TechPowerup - Nvidia Kepler GK104 PCB Drawings and power connector pics

[BallaTheFeared]

I don't really see why it matters if they test lower than optimal resolutions, those should just be ignored if you don't care about them.

"Who buys a 7970 for 1024x768" Is basically the same logic behind, who buys a 7970 for 30 fps on a $1000 screen. Nothing wrong with either stance, I get almost nothing from [H] reviews, others get nothing from TPU reviews, I get nothing from any review that doesn't use at least a 4.8GHz SB with CF/SLI testing.

Thankfully there are a ton of review sites that all doing something differently!

 

[railven]

Can they hot-install driver without the need to reboot like Nvidia?

Just commenting on this, on my HD 5870 I could install drivers withOUT requiring an restart, but last night I had to install my PhysX as a primary (GTX 9800) and it asked me to reboot after I installed the newest drivers.

Is this a function based on the generation of the card? Just curious. (I'm actually glad it told me restart because once I did it told me I had insufficient power which turns out I only connected 1 of the 6pin power, not both. But, on the AMD/ATI, if you forget to plug in both the screen just flashes with a warning/reminding you.)

 

[f1sherman]

Is this a function based on the generation of the card?

TBH dunno bud.
All I know is I do multiple installs/reinstalls of different Forceware versions and don't reboot even once.

 

[maddie]

Yeah, but what moron would buy a 7970 for 1024x768? They need to drop irrelevant resolutions from high end reviews, as they're factored into their fabled "relative performance - all resolutions" chart.

I believe that we have the right to be "morons" with our own money once no one forces you to do the same.

FREEDOM

 

[maddie]

I don't really see why it matters if they test lower than optimal resolutions, those should just be ignored if you don't care about them.

"Who buys a 7970 for 1024x768" Is basically the same logic behind, who buys a 7970 for 30 fps on a $1000 screen. Nothing wrong with either stance, I get almost nothing from [H] reviews, others get nothing from TPU reviews, I get nothing from any review that doesn't use at least a 4.8GHz SB with CF/SLI testing.

Thankfully there are a ton of review sites that all doing something differently!

Ditto.

If only we all felt the same about graphic card choices.

 

[Subyman]

Just commenting on this, on my HD 5870 I could install drivers withOUT requiring an restart, but last night I had to install my PhysX as a primary (GTX 9800) and it asked me to reboot after I installed the newest drivers.

Is this a function based on the generation of the card? Just curious. (I'm actually glad it told me restart because once I did it told me I had insufficient power which turns out I only connected 1 of the 6pin power, not both. But, on the AMD/ATI, if you forget to plug in both the screen just flashes with a warning/reminding you.)

Are you running XP by chance? I believe WIn 7 has the ability to hot-install drivers without reboot, but XP needs to be restarted.

 

[blackened23]

I believe that we have the right to be "morons" with our own money once no one forces you to do the same.

FREEDOM

TPU is so out of touch with reality its ridiculous. To even suggest that someone would pay 450+ for a GPU while spending 25$ on a cheap 20" wal mart TN panel is the height of stupidity, I cannot believe anyone would justify TPUs archaic testing methods.

I used 1280x1024 in the late 90s. Hello, this isn't 1998. Its time for TPU to wake up.

 

[maddie]

TPU is so out of touch with reality its ridiculous. To even suggest that someone would pay 450+ for a GPU while spending 25$ on a cheap 20" wal mart TN panel is the height of stupidity, I cannot believe anyone would justify TPUs archaic testing methods.

I used 1280x1024 in the late 90s. Hello, this isn't 1998. Its time for TPU to wake up.

And that statement, is why these forums often deteriorate into shouting matches. The world is NOT only about you.

All I'm saying is that let other people choose without your attempts to dictate. You lose nothing.

 

[Homeles]

And that statement, is why these forums often deteriorate into shouting matches. The world is NOT only about you.

All I'm saying is that let other people choose without your attempts to dictate. You lose nothing.

Someone's gotta tell someone they're being stupid.

 

[BallaTheFeared]

TPU tests all resolutions, nothing stupid about it.

 

[blackened23]

And that statement, is why these forums often deteriorate into shouting matches. The world is NOT only about you.

All I'm saying is that let other people choose without your attempts to dictate. You lose nothing.

Oh. Sorry. Getting 580 SLI for 850$ and a 20" wal mart TN panel for 25$ makes PERFECT sense. I guess I didn't see it.

 

[BallaTheFeared]

Just because nobody uses doesn't mean there isn't any reason to test it.

It also shows the gains we've seen not only in gpu power, but also cpu power since those days.

But I'm sure TPU cares what you think, you should send them a message though because the probability of them reading your responses here is quite low.

 

[MrK6]

TPU is so out of touch with reality its ridiculous. To even suggest that someone would pay 450+ for a GPU while spending 25$ on a cheap 20" wal mart TN panel is the height of stupidity, I cannot believe anyone would justify TPUs archaic testing methods.

I used 1280x1024 in the late 90s. Hello, this isn't 1998. Its time for TPU to wake up.

I think it's just the idea that TPU seems to waste a lot of effort where most people don't really care. Honestly, if they dropped the 1024x768 tests and instead spent that time ensuring they had the latest drivers installed, I think I and others would place more merit on their reviews.

 

[Tempered81]

They don't wanna be like everyone else. Wiz would rather cover all bases and update his drivers every 3-4 months to show the most data, like Kyle and Brent do their 'gameplay experience' thing. I like both sites, cuz you have about 25 others that do the 'normal' way. However, I agree with MrK6, would be nice if they did it with the newest drivers.

 

[Quantos]

Well, imo testing all resolutions isn't really that big of a deal because they show the overall performance per resolution at the end of the review. It does mean the "performance overall" is diluted by low resolution is pretty useless. It seems like a waste of time to me to test at 1024x768, but hey, it's their time. It just means that you have to look at the resolution you're looking for to get meaningful resulsts.

What bothers me the most, however, is the CPU bound game tested that, at any resolution, give the same results for each GPU.

http://tpucdn.com/reviews/AMD/HD_7970/images/sc2_2560_1600.gif

This kind of data is hurting the review's results overall.

 

[BallaTheFeared]

Meh, their problem is they're still using a 3.8GHz i7-920, and [H] was using a 3.6GHz last gen product. I should clarify why... [H] Was taking a lot of flack for using that processor with SLI setups, because Nvidia was known to require a faster cpu to get higher scaling/usage. They even did a review stating the large gains they saw switching to a 4.8GHz i5-2500k, however the're back on their 3.6GHz 1366 cpu once again for these new card tests.

That it itself invalidates their testing, but a lot of sites are using dated hardware to test top end gpus, even CF/SLI setups with dated processors from 2008 without even 4GHz overclocks.

Also:

Bulldozer is slower than a 2.66GHz processor from 2008! lolz

 

[3DVagabond]

Is there a link for this chart?

 

[notty22]

Is there a link for this chart?

HD 7970: Bulldozer vs. Sandy Bridge vs. Nehalem


http://www.techpowerup.com/reviews/AMD/HD_7970_CPU_Scaling/14.html

 

[Nintendesert]

Meh, their problem is they're still using a 3.8GHz i7-920, and [H] was using a 3.6GHz last gen product. I should clarify why... [H] Was taking a lot of flack for using that processor with SLI setups, because Nvidia was known to require a faster cpu to get higher scaling/usage. They even did a review stating the large gains they saw switching to a 4.8GHz i5-2500k, however the're back on their 3.6GHz 1366 cpu once again for these new card tests.

That it itself invalidates their testing, but a lot of sites are using dated hardware to test top end gpus, even CF/SLI setups with dated processors from 2008 without even 4GHz overclocks.

Also:

Bulldozer is slower than a 2.66GHz processor from 2008! lolz

http://www.techpowerup.com/reviews/AMD/HD_7970_CPU_Scaling/12.html



OH MAN Sandy Bridge gets smoked by Bulldozer in Metro at the highest resolutions!

 

[BallaTheFeared]

Yeah, no reply needed for that one.

 

[3DVagabond]

I'm not sure why we are comparing Bulldozer, but the whole review should be read. There's definitely a difference in performance between the processors, but SCII is bottlenecking all the CPU's at all resolutions, and pretty much a worse case. It's not the same with all games.

IIRC, the real problem with [H]'s X58 test system was the 3rd PCIe slot runs @ x4 and with nVidia they were using 3 cards and with AMD they were using 2 (both running @ x16).

 

[VirtualLarry]

Interesting design for the power connectors. I'm not sure I've ever seen that before. and 6-pin power connectors don't have latches to hold them in place.

WUT? All my cards in recent history that have either one or two 6-pin PCI-E connections, have latches.

 

[VirtualLarry]

You people just don't get it. 400-450$ for a performance part is insane. No Nvidia performance part has been this expensive.
9800GT supported tri-sli. How much was that? Hint: not 400-450.

What? Wasn't the 8800GTX like $600 when it was released?

It was ATI's small-die strategy and cheaper prices with the HD4850/70 that changed all that.

 

[AnandThenMan]

You people just don't get it. 400-450$ for a performance part is insane. No Nvidia performance part has been this expensive. If you could just admit that Tahiti is way overpriced in the grand scheme of the combined 28nm lineup of AMD+Nvidia, we would be actually getting somewhere with this discussion.
What idiot would pay that much money for a GK104 card? The same idiot that would pay 700-800 bucks for a GK110 and 1000 for GK104x2? Rrrrrright.

9800GT supported tri-sli. How much was that? Hint: not 400-450.

So much fail in this post.

 

[boxleitnerb]

So much fail in this post.

How about posting your own arguments instead of this simplistic response? Almost everyone agrees that the prices are bloated right now. I fully expect this to be set right once Kepler arrives.