ok, you've got me curious, I have a 9500 pro, the ram is in a straight line at the top of the card. I modded it, and now it shows up as a 9700 pro in winxp. I'll get that riva tuner thing to see what it says. I only see you guys mention the cards with the L shaped ram, is my card something special?
it may say that, but the only way you can really tell is if it is a 256mbit bus and not 128mbit. also, the pixel shader will tell you if it is a 9700pro if after the mod it was higher than before
If you flash the stright line version then you are likely for it to go very very wrong, as you are still doing the same approach as if the card is moddable.
What happened was that the folks doing the mods found that more of the ones with RAM in a straight line had "broken pipes" that were disabled. The mfrs knew this. They also typically stuck slower RAM chips on these cards.
So, what then happens?? First, the broken pipes actually are either not useable or display junk and, second, if the speed is tweaked beyond the RAM limits the RAM overheats. Since it is slower RAM, it does not overclock as far without dying or needning cooling help (chip heatsinks let a person get a bit more speeed out of it but not much).
Not just shape, but components used for that design by mfring specs.
it's infineon 3.3 ns ram, isn't that the good stuff? i thought the slower cards used 3.6 ns ram. I ran 3dmark2k3, the second test didn't show any crap on screen.
You got an exception to the general rule. You might want baby RAM chip heat sinks, though.
What happened, is that ATI bin sorts cores. The ones that test with broken pipes go to be sold for non-por cards. The ones that test no broken pipes go to be sold for pro cards.
Two places the wrong (better) core can be routed ot end up on a non-pro blank--
ATI's fab sorting might have stuck one good core in a not-perfect bin. then it was sold as cheaper and installed in card assembly (they are not tested just before wave-soldering on).
The assembler of card picked wrong part as card was bneing assembled, or the card mfr managed to mis-sort some cores to wrong assmbly station.
Why this is valid-- The card mfr's do not test for overability of cards, because testing all cards woudl cost more than the revenue from finding and relabelling the more functional cards. Thus, it was never OC tested and just tested to make sure it woudl work for what it was sold as.
There is one more important area of the card, though, and that is the oscillators that are long and narrow components that time the overall card. Those are hard to tweak. So, you might have everything except the right signal generating oscillator and end up with a card that can be OC'd some, show as the more powerful card, but not fully perform up to speed of the more powerful card due to the "heartbeat" or "pulse" signal being too slow.
Changing the signal generator might or might not work, if you know what signal gens are used on the better card and can milspec solder and desolder. Otherwise, I think you might get mediocre results unless someone completely built a pro hardware set on a non-pro blank and it got shipped.
Your card is a misbuild. Built too WELL and then undersold. That is the only thing from all the research I have been doing on this that makes sense.
What is likely with your card is that you can time to 275 MHZ RAM and have it stable, possibly lose it at 300 MHz, due to signal oscillator. I have a Sapphire like that here.
a) hynix 3.6ns memory blows. My hynix 3.3ns doesn't go far at all.
b) Pro speeds are 325/310 unless you are referring the 9500 pro which is 275/270 just like the 9700 np.
I have a 128mb 9500 non-pro and tried to softmod it via the Rivatuner script. It said that it was 256 bit but I got the checkerboard. It wasn't that bad, just in a few spots but it's still annoying. Is there any way to undo this? I tried un-installing and re-installing the drivers but I had no luck.
Thx.
Comments
so does this mean that all pipelines are active now?
I dont see how you could change the Bus though, as surely they would use different chips.....
NS
i think it gives a bit of a spedd increase but not sure
NS
So, what then happens?? First, the broken pipes actually are either not useable or display junk and, second, if the speed is tweaked beyond the RAM limits the RAM overheats. Since it is slower RAM, it does not overclock as far without dying or needning cooling help (chip heatsinks let a person get a bit more speeed out of it but not much).
Not just shape, but components used for that design by mfring specs.
The good RAM is the Hynix 3.6 and the Samsung 2.86ns. People think that the Hynix may actually be rated wrong as it performs so well.
Sorry Kanez.
NS
What happened, is that ATI bin sorts cores. The ones that test with broken pipes go to be sold for non-por cards. The ones that test no broken pipes go to be sold for pro cards.
Two places the wrong (better) core can be routed ot end up on a non-pro blank--
ATI's fab sorting might have stuck one good core in a not-perfect bin. then it was sold as cheaper and installed in card assembly (they are not tested just before wave-soldering on).
The assembler of card picked wrong part as card was bneing assembled, or the card mfr managed to mis-sort some cores to wrong assmbly station.
Why this is valid-- The card mfr's do not test for overability of cards, because testing all cards woudl cost more than the revenue from finding and relabelling the more functional cards. Thus, it was never OC tested and just tested to make sure it woudl work for what it was sold as.
There is one more important area of the card, though, and that is the oscillators that are long and narrow components that time the overall card. Those are hard to tweak. So, you might have everything except the right signal generating oscillator and end up with a card that can be OC'd some, show as the more powerful card, but not fully perform up to speed of the more powerful card due to the "heartbeat" or "pulse" signal being too slow.
Changing the signal generator might or might not work, if you know what signal gens are used on the better card and can milspec solder and desolder. Otherwise, I think you might get mediocre results unless someone completely built a pro hardware set on a non-pro blank and it got shipped.
Your card is a misbuild. Built too WELL and then undersold. That is the only thing from all the research I have been doing on this that makes sense.
What is likely with your card is that you can time to 275 MHZ RAM and have it stable, possibly lose it at 300 MHz, due to signal oscillator. I have a Sapphire like that here.
NS
b) Pro speeds are 325/310 unless you are referring the 9500 pro which is 275/270 just like the 9700 np.
Oh well, the RAM quality that I mentioned seems to be the general sensus on the Rage3D Site....
NS
Have you flashed it with Pro BIOS at all?
NS
Wonder how far I can push it, have you tried going any further?
NS
Thx.