are you blind?
break data into parts and assign device ID to its protocol information. 16 parts. one for each pci-lane. each piece of data contain protocoled information which will reach a decoder to decode protocol info. protocoled information contain device ID which will only be processed by GPU with correct device ID.
you said "(SLI Driver Path) requires no software changes according to your logic ..."
I'm referring to your "SLI Driver path"..Never even heard of it. . did you make that up? How do I imply on something I don't even know?
really. so what does SLI do?
Completely unrelated to this discussion..
each GPU does its own AA on its data..
please search for answers before ask.... tired of you popping up new questions that are completely unrelated to the discussion...
[/quote] each GPU does AA on its own data. that is it. same as in a dualcore cpu each core only process its own part...
And I'm making myself look like the fool? Its your words in my sig anyway, if it is the truth, what do you care? Unless you now know that you are wrong.....
This is TOO FUNNY!!! You've once again gotten me confused with someone else. I never said that it "sees" two GPU's, that was BFG10K. Also, I didn't say that it "saw" data. I said that the data HAS TWO destinations. In order for your precious data to be sent, it has to go somewhere. It doesn't just get shot out for nothing
You've even began saying what I'm saying in your later response to BFG10K:
Gee, why would it need to do that? Couldn't be because it has to go to two different places.....
Yep. That's right. And what are those profiles? (whispers: Shhhhhh, they're S. L. I. profiles)
From a hardware standpoint, I agree with that. However, from a software standpoint, it is two cards.
You are a complete idiot. You argue with common sense, you get those who are arguing with you confused and constantly misquote them all while butchering the language you speak. STOP POSTING.
---
If you want post on our forums, please spend some time improving your forum courtesy. If you want to be a prick to others, find another forum.
If you cannot observe the first option, we can and will enforce the second one.
AnandTech Moderator
Josh, you are just thread crapping and initiating personal attacks at this point. If you keep your sig, I'm going to PM mods.
Simple as that.
you replied "for" BFG from my reply to refute his idea of "software sees 2 GPUs", quoted:
break data into parts and assign device ID to its protocol information. 16 parts. one for each pci-lane. each piece of data contain protocoled information which will reach a decoder to decode protocol info. protocoled information contain device ID which will only be processed by GPU with correct device ID.
SLI software "sees" 16 PCI-e lanes as its destinations. GPU device ID is contained within protocol information part of the actual data.
Simple as that.
you replied "for" BFG from my reply to refute his idea of "software sees 2 GPUs", quoted:
break data into parts and assign device ID to its protocol information. 16 parts. one for each pci-lane. each piece of data contain protocoled information which will reach a decoder to decode protocol info. protocoled information contain device ID which will only be processed by GPU with correct device ID.
SLI software "sees" 16 PCI-e lanes as its destinations. GPU device ID is contained within protocol information part of the actual data.
so? whats your point?
Do you ever wonder why people are not agreeing with you nor backing you up when you get into arguments like these?
dramaqueen, I have to live by your words of wisdom. Don't force me to take your words out of my sig I could not live with out their divine wisdom. Why do you care so much that I have your own words QUOTED in my sig? Its not like plagarism or anything. That should be a compliment, if you think you are right that is.
You can talk about how data gets distributed all you want, but that doesn't change the fact that from a software/driver standpoint the 7950 is two cards. Why do you need SLI profiles if it is all one thing dramaqueen?
dramaqueen, I have to live by your words of wisdom. Don't force me to take your words out of my sig
I could not live with out their divine wisdom. Why do you care so much that I have your own words QUOTED in my sig? Its not like plagarism or anything. That should be a compliment, if you think you are right that is.You can talk about how data gets distributed all you want, but that doesn't change the fact that from a software/driver standpoint the 7950 is two cards. Why do you need SLI profiles if it is all one thing dramaqueen?
Oh I can't wait for dramaqueen's next broken english reply. :laugh:
---
The same warning applies to you as well as your brother. Be nice, or be somewhere else.
AnandTech Moderator
---
The same warning applies to you as well as your brother. Be nice, or be somewhere else.
AnandTech Moderator
BFG10K
Lifer
- Aug 14, 2000
- 22,709
- 3,002
- 126
Oh, pu-lease. You're the one slinging insults at people and looking down at them just because you've taken a C++ class or something.
Furthermore you appear to lack serious reading/comprehension skills and instead of addressing the core issues you play rhetorical and semantic games with a grasp of the English language akin to that of a child.
No, are you? It looks like a direct quote to me.
Yeah, and? Or in your world is a NIC the same thing as a GPU because they're both subjected to the mechanics of the PCIe protocol?
If the "hardware has nothing to do with the software" in your world the NIC must use the same drivers as the GPU, right? It's all PCIe in your world, right?
No, nVidia "made it up" when they implemented SLI support in the driver, something that was not necessary (or in fact undefined) for single cards.
Load balancing to two (or more) GPUs. It doesn't operate or even exist on single cards hence a 7950GX2 can't possibly be classed as single card.
The fact that it sits on PCIe and it only has one socket is irrelevant to the fact that it's TWO GPUs on one PCB and in terms of software behaves exactly like two separate GPUs sitting on separate PCBs.
The hell it's unrelated. You claim the 7950 is a single card yet by definition SLI AA requires composition from multiple GPUs. That is to say the concept of SLI AA doesn't even exist on single cards.
Right, the two GPUs that are on the card which means it's not a single card. This is unlike a true single card that doesn't even have the concept of SLI AA or even the concept of SLI.
:roll:
Okay beggerking, so the mods can depict your English afterall, but turned the other cheek when you posted this:
THREAD LINK
Doesn't seem too different from the "offensive" nickname I made. I didn't know that they listen to people who were already banned once? And what ever happend to this:
Found here:THREAD LINK
Yeah, you sure ignored it.:roll:
You still haven't answered my questions, and continued to name-call just as much as I have. So, I'll once again ask you very clearly:
If the 7950 is one card both hardware-wise and software-wise, why does it need SLI to be enabled to work?
THREAD LINK
Doesn't seem too different from the "offensive" nickname I made. I didn't know that they listen to people who were already banned once? And what ever happend to this:
Found here:THREAD LINK
Yeah, you sure ignored it.:roll:
You still haven't answered my questions, and continued to name-call just as much as I have. So, I'll once again ask you very clearly:
If the 7950 is one card both hardware-wise and software-wise, why does it need SLI to be enabled to work?
BFG, this is my last reply to you. Obviously you have no idea what you are talking about... You used the term "SLI Driver path" many times to explain your reasoning, yet when I ask you to define exactly what a "SLI Driver path" is, you reply with "No, nVidia "made it up" when they implemented SLI support in the driver, something that was not necessary (or in fact undefined) for single cards. " .
as a favor, I'm answering your question(s) for one last time.
Incorrect. NIC driver uses a different protocol in its data.
once again , incorrect. SLi was never a composition from multiple GPU. Its the other way around..it decomposes data into parts..
AA = antialiasing= filling in missing parts between 2 lines with calculated points. each pass add points inbetween spaces between points. more point = smoother lines = less jagged edges.
in oridinary AA, 4x = apply 4 AA passes on a whole screen with a single GPU. SLI AA = 2 GPUs each apply 8x on 1/2 screen which produce 8x AA.
concept of SLI = concept of windowsXP dualcore CPU support.
so , dual-core cpu is not a single GPU...ok...
Goodbye and have a good life.
to Josh:
as a favor, I'm answering your question(s) for one last time.
Incorrect. NIC driver uses a different protocol in its data.
once again , incorrect. SLi was never a composition from multiple GPU. Its the other way around..it decomposes data into parts..
AA = antialiasing= filling in missing parts between 2 lines with calculated points. each pass add points inbetween spaces between points. more point = smoother lines = less jagged edges.
in oridinary AA, 4x = apply 4 AA passes on a whole screen with a single GPU. SLI AA = 2 GPUs each apply 8x on 1/2 screen which produce 8x AA.
concept of SLI = concept of windowsXP dualcore CPU support.
so , dual-core cpu is not a single GPU...ok...
Goodbye and have a good life.
to Josh:
exact same reason why dualcore CPU / multiCPU system needs WindowsXP's support to make use of the 2nd core/CPU. aka load balancing/data distribution. Google and howstuffworks.com is your friend.
How is the data that was processed on the 2 GPUs recombined into 1 image and out through to the monitor?? I'm guessing it's the SLI drivers that recombine the data from the 2 GPUs.
If there was hardware/software on the card itself that recombined the image, THEN there would be no need for SLI drivers and hence you could MAYBE call it a single card since the on-card SLI would be transparent to the OS and would act just like a single card would.
Right, so from a software standpoint, it registers to TWO devices. This makes it a dual card from a software perspective.
Beggerking I am sorry that I offended you in a past post and I hope that it doesn't affect our future descusions.
I think that we are all missing out points. When you say the exact same reason why dualcore cpu/mulitcpu system needs WindowsXP's support to make use of 2nd core/cpu. That is what we are talking about. A specific software needs to be used by this card for it to make use of the 2nd core/gpu. We are basicly saying that when you look and hold this card and put it in your computer it is one card yes. However, you do need a software that can utilize this dual gpu card. That is why we are saying that it is one card when you are talking hardware and when we are talking software it acts like two cards.
Now when I say it acts like two cards, I don't know that it acts that way totally because I don't have that card. When I used two 7800gt's they where buggy and I just used one after that. As I understand it the 7950GX2 uses sli in the drivers, if that submits it to the same bugs then I could say that in software it acts like two cards. Aparently users from xtremesytems have liked their cards so I don't know yet what the fall out will be of using sli drivers with this card.
Another fact is that most motherboards need a bios upgrade to use this card. If software has nothing to do with hardware then why would you need to alter the software when putting different hardware in your motherboard?
These two facts, the bios upgrade for mobo, and the sli drivers needed for this card, are what have us saying from a software standpoint it is a dual card.
This fact alone doesn't take away from the feats this card accomplishes itself we are just stating that from a software view this is a dual gpu if not dual card. Perhaps calling it a dual gpu would be better, is that a compormise we all can agree on?
Wreckage
Banned
- Jul 1, 2005
- 5,529
- 0
- 0
No, it makes it dual GPUs from a software perspective.
Asus made a dual 6800 with 2 GPUs on the same PCB, is that 2 cards as well?
http://www.guru3d.com/newsitem.php?id=2422
It's one card and some are in denial because they are unhappy that it's the fastest single card on the planet. Get over it.
If it needs an SLI driver to operate, from a software standpoint--yes. From a hardware standpoint, no.
I'm agreeing with BFG10K. I doubt he's "unhappy that its the fastest single card on the planet" and niether am I. Nvidia does produce the "fastest" cards and I was never debating that. Yet you thought I was?
What if this thread is a sinister plot by evil viral marketers to keep the 7950GX2 in our minds?
BFG10K
Lifer
- Aug 14, 2000
- 22,709
- 3,002
- 126
You're the one that lacks a basic understanding of key issues but you accuse others of being stupid.
You don't know what a driver path is? It's a path that is taken in order to operate a given hardware device.
For most devices it's generic but since GPUs have vastly different features their driver pathways tend to differ.
Obviously nobody would expect the driver to program a DX7 card like the GF4MX in the same way as a DX9 card like the 7900GTX so the driver has a separate path. SLI is another driver path because it requires totally different programming to single cards.
What protocol? Certainly not PCIe since as far as it's is concerned the only real difference is less lanes.
That's your problem, you hide behind the PCIe protocol when it suits your argument but back-pedal when it doesn't.
What the hell are you talking about? That's exactly what it is. SLI composes one frame from multiple parts, each sitting on a different GPU.
SLI AA requires more than one GPU to function and that's why the 7950 can't be classed as a single card.
That's only one part of it.
Except you missed the most important bit, the offset from each GPU in SLI AA that is then merged into one frame. That's why it can't exist on single cards.
This is totally wrong and proves you don't have a clue what you're talking about.
SLI 8x AA is not 8x on each GPU, it's 4x on each GPU which is then given an offset/jitter and finally combined to give eight multi-samples and two texture samples per pixel.
This is simply not possible on single cards since they can't jitter and combine like that so a single card needs to apply 8xAA to the entire scene.
A dual-core processor is not one CPU, it's two combined onto one die to use one socket.
It was never a theory, and I clarified it for you in the "Replies here are out of control" thread since you wanted to troll there.
You were simply trying to prove the point that the hardware does not necessarily need a driver to throw an image on your screen. That is correct. But in the context of making the card useful and enjoying the $600+ you would pay for it, you would probably play games and therefore need the driver.
That is what BFG10K and I were telling you. Unless you prefer to play games in safe mode, not having the driver (which has to have SLI enabled through it for the card to work properly) makes the card worthless.
So you would consider a single P4 CPU with Hyperthreading as 2 CPUs since software sees it exactly the same way it sees a dualcore CPU?
I would say this:
In terms of ingenuity and design, no, nvidia hasn't produced the fastest single GPU with the 7950GTX
In terms of ingenuity and design, no, nvidia hasn't produced the fastest single GPU with the 7950GTX
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter