NVIDIA GeForce 4XX Series Discussion

[Zubasa]

What I also want to mention is, this thread is too centered around the "Fermi" for my likeling.
This is the GT30XX thread, and I believe is should be about the whole generation of cards.

Although GTX380/370/360 "Fermi" is the Leviathan that gets everyone excited, the mid-range cards are what really matters.
The dual-gpu behemoths are just a demonstration of power so to speak.
What really matters is the "GTS 350" and "GTS 340" this is by far the "meat" of the generation.
I doubt that 90% of gamers will shell out for the "extreme" high-end that the GTX380/GTX 390 is lying.
Despite many of us hate the G92 because of nVidia's naming scheme, it is by far the most sucessful graphics chip in recent years. This time they fail to deliever that "G92 of DX11" in time.

So at a stragetic point of view, nVidia already lose half of the battle.
They fail to deliever not only their high-end, but also their mid-range products before thanks giving and X'mas. We should all understand this is a great mistake in business.
If nVidia manages to release their "GTS 350" and ship them to this side of the world before Chinese New Year, they might be able to make up for what they lost in X'mas.

 

[Binge]

The GT200 thread was about the GTX 260/280/295. This thread should also be concerned with the high performance end of the 300 series. I mean we are enthusiasts here more than market analysts/critics.

 

[Zubasa]

The GT200 thread was about the GTX 260/280/295. This thread should also be concerned with the high performance end of the 300 series. I mean we are enthusiasts here more than market analysts/critics.

If the GTX 360 is going to be as powerful as the HD5870 as many believes, the GTS 350 should compete some where to the 5850 isn't it?
If so isn't a GTS 350 tri-SLi be more intersting to see for us enthusiasts?

The GT2XX thread was about the GTX series mainly because the GTS series did not exist until almost half a generation later.
If you wnat to mention the last generation, there is a good amount of users that own GTS 250 or GTX 260 and SLI setups.

 

[wolf]

lol, most of the troop is here

Currently Active Users Viewing This Thread: 8 (5 members and 3 guests)

wolf, Benetanegia, Binge, Bo_Fox, Zubasa

Awesome read guys keep it up.

 

[Binge]

If the GTX 360 is going to be as powerful as the HD5870 as many believes, the GTS 350 should complete some where to the 5850 isn't it?
If so isn't a GTS 350 tri-SLi be more instersting to see for us enthusiasts?

The GT2XX thread was about the GTX series mainly because the GTS series did not exist until almost half a generation later.

And NVidia has already released information that the first retail cards available will be the high performance cards. I don't think, I know, the whole span of Fermi architecture cards will be the most interesting to enthusiasts. If the GTS are not Fermi it means they are rebrands and are old news.

 

[Lionheart]

You do notice that pic of what we assume to be the GTX 380 has 8+6pin connectors?

Assuming that pic is real:
If they are going to make a dual GPU card out of that, they will most likely need @lease a dual 8-pin card with under clocked/cut out GPUs.
The Fermi might have higher transistor density, but that couldn't save them from the higher TDP from the higher power consumption.
It just sounds all too much like the last gen, I am sorry to say.
Just that this time they are a bit too late on the move.

I do now, thanks for clearing that up for me, seems like these are gonna be power hungry cards, lets hope not but will just wait and see, thats what we can only do for now!: play the waiting game!

 

[Zubasa]

And NVidia has already released information that the first retail cards available will be the high performance cards. I don't think, I know, the whole span of Fermi architecture cards will be the most interesting to enthusiasts. If the GTS are not Fermi it means they are rebrands and are old news.

Well hopefully the GTS 3XX should be fermi based, or else it is really a shame on nVidia.
What I mean by "G92 of DX11" is a chip that is as sucessful as the G92 is in the DX10 era.

 

[Binge]

Well hopefully the GTS 3XX should be fermi based, or else it is really a shame on nVidia.
What I mean by "G92 of DX11" is a chip that is as sucessful as the G92 is in the DX10 era.

We caught that Zubz, but I remember you saying....

What I also want to mention is, this thread is too centered around the "Fermi" for my likeling.
This is the GT30XX thread, and I believe is should be about the whole generation of cards.

If you're going to change the subject than I suggest you at least say something constructive or insightful. Saying shame on NV over and over only solidifies our understanding of your opinions of their business. Trust me when I say it's not so complex to warrant the redundancy.

 

[Bo_Fox]

What I also want to mention is, this thread is too centered around the "Fermi" for my likeling.
This is the GT30XX thread, and I believe is should be about the whole generation of cards.

Although GTX380/370/360 "Fermi" is the Leviathan that gets everyone excited, the mid-range cards are what really matters.
The dual-gpu behemoths are just a demonstration of power so to speak.
What really matters is the "GTS 350" and "GTS 340" this is by far the "meat" of the generation.
I doubt that 90% of gamers will shell out for the "extreme" high-end that the GTX380/GTX 390 is lying.
Despite many of us hate the G92 because of nVidia's naming scheme, it is by far the most sucessful graphics chip in recent years. This time they fail to deliever that "G92 of DX11" in time.

So at a stragetic point of view, nVidia already lose half of the battle.
They fail to deliever not only their high-end, but also their mid-range products before thanks giving and X'mas. We should all understand this is a great mistake in business.
If nVidia manages to release their "GTS 350" and ship them to this side of the world before Chinese New Year, they might be able to make up for what they lost in X'mas.

Well, the high-end part is usually the exciting part for many. You're right that the mainstream part will be selling more than the high-end part, but I think that more of us here are excited about the Fermi. In the links in my sig, more people voted on the high-end parts than on the mid-range ones.

I certainly do hope that Nvidia delivers its promise on ensuring the reliability of 40nm chips. Remember the 65nm fiasco? Pretty much half of the 8600 chips ended up dead so far.

 

[Benetanegia]

Thanks for the original viewpoint.

For anyone confused he said that NV will have a larger chip that will most likely produce a good bit of heat, and requires a good deal of power to operate. He feels this step is much like the GT200, and the "move" to GT300 is too slow. Zubz also found it fitting to bring up the validity of recent pictures of the GT300. I truly doubt anyone's brought up the subject before him

@Zubasa

The 8+6 pin connectors are meaningless. First of all they were seen in an engineering sample and secondly, the Tesla card with it's 3-6 GB of ram is rated at 186w TDP. The GPU will not be rated much higher.

 

[Zubasa]

@Zubasa

The 8+6 pin connectors are meaningless. First of all they were seen in an engineering sample and secondly, the Tesla card with it's 3-6 GB of ram is rated at 186w TDP. The GPU will not be rated much higher.

Afterall, it is a Tesla.
The workstation cards are often are clock differently and have different TDP to their Geforce counterparts.

 

[Benetanegia]

Afterall, it is a Tesla.
The workstation cards are often are clock differently and have different TDP to their Geforce counterparts.

Yes but it won't change much. Especially when the Tesla has much more memory and memory consumes its fair bit of power. Also as long as the TDP is just one watt higher than 225w 8+6 pins are almost guaranteed. 8pin+pcie slot = 150 + 75 = 225 OR 6+6pin + pcie slot = 225w ->> even if it consumes 226w you may want to put an extra pin to be safe and ensure good overclockability. At 226w the single GPU card you can still make a dual card, and yes it will be crippled in one way or another, just like the GTX295 was, and not like we should care TBH, 96 ROPs is just stupid to have on a graphics card...

EDIT: Think about the HD5750. it doesn't really need a power connector, but it does have one for the same purpose.

 

[wahdangun]

Yes but it won't change much. Especially when the Tesla has much more memory and memory consumes its fair bit of power. Also as long as the TDP is just one watt higher than 225w 8+6 pins are almost guaranteed. 8pin+pcie slot = 150 + 75 = 225 OR 6+6pin + pcie slot = 225w ->> even if it consumes 226w you may want to put an extra pin to be safe and ensure good overclockability. At 226w the single GPU card you can still make a dual card, and yes it will be crippled in one way or another, just like the GTX295 was, and not like we should care TBH, 96 ROPs is just stupid to have on a graphics card...

EDIT: Think about the HD5750. it doesn't really need a power connector, but it does have one for the same purpose.

but why nvdia didn't use 6+6 pin ? if the power consumption not that big,

it's sound like GTX 280 again but a bit late, plus 8+6 pin it's like gonna be a power hog(but maybe OC like crazy)

 

[Binge]

but why nvdia didn't use 6+6 pin ? if the power consumption not that big,

it's sound like GTX 280 again but a bit late, plus 8+6 pin it's like gonna be a power hog(but maybe OC like crazy)

We have another original guy with original observations

 

[Bo_Fox]

but why nvdia didn't use 6+6 pin ? if the power consumption not that big,

it's sound like GTX 280 again but a bit late, plus 8+6 pin it's like gonna be a power hog(but maybe OC like crazy)

Hey, I think Binge is probably just giving you a hard time for humorous intents, so just take him with a grain of salt.

This is a good observation, in that it's the first NV GPU to use 8+6 pin, so it's probably going to eat some amount more power than a GTX 280/285 at the very least. Perhaps you are right that it is to leave some room for factory overclocked cards that use a little more voltage. The 8800 Ultra and GTX 280/285 did eat a good amount of power at maximum load, so anything more than that is going to be something worth considering. Not everybody has PSU's that support 8+6 pin rails.

 

[Benetanegia]

but why nvdia didn't use 6+6 pin ? if the power consumption not that big,

it's sound like GTX 280 again but a bit late, plus 8+6 pin it's like gonna be a power hog(but maybe OC like crazy)

Because as I said 6+6 only givess enough power for 225w and they may want to have some headroom. The fact that the engineering sample has 8+6 doesn't mean the final card will have 8+6 anyway.

Anyway I have heard an explanation, which I don't know if it is true or not, but makes a lot of sense. 8+6 pins were put for the HPC market, so that they can choose between putting a 8 pin or 2x6-pin. That's very beneficial in the HPC market, because some servers are made of different aged racks often times and will have a very varied array of connectors.

 

[{JNT}Raptor]

This is a good observation, in that it's the first NV GPU to use 8+6 pin,

9800 GX2 used a 8+6 pin long before this.....Sweet card as well....I loved mine.

Before long your gonna see 8+8 or how about triple 6's.....mad power.

 

[Bo_Fox]

Because as I said 6+6 only givess enough power for 225w and they may want to have some headroom. The fact that the engineering sample has 8+6 doesn't mean the final card will have 8+6 anyway.

Anyway I have heard an explanation, which I don't know if it is true or not, but makes a lot of sense. 8+6 pins were put for the HPC market, so that they can choose between putting a 8 pin or 2x6-pin. That's very beneficial in the HPC market, because some servers are made of different aged racks often times and will have a very varied array of connectors.

Hopefully, Nvidia will allow GT300 cards to be powered on by 6+6 pins, instead of forcing 8+6 connections to be used.

9800 GX2 used a 8+6 pin long before this.....Sweet card as well....I loved mine.

Before long your gonna see 8+8 or how about triple 6's.....mad power.

LOL, I was talking about single-GPU cards.

The reason why I think the Fermi GTX 380 will eat more power than a GTX 280/285 is that with more than 2x the shaders unit, running at a higher speed especially (1700 MHz for more than 2x the GFLOPs performance), I'd expect it to need some more voltage.

A 5870 was just double the 4890 in everything, with identical stock speeds, so it was able to run at a somewhat lower voltage in order to remain within the same power envelope of a 4890. Of course, that is with the assumption that the 40nm process was refined enough for more efficient power usage.

Perhaps NV has optimized the 40nm design further than ATI has, but it's more complex with more trannies, and also more densely packed. Well, I do not know for sure, so we'll just have to wait and see. If it's gonna be only 226 watts, then I dont care! I have a PSU that supports 2x 8-pin's anyways.

 

[Lionheart]

Some ppl are saying that the GTX380's will be the same performance as the HD5970 or even outperform it. If so I could see myself getting 2 GTX380's and SLI them just for some stupid crazy pointless performance now that would be awesome!

 

[phanbuey]

Some ppl are saying that the GTX380's will be the same performance as the HD5970 or even outperform it. If so I could see myself getting 2 GTX380's and SLI them just for some stupid crazy pointless performance now that would be awesome!

its never pointless... you just keep cranking up details and AA and AF until you have the performance of a single card but STUNNING visuals.

make half life 2 look better than crysis lol.

or make gta IV ... oh wait... bad example.

 

[Bo_Fox]

make half life 2 look better than crysis lol.

or make gta IV ... oh wait... bad example.

 

[subhendu]

Heard that fermi has DDR5 memory...i have a 4 GB ddr2 800 mem .....i have doubt that it will reduce the performance of fermi

 

[Binge]

Heard that fermi has DDR5 memory...i have a 4 GB ddr2 800 mem .....i have doubt that it will reduce the performance of fermi

The relationship between CPU/main memory, and GPU/main memory is very different. If you end up with a bottleneck it would be at the CPU before it would be at the main memory.

 

[Bo_Fox]

Heard that fermi has DDR5 memory...i have a 4 GB ddr2 800 mem .....i have doubt that it will reduce the performance of fermi

The relationship between CPU/main memory, and GPU/main memory is very different. If you end up with a bottleneck it would be at the CPU before it would be at the main memory.

I think what he means is: Would his mainboard memory reduce the performance of the Fermi?

Nah, probably not, unless your CPU is clocked lower than say, 3GHz and only a dual-core processor. Some games like Far Cry 2 would be bottlenecked by the CPU at 30 fps or less if you have an Athlon X2 5500 or a Celeron for example, no matter how powerful your GPU is. Also, if your DDR2-800 is CAS 6 or higher, then that could also hurt performance. But if you're using an overclocked 3.6GHz Core 2 Quad with DDR2-800 at CAS 4 latency, then it would not reduce the performance of Fermi for most games.

 

[wahdangun]

but some benchies have said even OC Core i7 was bottlenecked HD 5970, we need more demanding game to use this fermi (or better CPU)