How NVIDIA made the 9600 GT gain extra performance..secretly

kllrnohj said:
Er, what? It isn't cheating... What is the difference between LinkBoost over clocking your card or BFG, XFX, eVGA, MSI, etc...? I'd guess that most reviews of the 9600GT were running @ stock frequencies, anyway, and I doubt a whole lot of people overclock the PCI-E bus (since it doesn't really help with anything).

Besides, how can nvidia "secretly overclock" the card. They are the ones that define what "stock" is anyway :confused:
Click to expand...

Spot-on. Plus, if LinkBoost really is off by default, it means that YOU have to turn it on, thus you are, in fact, OCing your card. Nothing sneaky about that. Besides, even if it was defaulted to on, LinkBoost is a feature of an nVidia motherboard.
 
After thinking about how an Evga SSC 9600 GT, with a default core clock of 740 Mhz, could possibly survive on a 780i motherboard, which viper john alleges runs at pci-e bus speed of 125 Mhz....



A rough estimate of w1zzard's forumla would put the actual core speed of that 9600 GT over 900 Mhz...


I put it to the test myself. My 9600 GT SLi set-up is currently OC'd at 740 Mhz core clock. I re-booted and upped the PCI-E frequency to 105 MHz. I made it into Windows, but as soon as I ran 3dMark, the system froze.


I clocked the PCI-e bus back down to 100, ran a 3dMark06 loop and just played 30 minutes of Crysis without a hiccup.

So it made me wonder how could a 9600 GT SSC run on a 780i motherboard?

As this girl has so excellently done:

http://forums.slizone.com/index.php?showtopic=16709&st=0



But after seeing the pic in the link below, a BIOS screenshot of an EVGA 780i motherboard, courtesy of Kyle, it seems that the 780i can run at 100 Mhz bus speed.

http://enthusiast.hardocp.com/image.html?image=MTIwMDU0ODQ4OUM0bVNCVFd6TnRfMl81X2wuZ2lm



So is Viper John mistaken? Is it just reference boards that are clocked at 125 Mhz, but production boards default to 100 mhz?


Would this still be considered "cheating", if what w1zzard claims in his article is true?
 
Ogilvy said:
So is Viper John mistaken? Is it just reference boards that are clocked at 125 Mhz, but production boards default to 100 mhz?
Click to expand...

I can say after dealing with John for over 5 years, the man knows his stuff...
I doubt very much he would post anything he wasn't 100% sure of.

He has an EVGA board, so its possible he is only talking about EVGA 780i's...you can shoot him an email and ask, he is always happy to answer any questions about video cards and the like..:)
 
Ogilvy said:
So it made me wonder how could a 9600 GT SSC run on a 780i motherboard?
Click to expand...

Good question - and I have no idea...

But after seeing the pic in the link below, a BIOS screenshot of an EVGA 780i motherboard, courtesy of Kyle, it seems that the 780i can run at 100 Mhz bus speed.

http://enthusiast.hardocp.com/image.html?image=MTIwMDU0ODQ4OUM0bVNCVFd6TnRfMl81X2wuZ2lm



So is Viper John mistaken? Is it just reference boards that are clocked at 125 Mhz, but production boards default to 100 mhz?


Would this still be considered "cheating", if what w1zzard claims in his article is true?
Click to expand...

All motherboards (including the 780i) default to a PCI-E bus speed of 100mhz. That is the stock speed for a PCI-E slot. It is only when paired with an nvidia video card with linkboost enabled that the PCI-E bus jumps to 125mhz. It is a motherboard setting, not a video card one.

lol I love this line

Ya know, secretly ... the same way you secretly drive your car at 85mph even though the arbitrary sign on the side says it's only supposed to do 70
Click to expand...

Except that sign isn't made and enforced by the same company that makes my car, so your analogy doesn't work :p
 
Crosshairs said:
I can say after dealing with John for over 5 years, the man knows his stuff...
I doubt very much he would post anything he wasn't 100% sure of.

He has an EVGA board, so its possible he is only talking about EVGA 780i's...you can shoot him an email and ask, he is always happy to answer any questions about video cards and the like..:)
Click to expand...



Don't get me wrong...I don't mean to disparage Viper John in any way. I know he knows his stuff...that's why I cited him a few posts back thinking that his information is reliable. But just as w1zzard found the 27 mhz crystal curious, I found VJ's information and w1zzard's observations curious. The BIOS screenshot in my last post IS of an EVGA 780i motherboard, reviewed by [H].


The truth is...I don't know much about any of this.

But if the claim is, nvidia intended to cheat on reviewers benchmarks, it would require a motherboard that has a higher bus freq. than 100, as VJ stated, or a motherboard that can automatically raise the bus freq., which I believe is what LinkBoost is supposed to do.


But again, LinkBoost would have to be active...and if it was...would it blow out a highly-factory overclocked 9600 GT like the EVGA SSC? Or would it be smart enough to know when to stop increasing the bus frequency?
 
Maybe for it to default to the "Linkboost", the setup must be ESA compliant or at least have nTune installed which most avoid?
 
Ogilvy said:
But if the claim is, nvidia intended to cheat on reviewers benchmarks, it would require a motherboard that has a higher bus freq. than 100, as VJ stated, or a motherboard that can automatically raise the bus freq., which I believe is what LinkBoost is supposed to do.
Click to expand...

No one is really claiming that - and it would be stupid for nvidia to do it for those reasons. A lot of the sites I checked out were using an Intel chipset for the 9600GT benchmarks - no LinkBoost.

But again, LinkBoost would have to be active...and if it was...would it blow out a highly-factory overclocked 9600 GT like the EVGA SSC? Or would it be smart enough to know when to stop increasing the bus frequency?
Click to expand...

It seems to me that nVidia either forgot about LinkBoost, or forgot to tell their AIC partners about the possibility... LinkBoost isn't "smart" at all - its either on or off, 100mhz (off) or 125mhz (on). On my M2N32-SLI Delux (nforce 590), LinkBoost defaults to "off". Not sure about other boards. LinkBoost is also completely independent of software and drivers - it is a BIOS level "feature" (that doesn't really DO anything - it was a marketing move more than anything)

Since there haven't really been any people complaining about crashing 9600GTs and artifacting appearing in factory overclocked cards, I'm guessing very few people have overclocked PCI-E buses and/or LinkBoost enabled :D
 
Since there haven't really been any people complaining about crashing 9600GTs and artifacting appearing in factory overclocked cards, I'm guessing very few people have overclocked PCI-E buses and/or LinkBoost enabled
Click to expand...

That, or the 9600GT can OC a lot better than we all thought it could, xD
 
Kowan said:
Maybe for it to default to the "Linkboost", the setup must be ESA compliant or at least have nTune installed which most avoid?
Click to expand...


Which is kinda my point. There are a set of circumstances that are unlikely to happen for pci-e overclocking of a 9600 GT overclocking to kick-in. For a reviewer to unwittingly publish higher-than-normal benchmarks for the 9600 GT seems equally unlikely.


What ever is going on....It's an interesting find by w1zzard, though.
 
kllrnohj said:
No one is really claiming that - and it would be stupid for nvidia to do it for those reasons. A lot of the sites I checked out were using an Intel chipset for the 9600GT benchmarks - no LinkBoost.
Click to expand...


Well, Viper John certainly implied that reviewers could be posting false benchmarks...and "he knows his stuff" so someone did say it. And I checked a few reviews myself...there were quite a few using intel chipsets.



kllrnohj said:
It seems to me that nVidia either forgot about LinkBoost, or forgot to tell their AIC partners about the possibility... LinkBoost isn't "smart" at all - its either on or off, 100mhz (off) or 125mhz (on). On my M2N32-SLI Delux (nforce 590), LinkBoost defaults to "off". Not sure about other boards. LinkBoost is also completely independent of software and drivers - it is a BIOS level "feature" (that doesn't really DO anything - it was a marketing move more than anything)
Click to expand...


Well, it does something now, doesn't it?


kllrnohj said:
Since there haven't really been any people complaining about crashing 9600GTs and artifacting appearing in factory overclocked cards, I'm guessing very few people have overclocked PCI-E buses and/or LinkBoost enabled :D
Click to expand...


Certainly true...no one's complaining yet. It will be interesting to see if someone does burn out a 9600 GT.



I know I have belabored this issue...it's because I own a couple of 9600 GTs and I was planning on getting a 780i motherboard.
 
Ogilvy said:
Well, Viper John certainly implied that reviewers could be posting false benchmarks...and "he knows his stuff" so someone did say it. And I checked a few reviews myself...there were quite a few using intel chipsets.
Click to expand...

While yes, it is theoretically possible that a reviewer was posting "false" benchmarks, it really wouldn't be false at all. It is something the end user may experience as well. However, from what I've seen over clocking tends to not help all that terribly much, so the results aren't going to be that skewed anyway. And since most major sites were using Intel chipsets, this whole point is moot.

Well, it does something now, doesn't it?
Click to expand...

Yes, but somehow I doubt that a couple years ago when they dreamed up LinkBoost they suddenly thought "Hey, in a couple of years we can release a midrange card that magically gets faster with linkboost by using a divider on the PCI-E bus to control the core clock speeds!" :D

I know I have belabored this issue...it's because I own a couple of 9600 GTs and I was planning on getting a 780i motherboard.
Click to expand...

This shouldn't affect your decision one way or another. You can still overclock the card the regular way, and most motherboards let you adjust the PCI-E bus manually anyway. LinkBoost isn't anything special, just a gimmick.
 
So in the end what's the difference of cranking the PCI-E speed for extra mhz vs just doing it yourself. :? What's the big deal?
 
The "deal" is AMD can't handle nVidia or Intel so this is the sort of thing they come up with.
 
watch these cards start burning out lol. and then they wont warranty it, it was "overclocked"
 
I cant believe some of you are defending these shady practices by nvidia. The fact is that this could be used to fool the average consumer into thinking that nvidia mobo + nvidia video card > other mobo + nvidia video card.

It's not the point that we could all overclock our pcie freq to match or that we could buy a factory overclocked 9600gt and get the same results, it is the fact that it happens without us explicitly knowing! I mean at least the video card box says "xtreme OC version" or something similar to let us know what we are getting into. I have never seen any thing like "faster with an nvidia motherboard"???

This whole thing just seems fsucked up...:confused:
 
ryko said:
I cant believe some of you are defending these shady practices by nvidia. The fact is that this could be used to fool the average consumer into thinking that nvidia mobo + nvidia video card > other mobo + nvidia video card.

It's not the point that we could all overclock our pcie freq to match or that we could buy a factory overclocked 9600gt and get the same results, it is the fact that it happens without us explicitly knowing! I mean at least the video card box says "xtreme OC version" or something similar to let us know what we are getting into. I have never seen any thing like "faster with an nvidia motherboard"???

This whole thing just seems fsucked up...:confused:
Click to expand...

Not at all. As has been previously mentioned, LinkBoost is a feature of nVidia motherboards. Therefore, it's perfectly alright that an nVidia mobo with LinkBoost should be able to produce an overclock as it supposedly does. And from what has been related so far, this doesn't happen without you explicitly knowing. LinkBoost is a BIOS feature, one that can be turned on or off (and which seems to start OFF).
 
Ogilvy said:
Don't get me wrong...I don't mean to disparage Viper John in any way. I know he knows his stuff...that's why I cited him a few posts back thinking that his information is reliable. But just as w1zzard found the 27 mhz crystal curious, I found VJ's information and w1zzard's observations curious. The BIOS screenshot in my last post IS of an EVGA 780i motherboard, reviewed by [H].
Click to expand...

I saw that screenshot after I posted , I'm more confused now than I was before....and thats pretty bad....
 
TerranUp16 said:
Not at all. As has been previously mentioned, LinkBoost is a feature of nVidia motherboards. Therefore, it's perfectly alright that an nVidia mobo with LinkBoost should be able to produce an overclock as it supposedly does. And from what has been related so far, this doesn't happen without you explicitly knowing. LinkBoost is a BIOS feature, one that can be turned on or off (and which seems to start OFF).
Click to expand...

yes,yes...you and i and every one else here understand about how this is a "feature" that can be turned on or off just how something can be overclocked or underclocked, but it just seems like the marketing folk got together with the engineers and said something like, "well can't we just make our cards work faster with our boards?"

and have we even determined if this is defaulted to off on the 680i, 780i? we know linkboost is not present on 650i or 750i, correct? and then someone mentioned ESA and ntune? what if when a 9600gt is installed, it automatically turns linkboast on? have we proven this doesn't happen? we need more info...
 
Crosshairs said:
I saw that screenshot after I posted , I'm more confused now than I was before....and thats pretty bad....
Click to expand...


It might be like this....


w1zzard discovered that 9600 GT's can be overclocked by the pci-e bus frequency...tied it into an obscure, rarely used nvidia technology called LinkBoost....then trumped it up a bit as something "shady" to get some attention for his website. Perhaps Viper John was trying to give him a "solid" toward that end.
 
ryko said:
and have we even determined if this is defaulted to off on the 680i, 780i? we know linkboost is not present on 650i or 750i, correct? and then someone mentioned ESA and ntune? what if when a 9600gt is installed, it automatically turns linkboast on? have we proven this doesn't happen? we need more info...
Click to expand...

ntune has no mention of LinkBoost at all for me. As I understand it, it is *supposed* to be completely autonomous.

As for the 780i, I'm guessing it too defaults to off, as the latest 780i motherboard on the [H] table shows a PCI-E speed of 100mhz when paired with an 8800GTX ( http://enthusiast.hardocp.com/image.html?image=MTIwMTM2NzM1NGlla2NOOTBwbXVfMl81X2wuZ2lm )

I'm looking to hopefully sell some old parts and buy a 9600GT here soon, so if someone hasn't verified before then I will :)
 
My question is: Will I be able to plug a 9X00 card into my Intel-based mobo and crank the PCIe speed up to overclock it?
 
dirtypop said:
My question is: Will I be able to plug a 9X00 card into my Intel-based mobo and crank the PCIe speed up to overclock it?
Click to expand...

If you can't, then it's because your mobo lacks such a feature and is not because nVidia is being "unfair". But I dunno why you'd want to OC through the PCI-E slot anyway.
 
TerranUp16 said:
If you can't, then it's because your mobo lacks such a feature and is not because nVidia is being "unfair". But I dunno why you'd want to OC through the PCI-E slot anyway.
Click to expand...

You wouldn't have to download extra software or use 3rd-party apps to overclock.
 
dirtypop said:
You wouldn't have to download extra software or use 3rd-party apps to overclock.
Click to expand...

Except those apps provide very useful information, such as GPU temp monitoring and artifact scanning - so you will still want them. But to answer your question, yes, you could.
 
Well since a lot of people seem to be a little slow let me break it down. Let's assume Nvidia is doing this shady thing to make their motherboards look better (presumably so they can sell more of them).

Sequence of events

1. Nobody cares about Linkboost
2. People notice that 9600GT is faster with Linkboost enabled
3. Somebody with a smidgen of common sense links the boost to the PCIe bus increase
4. Somebody else also with a smidgen of common sense (or maybe even the same guy in #3) overclocks the PCIe bus on a non-Nvidia board.
5. Everybody in the world knows that they can overclock the PCIe bus on any board and get the same increase
6. Nobody cares about Linkboost

See how that works? IMO you have to be really special to think that Nvidia did this as some sort of shady marketing gimick. Also amusing is how overclocking the PCIe bus magically went from something that nobody does to something that reviewers may be doing and inflating benchmark scores....sheesh :)
 
This is a little interesting. I am looking at new mobos atm, and was browsing the XFX site. I found this under the "Features" section:

NVIDIA SLI Certified Components
Look for other components including select NVIDIA GeForce GPUs and system memory that automatically increase system bus speeds when paired with NVIDIA nForce 590 SLI motherboards.
Click to expand...

Link
 
So they used Rivatuner 2.06 that doesn't even support G94 card and gives an incorrect reading that the author of the article mentions. How is this even news? What if I tried to make a hacked AOD work on a C2D Extreme and see it reports incorrect numbers too! Older versions of CPU-Z also reports incorrect info on newer chips.

This article talks about "Linkboost" causing PCI-E frequency being increased and causing the graphics cards to be overclocked. However "Linkboost" is an obsolete Nvidia motherboard technology that Nvidia discontinued it and its not present in newer 590 chipsets and the 680i chipsets according to Nvidia website.
 
TerranUp16 said:
This is a little interesting. I am looking at new mobos atm, and was browsing the XFX site. I found this under the "Features" section:



Link
Click to expand...

Yes, that is the LinkBoost and SLI Memory "features" that seemed to have died off
 
Yeah- well, the interesting part I pulled from there was that it is enabled automatically depending on the gpu/memory installed.
 
Does anyone remember the Abit BX 133-raid mother board? the agp/front side buse maltiplyer had 1/1 bus 66 mhz buse and 2/3 for 100 mhz. the BX chipset didnt support 133 mhz processers. when you used the the 133 bus PIII it added 44 mhz to the agp bus. At that point in time i had the fastest 3dmark2000 score in the world with a hurcules FX5200 and a PIII 800
 
lehmann said:
Does anyone remember the Abit BX 133-raid mother board? the agp/front side buse maltiplyer had 1/1 bus 66 mhz buse and 2/3 for 100 mhz. the BX chipset didnt support 133 mhz processers. when you used the the 133 bus PIII it added 44 mhz to the agp bus. At that point in time i had the fastest 3dmark2000 score in the world with a hurcules FX5200 and a PIII 800
Click to expand...

...and that's why drugs are bad, mmmkay?
 
spine said:
...and that's why drugs are bad, mmmkay?
Click to expand...

I don't know whats worse, bumping a 20 month dead thread to talk about the good ol days, gaining an obscure record (although on the internet that does classify as an accomplishment, just look at xtremesystems forums) about something which is relevant to the topic, but not the good kind of relevant.

I mean hey, I am pretty sure I had the highest windows stable HTT clock speed to be registered on an nforce 4 based socket 939 mobo, (DFI SLi-DR) ( I don't think that required it to be bench stable, but it was kept) and I could of shattered the record for highest stable clock 2x1 GB DDR1 Infineon BE-5 memory (well actually any 2x1GB DDR1), but Id always get 1 error in like 10 passes of memtest, which I am pretty sure was due to the memory controller of the particular A64 I had, but that didn't even matter because the processor I had wasn't really that great, so I couldn't get the dividers to get the memory fast enough and utilize what cpu OC, along with a low end psu, Basically cost me more than I would of wanted to spend to obtain a record I didn't really care about, well past the platforms hay day,

alrite I need to lay off the drugs.
 
You must log in or register to reply here.
Back
Top