dunkinkid22

DNOTTIS

hrmm I just disable them in device manager.  Then again I'm using a PCI-e X-FI Titanium.  I don't have any sound related issues though.

Are you sure that you have no irq conflicts?

With what?

onishchuk4

DNOTTIS

dunkinkid22

DNOTTIS

hrmm I just disable them in device manager.  Then again I'm using a PCI-e X-FI Titanium.  I don't have any sound related issues though.

Are you sure that you have no irq conflicts?

With what?

He probably referring IRQ conflicts with Nvidia sound drivers and your stock ones.

My on board is disabled and I have the X-Fi.  I disable the Nvidia audio drivers in the CP.  I've always done this though, even my HD5870 had HDMI audio driver I had to disable.

AHowes

One program will stop one of the cores on my i7 930? I looked on the eleet program and seen nothing to disable a core. I'd hate to reboot the system each time and disable a core in the bios everytime i wana play BFBC2.

Start game. Alt tab out, goto task manager, processes, right click BFBC2.exe, set affinity. Disable (uncheck) one core. 


If you disable the core from bios and boot up the performance was still choppy.

It's not choppy on my processor SLI is junk on 295s and 470s. On my 470s graphic driver stops working in DX10 and DX11. I can't tell a difference between one card and two with fraps and it seems better with one in DX9 and 470.

dunkinkid22

DNOTTIS

dunkinkid22

DNOTTIS

hrmm I just disable them in device manager.  Then again I'm using a PCI-e X-FI Titanium.  I don't have any sound related issues though.

Are you sure that you have no irq conflicts?

With what?

what i found is that the nvidia drivers were sharing some of my onboard drivers. you need to look in device manage under driver details and see which files and services they are sharing see my guide for a way to remove them
http://forums.evga.com/tm.aspx?m=531113

Actually, I have found with my X-Fi disabling the HD drivers is enough. The trick is to have the X-Fi driver installed and set up before installing the 285.96 driver with HD audio drivers. Yah, wish they didn't self install.

new drivers this week, not the BC2 fixed drivers, but should be something from the 260 branch.

i think part of the problem is that these drivers are designed to work with a wide range of cards (6,7,8,9,200,400 series) I think nvidia should just focus on each product specifically.  Cause each has its own architecture and own issues. drivers should be just for the 400 hundred series.

why do i have 90%'s gpu usage and other people don't yet i am using the same driver as everyone else?

DNOTTIS

Here we go guys - 



It seems this weeks beta driver has been pushed back until Monday so there is a chance I can get the fix in for Monday's driver release. I will let you all know.


--------------------
NVIDIA Forums Technical Advisor
-------------------------------------------------------------------------------------------------------
NVIDIA Software Downloads - http://www.nvidia.com/Download/index.aspx?lang=en-us
NVIDIA Knowledge Base - http://nvidia.custhelp.com/cgi-bin/nvidia....ser/std_alp.php
NVIDIA nZone Entusiast Website - http://www.nzone.com
NVIDIA SLI Zone - http://www.slizone.com
Twitter - http://twitter.com/nvidiacc

That would be awesome.


Thanks again DNOTTIS.

dunkinkid22

i think part of the problem is that these drivers are designed to work with a wide range of cards (6,7,8,9,200,400 series) I think nvidia should just focus on each product specifically.  Cause each has its own architecture and own issues. drivers should be just for the 400 hundred series.

no thanks.

mdswish

What are the pros and cons to having separate drivers for each family?  I can't think of any negatives other than convenience.  On the pro side, each driver package would be smaller in size and may be easier to code since there are less worries of making a change to a 400 series card feature that might break something in the 8000 series.  Maybe a person more educated on software or drivers than I am can educate us?

I'm a software developer and I can tell you that it is easier to create a 'one size fits all' solution.

Most of the cards are very similar they are based around the same idea (Programmable Shaders) and from the 8 series to the current 400 series the architecture has remained pretty much the same. The physical build of each GPU is different but the way in which the instructions are processed has remained the same. It is because of this that most graphics glitches can be fixed once and have an effect across all the cards.

If they went with a 'many driver per device' approach it would create a lot more work and a lot of work would get repeated. It would also require much more testing for each individual driver (slower releases) and it would become attractive for NVIDIA to drop support for older cards much faster.

The other issue is obviously people want to use multiple series cards in one system such as a 9 series card for PhysX or simply to have more displays attached to one computer.

WinMacLin

mdswish

What are the pros and cons to having separate drivers for each family?  I can't think of any negatives other than convenience.  On the pro side, each driver package would be smaller in size and may be easier to code since there are less worries of making a change to a 400 series card feature that might break something in the 8000 series.  Maybe a person more educated on software or drivers than I am can educate us?

I'm a software developer and I can tell you that it is easier to create a 'one size fits all' solution.

Most of the cards are very similar they are based around the same idea (Programmable Shaders) and from the 8 series to the current 400 series the architecture has remained pretty much the same. The physical build of each GPU is different but the way in which the instructions are processed has remained the same. It is because of this that most graphics glitches can be fixed once and have an effect across all the cards.

If they went with a 'many driver per device' approach it would create a lot more work and a lot of work would get repeated. It would also require much more testing for each individual driver (slower releases) and it would become attractive for NVIDIA to drop support for older cards much faster.

The other issue is obviously people want to use multiple series cards in one system such as a 9 series card for PhysX or simply to have more displays attached to one computer.

took the words right out of my mouth.

dunkinkid22

i think part of the problem is that these drivers are designed to work with a wide range of cards (6,7,8,9,200,400 series) I think nvidia should just focus on each product specifically.  Cause each has its own architecture and own issues. drivers should be just for the 400 hundred series.

+1 Amen to that.  I wish Nvidia would take note and focus on each independently as you stated the architecture between the cards is different enough that it warrants it's own development.

squall-leonhart

WinMacLin

mdswish

What are the pros and cons to having separate drivers for each family?  I can't think of any negatives other than convenience.  On the pro side, each driver package would be smaller in size and may be easier to code since there are less worries of making a change to a 400 series card feature that might break something in the 8000 series.  Maybe a person more educated on software or drivers than I am can educate us?

I'm a software developer and I can tell you that it is easier to create a 'one size fits all' solution.

Most of the cards are very similar they are based around the same idea (Programmable Shaders) and from the 8 series to the current 400 series the architecture has remained pretty much the same. The physical build of each GPU is different but the way in which the instructions are processed has remained the same. It is because of this that most graphics glitches can be fixed once and have an effect across all the cards.

If they went with a 'many driver per device' approach it would create a lot more work and a lot of work would get repeated. It would also require much more testing for each individual driver (slower releases) and it would become attractive for NVIDIA to drop support for older cards much faster.

The other issue is obviously people want to use multiple series cards in one system such as a 9 series card for PhysX or simply to have more displays attached to one computer.

took the words right out of my mouth.

I am also a software developer and from a business point of view a one size fits all model makes sense.  But from a customer/user point of view it does not.  There are large enough differences in the card architectures that it indeed does affect how they code and release their drivers.  From a user base I know from experience that if I received a driver that was specifically written for my hardware that it would function much more efficiently and would most likely be of much higher quality. I could argue both ways but for now because I own the 400 series cards and I am experiencing issues with them I am leaning towards the side of the user and wishing they did in fact separate the series out and develop drivers independently of each other. From a business stand point I don't see this ever happening though.  It used to be this way not to terribly long ago though.

CREUSS
I am also a software developer and from a business point of view a one size fits all model makes sense.  But from a customer/user point of view it does not.  There are large enough differences in the card architectures that it indeed does affect how they code and release their drivers.  From a user base I know from experience that if I received a driver that was specifically written for my hardware that it would function much more efficiently and would most likely be of much higher quality. I could argue both ways but for now because I own the 400 series cards and I am experiencing issues with them I am leaning towards the side of the user and wishing they did in fact separate the series out and develop drivers independently of each other. From a business stand point I don't see this ever happening though.  It used to be this way not to terribly long ago though.

If they were to do one set of drivers per card series you would still find that the code among them all would be duplicated. When they create a new card they are not going to go back to the drawing board and completely redesign their entire driver architecture and how the cards interperate 3D API's. And that is because of the following reasons:

1. They already have an efficient design you can only reinvent a system a few times before you run out of ideas for making it more efficient. You have to wait for the rest of the industry to take further steps forward before you can redesign something that is not just made differently but better than your old design.

2. All your games would break. Well maybe not all of them, but quite a lot would. It is not uncommon for game developers to place code in the game to target a specific card makers drivers and architectures. If NVIDIA restarted from scratch it would drive game developers insane and it would be like having a 3rd GPU maker releasing cards (from the game developers perspective)

3. If they had to start over with each series Fermi would have been delayed until 2012, or NVIDIA would be bankrupt.

4. NVIDIA develop the Drivers and Cards together at the same time. The cards are designed to stay within a unified specification that NVIDIA has outlined so that the cards can work reliably with the current driver architecture and the Drivers are designed to work with a very standard and wide range of cards efficiently. They aren't simply taking an old driver and messing with it to make new cards work, it is designed from the start to be easily portable to new cards.

5. Core Logic vs Game Fixes. It is much easier for NVIDIA to apply a game specific bug fix to the driver than it is to add that fix in to the core logic. Especially if the fault is not with NVIDIA but with the game designer. e.g. they cant/shouldn't alter their cards hardware to support a game designers poor coding, this means they need to apply the fix in software and it simply makes sense to use 1 software fix across the entire range of cards instead of tailor making a fix for five series of cards or more.

This is the same reason that we don't have AMD and Intel builds of applications. The processors are completely different in physicality but they support a standard (x86) which abstracts all the nonsense in each design to give us a basic outline to target. With every piece of silicon available we can choose to go with a unified approach (Single Binaries supporting a specification) or we can go with a tailored approach (Multiple binaries supporting specific silicon). But when it comes down to whats best for the developer, original equipment manufacturer and consumer unfinified ALWAYS wins. Everyone gets a better product.

And this is true with everything. A single unified approach always wins. Can you imagine what would happen if Direct X didn't exist and every card maker still supplied their own Graphics API? I shudder at the thought!

squall-leonhart

Actually, we do have intel and amd specific builds of applications. though these are usually open source. most of the time, processor specific optimisations are hand written in with detection strings.

ICC is used quite often to produce processor specific PGO builds.

Well yes you can optimize for specific processors but almost no one does. Most programmers leave it up to the compiler to interpolate there code and the ones that do go out of there way to optimize for specific processors do so in a single binary it's very rare to see AMD and Intel specific builds distributed as separate entities. And it is even rarer to see developers target specific CPU series under a manufacturer unless it's for raising the minimum requirements (like the move from 32-bit to 64-bit only software has allowed developers to begin targeting the Core 2 range of Intel as the bottom line and thus make use of instructions that earlier processor series lacked)

And really this is what NVIDIA does they build in some specific code for different GPU's but most of the code is going to work across there full range. Just like in x86 development.