So - 6800u - Low Volume Review Board...
- Thread starter whql
- Start date
Well, out of three review samples we know about, I believe PCPop thought theirs was 16-pipe, Ufon received an 8-pipe one, and that HK site seemed to get a normal 12-pipe one. Fun!plat said:to be fair, has any other review site reviewed a 6800nu with different clock speeds?
Let me preface all of this by saying I think we're splitting hairs...
But my point is fairly simple: without that "bunch of wires between multiple layers of plastics," you don't have *any* interface between the host system bus, be it AGP, PCI, PCIe, or whatever, and your gpu. So obviously, the role of the pcb is critical. You seem to be quite detailed except when discussing the pcb, which you see apparently as inert and "just a bunch of wires between multiple layers of plastics"...
The gpu pcb (the pcb the gpu is mounted on), too, must support the host bus interface, right down to the connecting pins, or the gpu mounted on the pcb won't be functional in the host system. What is there in the gpu or substrate, for instance, that supports voltage changes and regulation from one host-bus standard to another (eg, the differences between AGPx1 (PCI66), x2-x4 and AGP x4-x8 )? All of that circuitry is extraneous to the gpu or the gpu substrate and is a component of the pcb on which the gpu and substrate is mounted. Clearly, the pcb for these products is "active" and critical in every sense of the word. In fact, I guess we need to make a distinction between gpu and gpu substrate, as that distinction also seems lost in the wash...
Which must be connected to the host bus via the pcb circuitry, otherwise there is a disconnect between the host and the gpu and the gpu will not function in the host system. The gpu connects with the gpu pcb which in turn connects with the pins in the physical mboard slot, the host system bus's electrical interface with the card.
I'm referring to the pcb on which the gpu is mounted. The gpu interfaces with the pcb, but the pcb interfaces with the host system bus. Instead of pcb, would you be more comfortable if I called it a "card", instead?
Think about all of the various PCI gpus nVidia did, and on how their nomenclature changed while still running on PCI--ditto, AGP. You might expect them do that, but I wouldn't, because all of their previous chips received numbers based on architectural differences--with gaps of "5" denoting newer architures. What was the difference between the substrate for nV30 & nV35, or nV35 and nV38, for instance? All were AGP x4-x8 reference designs, as I recall.
IMO, nVidia has departed from its traditional nomenclature by referring to an nV40 mounted on a PCIe-interfacing pcb as "nV45." Now, if they'd have called it "nV43," instead, I'd have had no comment, as it's only the "5's" that have traditionally denoted new nVidia architectures (eg., nV10, nV15, nV20, nV25, nV30, nV35, & nV40--nV30 & nV35 were both AGP x4-x8 reference designs, and nV35 happened as it did because nVidia cancelled nV30.) Which chip among those listed here was an exact duplicate of the one before, the only difference being its connection to the host system bus? That's my point.
Look at it this way: what is nV40 in terms of difference with nV45, as reported thus far? It's nV40 without the bridge, mounted on a pcb designed to electrically support and connect to an AGP x4-x8 host bus slot. nV45, then, is nV40 *with* the bridge, mounted on a pcb designed to electrically support and connect to a PCIe host bus slot. Pretty simple, it seems to me. Unlike the differences between nV25 and nV30/35, which were architectural quite apart from substrate considerations.
Well, the info is still buried somewhere on the ATi site, I would presume, as ATi put it there circa R300, and I referenced it in similar threads here months ago. R300 would have been just as comfortable with PCI66 as AGPx8, according to ATi, but it's for certain you could not have taken an R300 mounted on an AGPx1 pcb and plugged it into an AGPx8 motherboard slot, could you? Voltage support and pins on the pcb are different--but the gpu is the same. Which was my point. ATi only offered the R300 in AGP x4-x8 reference designs, but that had nothing to do with what was theoretically possible to do with the vpu in terms of host-bus interface.
But my point is that an R300 mounted on a PCI66(AGPx1) pcb can't do any "talking" to the outside world at AGP x8, unless the same gpu is mounted on an AGP x8 pcb which can handle the pin connections with the host bus and the differences in the voltages. This is why you can't say the pcb is irrelevant...
But it is *the same nV40* that's doing the talking, isn't it?... I just don't see how you can say with a straight face that "there are no architectural differences between nV40 and nV45," which is true, and then add, "But the differences between the nV40 and nV45 reference design are not extraneous to the gpu".... Make up your mind...
incurable said:
But my point is fairly simple: without that "bunch of wires between multiple layers of plastics," you don't have *any* interface between the host system bus, be it AGP, PCI, PCIe, or whatever, and your gpu. So obviously, the role of the pcb is critical. You seem to be quite detailed except when discussing the pcb, which you see apparently as inert and "just a bunch of wires between multiple layers of plastics"...
The gpu pcb (the pcb the gpu is mounted on), too, must support the host bus interface, right down to the connecting pins, or the gpu mounted on the pcb won't be functional in the host system. What is there in the gpu or substrate, for instance, that supports voltage changes and regulation from one host-bus standard to another (eg, the differences between AGPx1 (PCI66), x2-x4 and AGP x4-x8 )? All of that circuitry is extraneous to the gpu or the gpu substrate and is a component of the pcb on which the gpu and substrate is mounted. Clearly, the pcb for these products is "active" and critical in every sense of the word. In fact, I guess we need to make a distinction between gpu and gpu substrate, as that distinction also seems lost in the wash...
Which must be connected to the host bus via the pcb circuitry, otherwise there is a disconnect between the host and the gpu and the gpu will not function in the host system. The gpu connects with the gpu pcb which in turn connects with the pins in the physical mboard slot, the host system bus's electrical interface with the card.
WaltC said:
I'm referring to the pcb on which the gpu is mounted. The gpu interfaces with the pcb, but the pcb interfaces with the host system bus. Instead of pcb, would you be more comfortable if I called it a "card", instead?
Think about all of the various PCI gpus nVidia did, and on how their nomenclature changed while still running on PCI--ditto, AGP. You might expect them do that, but I wouldn't, because all of their previous chips received numbers based on architectural differences--with gaps of "5" denoting newer architures. What was the difference between the substrate for nV30 & nV35, or nV35 and nV38, for instance? All were AGP x4-x8 reference designs, as I recall.
IMO, nVidia has departed from its traditional nomenclature by referring to an nV40 mounted on a PCIe-interfacing pcb as "nV45." Now, if they'd have called it "nV43," instead, I'd have had no comment, as it's only the "5's" that have traditionally denoted new nVidia architectures (eg., nV10, nV15, nV20, nV25, nV30, nV35, & nV40--nV30 & nV35 were both AGP x4-x8 reference designs, and nV35 happened as it did because nVidia cancelled nV30.) Which chip among those listed here was an exact duplicate of the one before, the only difference being its connection to the host system bus? That's my point.
Look at it this way: what is nV40 in terms of difference with nV45, as reported thus far? It's nV40 without the bridge, mounted on a pcb designed to electrically support and connect to an AGP x4-x8 host bus slot. nV45, then, is nV40 *with* the bridge, mounted on a pcb designed to electrically support and connect to a PCIe host bus slot. Pretty simple, it seems to me. Unlike the differences between nV25 and nV30/35, which were architectural quite apart from substrate considerations.
Well, the info is still buried somewhere on the ATi site, I would presume, as ATi put it there circa R300, and I referenced it in similar threads here months ago. R300 would have been just as comfortable with PCI66 as AGPx8, according to ATi, but it's for certain you could not have taken an R300 mounted on an AGPx1 pcb and plugged it into an AGPx8 motherboard slot, could you? Voltage support and pins on the pcb are different--but the gpu is the same. Which was my point. ATi only offered the R300 in AGP x4-x8 reference designs, but that had nothing to do with what was theoretically possible to do with the vpu in terms of host-bus interface.
But my point is that an R300 mounted on a PCI66(AGPx1) pcb can't do any "talking" to the outside world at AGP x8, unless the same gpu is mounted on an AGP x8 pcb which can handle the pin connections with the host bus and the differences in the voltages. This is why you can't say the pcb is irrelevant...
But it is *the same nV40* that's doing the talking, isn't it?...
I just don't see how you can say with a straight face that "there are no architectural differences between nV40 and nV45," which is true, and then add, "But the differences between the nV40 and nV45 reference design are not extraneous to the gpu".... Make up your mind...thatdude90210
Regular
There... officially shipping.
Whoa! If we can get some confirmation on that card being sold I guess I might have been wrong with my July prediction?!?!
Bouncing Zabaglione Bros.
Legend
They shipped those out last Tuesday and I've just finished testing one. Max clock for some single tests was 492/1302, but way back at 475/1250 to run everything thrown at it.
Gainward have a BIOS that runs the board at 470/1200 from boot.
It's not really worth the money, but if your disposable income is high enough, it's the fastest graphics card in the world at the time of writing(with the big disclaimer that no XT PE is shipping in the same volume).
Rys
Gainward have a BIOS that runs the board at 470/1200 from boot.
It's not really worth the money, but if your disposable income is high enough, it's the fastest graphics card in the world at the time of writing(with the big disclaimer that no XT PE is shipping in the same volume).
Rys
No that's not possible, people say the 6800s are not shipping, so should be much morejvd said:
And people buys the card @ 600£
http://www.komplett.co.uk/k/ki.asp?sku=300416&cks=SER
I'm not denying that they've sold a few Evildeus, but do you really consider their salting of the market as "it's out!!!"? :|
Evildeus said:No that's not possible, people say the 6800s are not shipping, so should be much morejvd said:
And people buys the card @ 600£
http://www.komplett.co.uk/k/ki.asp?sku=300416&cks=SER
Guess they aren't shipping in volume.
Hell someone can sell me a vodoo 5 6000 and i guess that means that it shipped ? Or someone can sell me a ramapage and that means it shipped ?
Well I have to agree with DW and jvd on this one , and also Evildeus, they are shipping but only certain models and then you have to be quick.
I ordered one of the CoolFX's but moved my order from Komplett to Dabs, which was probably a mistake as I know Dabs got some last week but mine still says awaiting stock.
I've actually cancelled it now as the Gainward Extreme GS 2600 with the beefed up air cooling is available in August and does 450Mhz for a lot less.
Rys, from being nvidia's golden boy with the 5700U you have now become the Ati golden boy with first your review of the 6800 using 8xAA at 1024x768 ( it's worst possible setting ) to using the Gainward with trliniear optimisations off and Ati with optimisations on .. so now the playing field is possible to be level again you decide not to.
Ah, but you will say, Ati optimisations look better than nvidia ones .. which you can argue, but you have just turned an opjective test into a subjective test, which is not what reviewers should be doing.
Or at least turn it all on and turn it all off and give screenshots to show the difference.
I ordered one of the CoolFX's but moved my order from Komplett to Dabs, which was probably a mistake as I know Dabs got some last week but mine still says awaiting stock.
I've actually cancelled it now as the Gainward Extreme GS 2600 with the beefed up air cooling is available in August and does 450Mhz for a lot less.
Rys, from being nvidia's golden boy with the 5700U you have now become the Ati golden boy with first your review of the 6800 using 8xAA at 1024x768 ( it's worst possible setting ) to using the Gainward with trliniear optimisations off and Ati with optimisations on .. so now the playing field is possible to be level again you decide not to.
Ah, but you will say, Ati optimisations look better than nvidia ones .. which you can argue, but you have just turned an opjective test into a subjective test, which is not what reviewers should be doing.
Or at least turn it all on and turn it all off and give screenshots to show the difference.
