is a better purchase. Example, hybrid HEVC 8bit DXVA decode, color compression, lower driver overhead (hope amd can improve its driver overhead), lower power consumption compared to upcoming extra overclocked 390x. No idea what the hell amd is doing, two of my R7 260X sitting there doing nothing because no 'working' crossfire profile for GTA V, Project Cars and Witchers 3. GTA V = weird blocky black checker board pattern on bottom right of screen. Project cars = need to use modded driver dll from windows 10 preview for better performance. Witchers 3 = followed all AMD knowledge base solution to reduce flicker, still flickering hard especially at watery area. |
|
|
Jun 6 2015, 03:53 PM
Return to original view | Post
#32
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(JohnLai @ Jun 6 2015, 11:05 AM) Mantle 'push' Microsoft? No such thing. Pretty much you have never tried Mantle before & just take whatever you read in the internet as 100% accurate. The whole point of Mantle is to reduce CPU overhead & it does work.DX12 is already in development long before AMD announced its Mantle initiative. Beside, the so called Mantle low level is kinda useless because Catalyst driver cpu overhead. DX development doesn't take one year to finish, it takes several years for testing suitability, acceptability and feasibility. I think Mantle kinda rushed to the market. If you take a look on the mantle documentation, it is kinda......messy and feel rushed. Really wonder how DICE is able to use the mantle renderer. Mantle is destined for failure in first place. Reason, AMD has somehow low discrete gpu market share. For console argument; PS4 dev tool uses PlayStation Shader Language (PSSL). Xbox One on the other hand uses DirectX HLSL with dx 11.2 API. We don't expect Sony to support mantle or microsoft to use its competitors API, no? Reason for people saying some documentation from mantle is similar to dx12 documentation? Simple, AMD contributed some of the code to Microsoft to review and included in DX12. It is not copying. |
|
|
Jun 7 2015, 12:47 AM
Return to original view | Post
#33
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(kizwan @ Jun 6 2015, 03:53 PM) QUOTE(JohnLai @ Jun 6 2015, 11:05 AM) Pretty much you have never tried Mantle before & just take whatever you read in the internet as 100% accurate. The whole point of Mantle is to reduce CPU overhead & it does work.QUOTE(JohnLai @ Jun 6 2015, 07:15 PM) ........I think you confused about something. Are you sure your're not the one confusing here? Well I'm not talking about DX9 or DX11 here but Mantle. (Shortened previous reply above to show that is the only thing I'm commenting.) You said it's useless which is not true. Did you ever try it? I don't think so. AMD has been updating Mantle at least few times since release, are sure it wasn't new update of Mantle in the driver that fixed the problem with 285? I have been using Mantle since release in BF4 & never look back. Nowadays I'm no longer playing games @1080p. Now 1440p or 4K resolution for me. Now I'm using DX11 again even in BF4. At the moment I'm only playing two games, BF4 & GTAV with DX11 & 1440p resolution. Well I don't have any complain with performance whatsoever.Allow me to elaborate; First thing first, you need to differentiate between driver and API. Easier if I grab some definition from wikipedia. 1) In computing, a device driver (commonly referred to as a driver) is a computer program that operates or controls a particular type of device that is attached to a computer. A driver provides a software interface to hardware devices, enabling operating systems and other computer programs to access hardware functions without needing to know precise details of the hardware being used. I was talking about AMD catalyst driver overhead. 2) In computer programming, an application programming interface (API) is a set of routines, protocols, and tools for building software applications. An API expresses a software component in terms of its operations, inputs, outputs, and underlying types. An API defines functionalities that are independent of their respective implementations, which allows definitions and implementations to vary without compromising each other. A good API makes it easier to develop a program by providing all the building blocks. A programmer then puts the blocks together. Mantle and DirectX are API. Now, DirectX 9 - 11 are known to have some API overhead. The purpose of Mantle and upcoming DirectX 12 is to reduce/optimize the API overhead by making low level access available. And yes, as you said, Mantle successfully reduce the API cpu overhead. Next, the second part is the DRIVER cpu overhead. This is the section AMD has its worst. Compared to nvidia driver, the issue of AMD catalyst driver overhead in dx9 - 11 game is quite serious. You can find more about it at guru3d forum. http://forums.guru3d.com/showthread.php?t=398858 When AMD released its R9 285 (unofficially known GCN 1.2), even when using Mantle renderer, there was some performance issue. This was solved after AMD optimized R9 285 driver overhead later. Edit: some grammer and wording mistake. You're talking about driver overhead but keep linking the issue with DX9 - DX11. What about driver overhead with Mantle then? Have you seen AMD FX CPU with quad 290X rig? Note: he doesn't play @1080p of course. EDIT: Well when I said CPU overhead, obviously I'm referring to API overhead, duh! It pretty much affected the CPU utilization. This post has been edited by kizwan: Jun 7 2015, 01:25 AM |
|
|
|
|
|
Jun 7 2015, 03:09 AM
Return to original view | Post
#34
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
Using 3dmark API Overhead test as a reference, 290X will be able to use DX12 multi-threading. Hopefully the potential shown in the benchmark translating to actual games performance when DX12-enabled games available.
http://www.pcworld.com/article/2900814/tes...-is-insane.html NOTE: The API Overhead test is not a tool for comparing GPUs. So hopefully no one comparing GPU A & N based on the draw calls. There's more to it than that. |
|
|
Jun 7 2015, 04:31 AM
Return to original view | Post
#35
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(JohnLai @ Jun 7 2015, 12:58 AM) Well, why don't you then? lol We all know the issue with AMD driver is high DX11 CPU overhead & usage. This is what driver overhead is referring to actually. When you said Mantle or DX12 is useless if AMD doesn't fix the driver overhead issue, it doesn't make sense to me. To solve driver overhead issue, AMD introduced Mantle & thanks to Mantle, now Microsoft introducing DX12. Using BF4 as an example, since it's DX11-enabled & Mantle-enabled game, performance when using Mantle much better than when using DX11 because the API CPU overhead & CPU usage are reduced, allowing more draw calls per second. However, while Mantle work great with GCN 1.1 card (Hawaii), it doesn't look so great with older cards. I'm afraid this trend may carry forward when Win 10 & DX12-enabled games available. We'll see. Right now I think to really utilize DX12, you need at least Hawaii card(s) if you go with AMD cards. If DX12 is anything like Mantle, we will see better performance once Win 10 & DX12-enabled games available. However, AMD will still struggling when running Nvidia-optimized games though. So we will still see the same crap that we're seeing currently. With the closed codes, e.g. Gameworks, it's hard for AMD to release optimized drivers for that games. |
|
|
Jun 7 2015, 12:06 PM
Return to original view | Post
#36
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(area61 @ Jun 7 2015, 09:49 AM) Hard to see anyone running AMD processor and pairing it up to enthusiast Radeons. Basically The issue of CPU overhead is moot when running on Intel processor. Not true though. You can still have CPU overhead issue with multi-GPU & with Intel processor when running CPU intensive game & at 1080p resolution.Saying AMD processor can't run enthusiast level Radeon in multi-GPU config is not based on real fact. Take a look link below. I'm pretty sure someone will point out this & that benchmarks scores is lower than Intel this & that benchmarks scores. Synthetic benchmarks always favour one camp over another. Rather than look at benchmark score, look at real games performance instead. Keep in mind this is not about Intel vs. AMD but it's done to show even AMD CPU also can handle multi-GPU & multi-monitor setup. http://www.overclock.net/t/1473361/amd-hig...0#post_23607538 This post has been edited by kizwan: Jun 7 2015, 12:07 PM |
|
|
Jun 7 2015, 03:43 PM
Return to original view | Post
#37
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(empire23 @ Jun 7 2015, 02:30 PM) Well if you can afford it, why not. Well in my opinion, if you can afford a pair of high-end cards, you should be able to afford 1440p monitor at least. A pair of 290/290X are too powerful for 1080p anyway. I just want to point out that people should be able to differentiate between driver overhead issue and driver optimization issue. Even if the API properly coded & the driver itself coded properly to utilized the API, you may still not get good performance in the games if the driver itself is not optimized for that games. This is the problem with AMD drivers especially with Nvidia-optimized games. Well I'm not arguing but just discussing, if you're referring that to me but probably not. I always keep in mind that forum can't conveys tone. This post has been edited by kizwan: Jun 7 2015, 03:44 PM |
|
|
Jun 8 2015, 02:27 PM
Return to original view | Post
#38
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(empire23 @ Jun 8 2015, 05:20 AM) That's why I believe the discussion about driver/api overhead is slightly moot without an understanding of the underlying principles of computer engineering. You have a point there but the discussion wasn't THAT deep though. The discussion was around the misunderstanding between driver overhead issue & driver optimization issue. It's not why there's driver overhead issue.Regarding AMD driver overhead issue, just in the span from around last year to now, I have seen a lot of improvement in DX11 CPU overhead. The major issue nowadays is the lack of optimization in the drivers for certain popular games, especially Nvidia-optimized games. What we can do right now is just wait for AMD to release the magic drivers, or jump to green camp. If I have computer engineering degree, I would be at AMD forum, helping them what to do. And the engineers that read it will said among themselves: "Look, this newbie trying tell us how to do our job". |
|
|
Jun 8 2015, 08:15 PM
Return to original view | Post
#39
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(Narako @ Jun 8 2015, 06:26 PM) Hey guys, everytime AMD catalyst says to update Likely the drivers is updated but the reg key that hold the version number in the registry not updated. I don't use "Software Update" feature in CCC. Better when updating drivers, download directly from AMD website & you should use the latest driver, including the beta drivers. Uninstall the old driver properly & install the new driver using the *.exe file.it says current version 14.10.1006 and the update is version 14.201.1008. But after update the version stayed the same like nothing changed why is that happening? |
|
|
Jun 8 2015, 10:43 PM
Return to original view | Post
#40
|
![]() ![]() ![]() ![]() ![]()
Senior Member
826 posts Joined: Jan 2009 |
QUOTE(Narako @ Jun 8 2015, 08:32 PM) i realized the problem. AMD catalyst didnt provide updates for windows 8 64bit. Sorry, no idea. I do not use Windows 8 or 8.1. However, if AMD doesn't provide drivers for Windows 8, where the "Software Update" get the driver update? Probably it also use driver for Windows 8.1 in Windows 8?! they only provide updates for windows 8.1. which is odd. If i install the driver from windows 8.1 will it cause issues? |
|
Topic ClosedOptions
|
| Change to: | 0.0558sec
1.05
7 queries
GZIP Disabled
Time is now: 30th November 2025 - 07:09 AM |