Tech Report: CrossFire DUAL SLAVE !
- Thread starter Jawed
- Start date
Unknown Soldier
Veteran
Jawed said:
Actually I was refering to that .. in case someone didn't read the article.
DaveBaumann said:
Is the sole factor in this the card design or will new motherboards help too? Does the current setup of two fairly vanilla 8x slots present problems with bandwidth (solveable by going to proper dual 16x?) or latency (solveable with a PCIe implementation focused on lower latency between 16x slots?)? Or are current/near-future dual-slot mobos "good enough" for the forseeable future?
Anybody see the Inq SLI2 piece recently? I wonder if NV is thinking of trying to shift any special hardware capability out of the graphics cards (other than the connector, of course) into the mobo chipset. If so, maybe what they are after is trying to maintain incompatibility for marketing reasons. . .make their mobos incompatible with ATI's solution, and make their SLI mobo chipset a real technical requirement (i.e. impossible to unhack a fake technical requirement enforced in the drivers) to SLI NV cards.
Unknown Soldier
Veteran
Wouldn't they just be shooting themselves in the foot? I mean, most people think that Nvidia mobo's are tops .. but if those mobo's become incompatible with other graphics cards .. then they are gonna loose bigtime imo.
US
US
This is kind of picky, but I wish folks would be more careful with the word "dongle"... Its casual use in the current context has me thinking there's some blurriness surrounding the idea of just what an actual "dongle" is and what a dongle does.
Literally, a hardware "dongle" is a device appendage which has no other function whatsoever than to authorize use of the device to which it is attached. A dongle itself has no native function and contributes nothing to the operation of the device aside from authorizing the device to function. For instance, some of my old computer-controlled A/B roll editing hardware required the presence of a pass-through hardware dongle attached to my parallel port. Although the dongle had no native function itself, its presence on the parallel port was required else the A/B roll editing hardware would not function. This step was taken by the manufacturer to curtail the pirating of the A/B roll-editing software (which was custom, although the hardware it controlled was 98% generic.) Dongles are by nature annoying and displeasing for all the obvious reasons, and hence the term "dongle" itself is generally used with at least a hint of sarcasm...
In this case I see nothing "dongle-ish" about the methods used to connect a pair of SLI or X-Fire cards so that they may be synchronized to operate in tandem, because these connectors regardless of superficial physical form are absolutely required for the pair of cards to function as a unit. Hence, their functional purpose is far greater than that of a mere dongle. Calling these connectors "dongles" is no more appropriate than, say, calling the VGA connector cable between monitor and graphics card a "dongle," or the cord that connects a keyboard to a USB port a "dongle."
That aside, there is one aspect of nV SLI that I do indeed consider operates like a dongle, but it has nothing to do with the way the two nV cards are physically connected to each other. I am referring to the hardware on the dual-slot PCIe motherboards that has to be manually set by the user so that the motherboard will operate in dual-slot PCIe mode so that nV SLI is supported. Along with the fact that mobo makers are restricted from even advertising "SLI-capable" motherboards without paying nV a licensing fee, a dual-slot PCIe mboard ought to autoconfigure through its bios for dual-slot operation necessary to support *either* SLI or X-Fire, it seems to me. The fact that the PCIe bus specs as promulgated by Intel natively support dual-slot operation should have meant that a proprietary switching solution--especially one consisting of hardware on the motherboard external to the core logic--should be unnecessary. To this end it would seem the X-Fire switching, which apparently does not require control circuitry on the mboard external to the core logic, is much closer to the original Intel specification.
Its casual use in the current context has me thinking there's some blurriness surrounding the idea of just what an actual "dongle" is and what a dongle does.Literally, a hardware "dongle" is a device appendage which has no other function whatsoever than to authorize use of the device to which it is attached. A dongle itself has no native function and contributes nothing to the operation of the device aside from authorizing the device to function. For instance, some of my old computer-controlled A/B roll editing hardware required the presence of a pass-through hardware dongle attached to my parallel port. Although the dongle had no native function itself, its presence on the parallel port was required else the A/B roll editing hardware would not function. This step was taken by the manufacturer to curtail the pirating of the A/B roll-editing software (which was custom, although the hardware it controlled was 98% generic.) Dongles are by nature annoying and displeasing for all the obvious reasons, and hence the term "dongle" itself is generally used with at least a hint of sarcasm...
In this case I see nothing "dongle-ish" about the methods used to connect a pair of SLI or X-Fire cards so that they may be synchronized to operate in tandem, because these connectors regardless of superficial physical form are absolutely required for the pair of cards to function as a unit. Hence, their functional purpose is far greater than that of a mere dongle. Calling these connectors "dongles" is no more appropriate than, say, calling the VGA connector cable between monitor and graphics card a "dongle," or the cord that connects a keyboard to a USB port a "dongle."
That aside, there is one aspect of nV SLI that I do indeed consider operates like a dongle, but it has nothing to do with the way the two nV cards are physically connected to each other. I am referring to the hardware on the dual-slot PCIe motherboards that has to be manually set by the user so that the motherboard will operate in dual-slot PCIe mode so that nV SLI is supported. Along with the fact that mobo makers are restricted from even advertising "SLI-capable" motherboards without paying nV a licensing fee, a dual-slot PCIe mboard ought to autoconfigure through its bios for dual-slot operation necessary to support *either* SLI or X-Fire, it seems to me. The fact that the PCIe bus specs as promulgated by Intel natively support dual-slot operation should have meant that a proprietary switching solution--especially one consisting of hardware on the motherboard external to the core logic--should be unnecessary. To this end it would seem the X-Fire switching, which apparently does not require control circuitry on the mboard external to the core logic, is much closer to the original Intel specification.
Bouncing Zabaglione Bros.
Legend
Unknown Soldier said:
It is a risky strategy, that can only work if Nvidia has a total stranglehold on the market (ie Microsoft levels of dominance). If everyone is buying Nvidia motherboards anyway, then this would lock everyone into also buying Nvidia graphics cards for those that want to use SLI, or might want the option to use SLI in the future.
Of course the fact is that Nvidia does not have that kind of dominance, so is just as likely to drive people toward ATI, Intel, or VIA based motherboards as they are to encourage people towards Nvidia graphics cards.
People buy overspecced motherboards and graphics cards because of the extra features that they may want to use at some point in the future. By restricting users to Nvidia graphics cards for one of the major gaming components would go against this "keep your options open for future proofing" philosophy, and could very well work against Nvidia in the long run.
Nvidia may try to walk a very fine line and only include this restriction for multi-card setups, and still allow any single card regardless. There might be legal issues due to various competition laws, though no doubt Nvidia would cite testing and interoperability issues even though the cat is out of the bag because we now know they cripple cheaper SLI-capable chipsets in their drivers for marketing reasons.
WaltC said:
My understanding is that all the switch card does is reroute eight PCIe channels from the back end of slot one to the front end of slot two. The DFI version uses sets of jumpers instead, which suggests that there's no additional electronics needed. Furthermore, while ATI is pushing BIOS-enabled switching, there are still going to be some Crossfire mobos with an external switching card just like nVidia's because it's cheaper - doing it in BIOS requires a set of switching chips on the mobo, which cost a dollar or so a pop and you need five or six. There are actually SLI boards on the way, I believe, with BIOS-based switching, as it's not a huge extra cost for a high-end board.
Walt, this is really ridiculous. First, you present a lengthy explanation how the word "dongle" has been misused, then you misuse it just a few lines on something that is nothing like the definition you told before. This routing card has absolutely nothing to do with authorization, it is a connector just like a cable is, and in those specific implementations is absolutely required for the pair of cards to work. Just like the Crossfire cable is required in this specific implementation, but could be replaced by something else.WaltC said:
Calling these connectors "dongles" is no more appropriate than, say, calling jumpers on a mainboard "dongles".
Do you think Crossfire certification comes for free?
This implementation of switching between two slots working in 8x mode or one slot working in 16x mode is completely within the PCIe spec, it's not at all tied to SLI and not even all SLI boards have it. It's just cheaper, because you need less traces and no special switch chip.
The board manufacturers could even have designed the boards with two 8x slots and leave it at that - but that's just bad for marketing.
Oh, and I wish people would stop replacing "cross" or similar things with "X". That is just sooooo annoying...
http://www.theinquirer.net/?article=23593
Here's the actual article. Thinking back on Josh's interview with NV, maybe they're after making the pcie connection wide enough to do away with the intra-card connector without impacting performance? Folks might like that. Don't know what it would do to compatibility, if anything.
Here's the actual article. Thinking back on Josh's interview with NV, maybe they're after making the pcie connection wide enough to do away with the intra-card connector without impacting performance? Folks might like that. Don't know what it would do to compatibility, if anything.
Blazkowicz
Legend
I agree there's a negative connotation with the word 'dongle', and that there isn't a dongle on either ATI or NV multi-GPU.
(what's fun with these old dongles is that they still could be bypassed by cracking the software..)
(what's fun with these old dongles is that they still could be bypassed by cracking the software..)
Blazkowicz
Legend
geo said:
maybe the doubling of PCIE lane is for the quad GPU SLI, we already saw the gigabyte board doing this.
NV might prepare a chipset with double the lanes instead of using gygabite's solution of having both a nforce SLI intel edition and a nforce 4 SLI.
Xmas said:
The reason why I might loosely label that particular aspect as akin to a dongle is this: the core logic ought to be able to seamlessly configure a dual-slot PCIe-board without the need for manual "rerouters" or jumpers or any manual switches. The dual-slot PCIe config, remember, isn't custom or alien in any way to the PCIe spec advanced by Intel since the start. What nV's done, here, seems like deliberately configuring its nf chipset not to auto-switch transparently and to *require* some kind of artificial, chipset-extraneous mechanism to do the switching so that nV can call it proprietary and charge licensing fees for use of the circuitry. (Which makes sense considering that apparently nV has deliberately departed in this area from the Intel spec.) Again, of course, there's nothing dongle-ish about connecting the two cards to each other--it's what's happening at the chipset level on the mboard that I'm talking about.
IE, the manual switching mechanism is simply not needed in my view and is simply an artificially manufactured requirement for nF chipsets so as to make support for nV SLI proprietary and income producing.
Xmas said:Oh, and I wish people would stop replacing "cross" or similar things with "X". That is just sooooo annoying...
Heh...
Yes, maybe we should abandon acronyms altogether. I mean, why write "IBM" when it's so much less annoying to write "International Business Machines"...? (Or is it?...)Seriously, acronyms are generally used to alleviate annoyance as opposed to creating it. If people understand that "X-Fire" means "Crossfire" (and in this context it's difficult to see how that might be misunderstood), what's the harm? Using lengthy marketing verbosity inside informal posts is what I call "annoying," but that's just me...
Walt, at what point does this
have anything to do with the original topic? Here I am reading a thread that I thought was going to explain how two slave cards can work together and you come in and ramble on about dongles and how ATI is much closer to the original Intel specification than Nvidia is. I guess if you have an opportunity to bash Nvidia, you take it eh Walt?
have anything to do with the original topic? Here I am reading a thread that I thought was going to explain how two slave cards can work together and you come in and ramble on about dongles and how ATI is much closer to the original Intel specification than Nvidia is. I guess if you have an opportunity to bash Nvidia, you take it eh Walt?
WaltC said:
In the midst of all your ranting you do know that ATi's Crossfire motherboards have several options for configuring the dual-GPU mode, one of them being the use of a similar "switching card" used in most SLI motherboards. The software option is simply the most expensive. Also, as demonstrated by Asus it is quite possible to implement software based switching for SLI as well (for which they also charge a premium). So exactly what is your point again?
WaltC said:
Of course it isn't. What Nvidia pioneered though was making the routing of the PCI-Express lanes reconfigurable by the end user. Intel's solution for dual PCIE graphics so far has been dedicating 16 lanes to one slot and 4 to the other, no configuration there at all 'seamlessly' or otherwise. Nvidia was the first to offer the choice between 1x16 or 2x8, and ATI has now opted for the same solution.
I'm not sure if Intel *has* a spec for end user configurable PCI-Express lanes. Perhaps you have a link to an Intel spec that Nvidia deliberately departed from.
The manual switching system was simply the first method that was available. The existence of Asus' soft-SLI confirms that it's neither needed nor a requirement. I suspect software switching will eventually become the norm on SLI motherboards as well because it is more convenient.