The GT300/Fermi Thread

[neliz]

That looks more like a camera blur, also it was only 2 connectors.

That much was very easy to see.

it is indeed a reflection.

There's two connectors, true, but there's 14 cables (8+6) going in the 8pin and 6 in the 6pin.

[N19h7m4r3]

That's because both the 6 and 8 pin connectors are on the same cable.

It's one of those odd PSUs where instead of having a set of cables for each connector they have them on the same one, where it splits near the end.

It's basically in multi-rail PSUs where they have one cable for PCI-E for each rail, meaning each card gets its own rail.

[mikeyakame]

Same here Scaniris
I have had to RMA 3 GTX 280 cards
My first one lasted for ~7months before the screen started to get pink dots.
Second one (RMA replacement) was a DOA
Third one lasted for ~9months and then started to run extremely hot (loud fan too) and develop black flashing retangles in DirectX 10 games such as Clearsky, Crysis and (yes it's not a game) 3dmark Vantage.
I have a 285 on the way, but it is delayed because of the snow

Anyway back on topic I agree that stability is key and performance too, but stability more so.

The ATi HD 5970 is not out yet in the UK , perhaps when it is released here the drivers would be much better. The 5870 series stocks are very low here, but they have been out since December.

nVidia really do need to hard launch, and launch well as ATi's launch as been woeful here.

Charlie is semi accurate. The Tesla Fermi does indeed have 448 cores, however Quadro varients have a 448 3GB version coming EARLY Q2 2010 and a 512piper 6GB version coming sometime Q3 2010.

I would assume the same could be said for desktop cards (except perhaps the memory halved) and then perhaps a sneaky refresh of the 448 part @ 480pipes (rather like the GTX 260-216)

As for power consumption, it would not surprise me if it the test/beta cards have 2x 8pin, but the retail cards are 2x 6pin or 1x8pin and 1x6pin.
Remember that this product is still not fit for human consumption, nVidia have 6weeks? or so of tweakage to go.

John

That's pretty horrible luck with your 280s, my 280 I bought a few weeks after release is still going strong overclocked. I guess it's probably around 18-19 months old now. It's always run extremely cool and fan is running on auto mode, so up and down all over the place duty cycle.

[annihilat0r]

All the people giving Fermi performance numbers or estimates, have the clocks been finalized anyway? If so, what are they?

[saaya]

All the people giving Fermi performance numbers or estimates, have the clocks been finalized anyway? If so, what are they?

think about it... why would nvidia finalize them already?
it would be stupid... they are definately still tweaking the design...
they wont settle for final clocks until very few weeks before the launch... and thats still months away...

i heard several rumors though that range from 500-750mhz and shader clocks of 1300-1600
i think the lower numbers were a1 and a2 which had clock problems, i think the final clocks will be about the same as the 280...

[LordEC911]

think about it... why would nvidia finalize them already?
it would be stupid... they are definately still tweaking the design...
they wont settle for final clocks until very few weeks before the launch... and thats still months away...

i heard several rumors though that range from 500-750mhz and shader clocks of 1300-1600
i think the lower numbers were a1 and a2 which had clock problems, i think the final clocks will be about the same as the 280...

More like GTX285. It really depends on what they are able to do with their bins.
I still have doubts that they will have a fully enabled 700mhz part, unless it is a reviewer edition for the first quarter or so.

Edit- Short answer to your question, no clocks are not finalized though they have a good idea where they will land.

[saaya]

well 280 was 600, and 285 was 650 ref clocks...
i think they will be more around 600 than 650...

fully enabled? i thought even their server parts wont have all units enabled?
and they seem to be heat limited... a lot... unless they have perf issues and cant beat an overclocked 5870, which would surprise me a lot, i dont see why they would go at great lengths to clock their cards high and run into a lot of heat and power complications... not really worth it i think...

[LesGrossman]

Am i the only one thinking right now that fermi will be a massive flop?

I mean 5xxx are selling like crazy, nvidia cards not much, why no official performance numbers from nvidia? It would only hurt AMD. The card is up and running depiste the crashes, so why no numbers? The only reason i can thing of, is that it would make the few people still waiting for fermi to stop waiting and go buy a ATI card.

[570091D]

Am i the only one thinking right now that fermi will be a massive flop?

I mean 5xxx are selling like crazy, nvidia cards not much, why no official performance numbers from nvidia? It would only hurt AMD. The card is up and running depiste the crashes, so why no numbers? The only reason i can thing of, is that it would make the few people still waiting for fermi to stop waiting and go buy a ATI card.

you're correct, telling people the performance numbers of an unfinished card would make potential buyers turn to the competition.

[JohnZS]

That's pretty horrible luck with your 280s, my 280 I bought a few weeks after release is still going strong overclocked. I guess it's probably around 18-19 months old now. It's always run extremely cool and fan is running on auto mode, so up and down all over the place duty cycle.

I wish my GTX 280 lasted that long.
My luck got worse. GTX 285 was able to get here...however it is DOA. Doesnt POST. In either slot I get the beep codes for VGA not detected. So back it goes. I hope Fermi is more reliable for me.
John

[eric66]

you're correct, telling people the performance numbers of an unfinished card would make potential buyers turn to the competition.

who cares if its unfinished or not numbers will turn ati buyers to fermi waiters you know that everybody does

[Nedjo]

Did we discussed together before? I don't remember this so Hi, nice to meet you there, I'm Cyril and I'm happy to be there sharing informations and thoughts with you.

since this is public forum it makes no difference did you've discussed with me before or not. Anyhow, nice to meet you Cyril

And before talking about me Fuding there, you should know 3 things:

- I'm working for one of the 4 world major IT manufacturers as ITAC, and
I HATE Fuding because it's opposite of what I'm doing in my daily business, absolutely not compatible with my ethic @ work and in the real life , I'm neither a sales guy nore a marketing guy. I'm there to share, have fun, learn and have objective and pragmatic discussions with others, I'm not there to fight or to fud

congrats on your job! I'm glad we agree on FUDing subjects! Thanks for providing the links

- Did you already had a look into the ATI forum and into the watercooling forum? If not, a short refresh about latest news, you'll see people, me included, trying to find workarounds... with this HD5970

Here: http://www.xtremesystems.org/forums/...=241616&page=2
Here: http://www.xtremesystems.org/forums/...=239276&page=3
Here: http://www.xtremesystems.org/forums/...=240358&page=2
Here: http://forums.amd.com/game/messagevi...&enterthread=y

Another one on the EVGA forum.

I see that you've spread the word of your problems around the Net and forum. Did you actually tried to RMA the card it self, 'cos you know that any piece of any hardware is prone to malfunctioning?!

- Last bnl: I thought I was editing my message, but instead of that, I was posting a new one. So it's not wanted to have a reply only with a smiley.

So please...don't teach me about what Fud is please. And moderate yourself when you talk with people you don't know, you're on a forum yes but behind keyboard there are real human people.

you really need to reconsider your emotional approach to forum debates. It's really nothing personal. I'm sure that you're great French guy! If you read my post agin, you'll see that I've only asked you for the links 'cos w/o them writing those kind of posts does look like FUD-ing, as I'm sure you're aware of that!

Keep some critical distance and try to stay objective with me and others. I don't know you so I hope it's some kind of misunderstanding because of this lonesone smiley.

Of course I don't know you, as well as you don't know me! That's exactly why I didn't write you in any personal manner! Again please read my post again and I'm sure you'll see that.

Hope everything is crystal clear now about my intentions. If you have a look on previous posts in this thread you'll see that I moderate my conclusions after the press news I gave about high temps. I'm not that kind of guy with kid's maturity and negative mindset.

I've never insinuated on your attentions. I've simple asked for the source links so that you'r post have appropriate backing, and not something that I'm sure you haven't intended to have - FUD.

PS: and remember: I'm french and english isn't my first language. It's not alway easy for me to be as precise and fine with English as with French. But I take care of details as far as I can to avoid such situations where language may provoke misunderstandings. Why would I loose my time to make efforts speaking in english if it's for being a troll? It would be easier to do it on french forum right?

any forum member regardless of the origin is my fellow forumer


P.S
please consider RMA-ing your card

[LesGrossman]

you're correct, telling people the performance numbers of an unfinished card would make potential buyers turn to the competition.

wow nice, Nvidia just runned unigine with 3 "unfinished" cards

[Chewbenator]

I've had pretty bad luck with my GTX260. Two RMA's so far with EVGA. I am only looking forward to Fermi so that there's an ATI pricecut. Sad that Nvidia's problems are stopping me from buying an ATI card.

[Chumbucket843]

Am i the only one thinking right now that fermi will be a massive flop?

I mean 5xxx are selling like crazy, nvidia cards not much, why no official performance numbers from nvidia? It would only help AMD. The card is up and running depiste the crashes, so why no numbers? The only reason i can thing of, is that it would make the few people still waiting for fermi to stop waiting and go buy a ATI card.

fixed.

[FischOderAal]

I wish my GTX 280 lasted that long.
My luck got worse. GTX 285 was able to get here...however it is DOA. Doesnt POST. In either slot I get the beep codes for VGA not detected. So back it goes. I hope Fermi is more reliable for me.
John

You sure it's not your mainboard? Imho you should test it in a friend's computer as well. The chances to get that much hardware doa are small, don't you think? Only had one piece of hardware arrive dead.

who cares if its unfinished or not numbers will turn ati buyers to fermi waiters you know that everybody does

Exactly what I'm thinking. Would it be that hard to show a few numbers, even on early silicon? I mean, they could say "it's early hardware so it can improve" and keep ppl from buying HD-5k.

[RPGWiZaRD]

The sooner they reveal the performance the easier it will be for AMD to know exactly how much performance they need to aim for their refresh, that's why you don't usually want to reveal anything until 1-2 months before, that's not enough time for the competitor to make any rough changes but enough time to create some hype before launch.

And what if they showed numbers now but the figures was worse than expected?

[570091D]

wow nice, Nvidia just runned unigine with 3 "unfinished" cards

why would the cards need to be finished to run unigine? no eng samples or test boards? it was running with the final clocks and shader count? c'mon man....

we all know how news is disseminated. most of the people waiting to make a decision are not hawking forums for the latest, they are just waiting for any numbers before they pull the trigger. i am certain that nvidia wants them to see the final results (for better or worse). beyond that, a certian competator of theirs has been waging quite an aggressive pr campaign with the express intent to sully nvidia's name (not that they can't do that themselves). i'm sure they want to release some performance data, but they don't need to give thier rivals or critics any more ammo. the nda comes off next week, after waiting this long a few days more won't kill us.

[Aerwidh]

who cares if its unfinished or not numbers will turn ati buyers to fermi waiters you know that everybody does

Only if the numbers are good enough, disappointing numbers could drive people to their competitor. They probably just want to look their best when they release any numbers.

[saaya]

you're correct, telling people the performance numbers of an unfinished card would make potential buyers turn to the competition.

only if they suck... if they are great it will make people camp on their cash

wasnt that the whole point of nvidias pr hype late last year at GTC?
dont buy atis 5xxx cards, fermi is coming soon and look how great it performs and what you can do with it!

now that they COULD create some REAL hype with actual numbers and performance... they only have a very small show at ces and dont talk about perf at all... hmmmmmmmmm why could that be?

im not saying fermi perf sucks, but its clearly not as amazing and way faster than atis cards as some people think, and as nvidia claimed... otherwise they would def show off some numbers or at least make some bold claims like "up to 50% higher perf" and "we will blow ati out of the sky" etc etc...

[Jowy Atreides]

wow nice, Nvidia just runned unigine with 3 "unfinished" cards

Yes, they did run (or ran, not runned..) uniengine with unfinished cards.
What's the meaning of your statement?

The redundancy is lost on me. (pointing out the obvious if your english is insufficient)

[WeeMaan]

. otherwise they would def show off some numbers or at least make some bold claims like "up to 50% higher perf" and "we will blow ati out of the sky" etc etc...

My guess is that they are saving the best to the "deep dive into fermi" that will come tomorrow.

[JohnZS]

You sure it's not your mainboard? Imho you should test it in a friend's computer as well. The chances to get that much hardware doa are small, don't you think? Only had one piece of hardware arrive dead.

Good point, a 8800GTS (G92) and 7800GTX run fine on the board

It is a dead card as the fan spins, and I get the BIOS beep codes for "no VGA detected".

I have tested using both PCIe x16 2.0 slots and DVi connections.

It's technically only my second DOA as the other 2 cards became defective after time (1st was the notorious pink dots problem, the other developed black flashing black retangles and the Black Screen of Death issue).

Inbetween those 2 was a DOA which posted, but died after 10mins.

If you ask me BFG should have been stricter with the binning on their GT200 cards, I cannot comment for GT200b cards as this one was a DOA and would not even POST.

I guess these are the issues you get with huge dies and cores, meaning that the slightest issue can manifest iteself over time?
Fingers crossed Fermi does not go through this.....but I fear with the high power consumption it may do, as more power = more heat = damage.

I am losing faith in BFG and dare I say it nVidia after the issues I have had with GTX 280 and 285. If only the grass truly was greener on the ATi side....

John

[orangekiwii]

my gtx285 is dying as well :/

its artifacting and the heatsink barely keeps the card running under stress (100c at 100% fan speed anyone?)

It's not doing everything I want either, so it's wait for Fermi or get a 5970 now. Since neither of those are possible atm, I'll wait for Fermi.

[Manicdan]

its artifacting and the heatsink barely keeps the card running under stress (100c at 100% fan speed anyone?)

you shouldnt be having worse temps over time as long as you keep the dust out. the cards shouldnt be using up more heat, and the heatsink and fan are not changing, so in theory the thermal properties are identical from day one. you might want to reset the heatsink and take a vacuum to it.