Sign in with
Sign up | Sign in
Your question

The GF310 is here!!! - Page 3

Last response: in Graphics & Displays
Share
a c 130 U Graphics card
December 1, 2009 8:36:03 PM

one-shot said:
I don't want to get into an argument with you. Yes, Randomizer said it first as a joke. That doesn't mean you still couldn't make a joke off of that. I assumed you would have picked up on that....

I'll list the specs for you:

GF310
CUDA Cores : 16
GPU Clock : 589MHz
Core Clock : 1402MHz
Memory Clock: 500MHz
Capacity : 512MB DDR2
Memory Interface Width : 64bit

GF210
CUDA Cores : 16
GPU Clock : 589MHz
Core Clock : 1402MHz
Memory Clock: 500MHz
Capacity : 512MB DDR2
Memory Interface Width : 64bit

You can clearly see that both cards have the exact same specs. Both cards are even DX 10.1.

Sources:

http://www.nvidia.com/object/product_geforce_310_us.htm...

http://www.nvidia.com/object/product_geforce_210_us.htm...

You also stated:

"Its not just a name change. Try looking the card up before you start putting it down. Its the very bottom of the pile OEM only so far as well, and is as i said earlier designed with the HTPC in mind. Sure its not a gaming card but for what its made for it does very nicely thank you."

I wasn't putting the card down. I made a comment on the similarities and my inability to tell the difference between two exactly similar cards with different names.

If it's not just a name change, again, please tell me. We know it's at the very bottom of the OEM. We know it's designed for HTPC. We KNOW it does what it's made to do. You STILL have not answered what is different about the two cards besides a name change. Instead, you ignore the question and state obvious things we all know. If you would just tell us what the difference is, I'd be happy to let this rest. I don't want to miss out on anything if indeed you DO know something we don't already know.


Firstly im sorry but needed to make a point here.

Thank you with your last paragraph you have just made my point for me.
JD has done just about every thing you listed and yet I'm the only one pulling him up on the fact that none of what he is stating about Fermi is actually known.
We know everything he has posted about drivers and refreshes and how the 4890 performed etc. I asked for links regarding claims that Fermi wont shrink well more than once and got flanneled off with talk of a forum post that "if you know who is saying what will make some sort of sense"
Im quite happy to tell you that i have no idea whats actually differant about the "newer" card if indeed anything. There may or may not be newer instruction sets i really dont know.

Now your turn JD, Do you actually know anything usefull about Fermi or not ? As i said before you make a good argument but its usual etequete on these forums during such a discussion when making specific claims to post suportive links, especially when asked for. Had you said watch this guys posts in this forum then fair enough but to just say work out who's saying what for yourself isnt very helpful.

I just proved that claims are expected to be suported. Sorry again one shot :( 
I didnt like doing that but you just couldnt or wouldnt see what the issue was even after i spelled it out for you. I wasnt just posting for myself. You need to remember not everyone has your level of knowlegde of contacts/who's who and whats what.

Mactronix
m
0
l
a b U Graphics card
December 1, 2009 10:10:11 PM

Just something that may be of interest to those wanting more info on the Fermi design, this is the latest, and likely best summation of the info available to date, even mentioning a little obout textures and how tesselation will be handled in 'software' mode;

http://techreport.com/articles.x/17815

PS, for those of you who don't know, Rys does the reviews for Beyond3D.
m
0
l
December 1, 2009 10:54:41 PM

mactronix said:
Firstly im sorry but needed to make a point here.

Thank you with your last paragraph you have just made my point for me.
JD has done just about every thing you listed and yet I'm the only one pulling him up on the fact that none of what he is stating about Fermi is actually known.
We know everything he has posted about drivers and refreshes and how the 4890 performed etc. I asked for links regarding claims that Fermi wont shrink well more than once and got flanneled off with talk of a forum post that "if you know who is saying what will make some sort of sense"
Im quite happy to tell you that i have no idea whats actually differant about the "newer" card if indeed anything. There may or may not be newer instruction sets i really dont know.

Now your turn JD, Do you actually know anything usefull about Fermi or not ? As i said before you make a good argument but its usual etequete on these forums during such a discussion when making specific claims to post suportive links, especially when asked for. Had you said watch this guys posts in this forum then fair enough but to just say work out who's saying what for yourself isnt very helpful.

I just proved that claims are expected to be suported. Sorry again one shot :( 
I didnt like doing that but you just couldnt or wouldnt see what the issue was even after i spelled it out for you. I wasnt just posting for myself. You need to remember not everyone has your level of knowlegde of contacts/who's who and whats what.

Mactronix


I took what you said personal, rather than the way you meant it, Mac. I was hoping there was actually something different so it would make nVidia seem like they didn't just rebrand another product. I would have been happy to be wrong on that. :( 
m
0
l
Related resources
a b U Graphics card
December 2, 2009 12:40:18 AM

What Im saying mac, its being debated now, as weve seen nothing from nVidia that shows it can be done.
Thats why to see if the breakdown is efficient enough to actually make a difference at smaller designs. It doesnt appear that the 200s are doing that, and it appears the 300 from what we do know is a slight divergence from the 200 series, but not in the G92 area at all.
They gave up alot of their compute density with G200, which even the G92 cant match compared to the 4xxx series.
Going monolithic/large has been a problem for nVidia, and as I said, it still doesnt appear to have changed that much for smaller designs.
Its still up in the air, but the infos there, If clocks can be ramped, this will all change, provided shaders and core clocks remain balanced at higher clocks, but again, thats an unknown, and something to keep an eye on.
The cut down slightly smaller G200s did well, allowing for higher clocks, but the 280 was dead to ocing in what wed seen compared to the G92, and since, the clocks have actually gotten slower. Again, we know the smaller 200s dont offer anything over the G92s, and if their clocks arent achieveable either, theres simply not alot to work with.
This couldchange, but so far, weve seen nothing to change this pattern.
The design appears to be more robust on the 300, but beyond that, will this also transcend down to the lower levels, if clocks continue to be a burden?
Or, by cutting it down, does it lose it robustness, even if clocks are good?
Either way, what were seeing lately is, something needs to change, and it may be more of the same, and thats whats been in question.
So far, it hasnt been shot down, and thats up to nVidia to surprise
m
0
l
a b U Graphics card
December 2, 2009 12:57:26 AM

Nice read there TGGA. No if only NVidia would get the dumb thing out. I really would like to see what hit software Tesselation does.
m
0
l
December 2, 2009 1:10:02 AM

http://techreport.com/articles.x/17815/4
http://techreport.com/articles.x/17986/2

take a look at both tables of comparisson

taken the numbers from both and merging them I get this

| 295 | GF100 | 5870
-----------------------------------------
GTexels |92.2 | 83.2 | 68.0
GPixels |32.3 | 31.2 | 27.2
Memory GB/s|223.9| 201.6 | 153.6


It seems Fermi is going to be slower than 295. And that is if they hit 650/1700 Mhz.
Boy they will be in trouble if they hit only 600/1500.

But even if they hit their mark and looking at current numbers 295 is about 10% faster than 5870 (or even less), and taking into consideration of few more months of maturing drivers for 5870, if Fermi is slower than 295 then it will be total disapointment as it will not be able to beat 5870 at all, may be at most equal = failure

I am rooting for nVidia at the moment as I dont want to see another iteration of monopoly leaving it all to ATI while we wait next nVidia/Larabee cards to try to compete.

m
0
l
a c 130 U Graphics card
December 2, 2009 10:26:12 AM

EXT64 said:
Nice read there TGGA. No if only NVidia would get the dumb thing out. I really would like to see what hit software Tesselation does.



+1 Very good article

Mactronix
m
0
l
a b U Graphics card
December 2, 2009 5:13:24 PM

WooHoo!!

The REAL Fermi card, all they needed to do was to add a 5 !!

Now for some Performance #s ! :bounce: 

m
0
l
December 2, 2009 5:18:16 PM

a real fremi card in a
HP Pro 2000 Business PC?
esp when looking at the other cards
Intel Graphics Media Accelerator X4500
ATI Radeon HD 4350 (512MB) FH PCIe x16 Card
ATI Radeon HD 4550 (256MB) DH PCIe x16 Graphics Card (available through January 2010)
ATI Radeon HD 4550 (512MB) DH PCIe x16 Graphics Card (available after January 2010)
ATI Radeon HD 4650 1GB Dual Head PCIe x16 Card
NVIDIA GeForce G210 512MB ATX PCIe x16
NVIDIA GeForce GT230 1.5GB PCIe x16 Card
NVIDIA GeForce 205 512MB ATX PCIe x16

some of them have G's some of them have a GT some of them have no G or GT and one of them is a GT230 with 1.5gb of ram!

I really don't get nvidia's naming scheme anymore, too confusing; you have to be smart to make a naming scheme so dumb.

http://www.nvidia.com/object/product_geforce_gt_240_us....
under specs
"1700 MHz GDDR5, 1000MHz GDDR3, 900MHz DDR3" wait what when did that happen haha
m
0
l
a b U Graphics card
December 2, 2009 5:20:47 PM

IzzyCraft said:
a real fremi card in a
HP Pro 2000 Business PC?


Sure it's supposed to be mainly a cGPU card anyways, right ?!? ;) 

Go GO Gadget Workstation !!
m
0
l
a b U Graphics card
December 2, 2009 8:07:03 PM

A little poking around during lunch and I found the specs for the G315 (mistakenly put as GT315) looks like a rebranded GT216 (interesting they included the gen # 200 and the spin A2) ;

http://h18000.www1.hp.com/products/quickspecs/13490_div...

Bottom of Page 22

NVIDIA GeForce 315
1GB PCIe x16 Card
Form Factor ATX Full Height
Display max resolution Analog VGA: 2048x1536x32bpp@85Hz
DVI: 2560 x 1600 x32bpp @60Hz
Input/Output connectors DVI-I +VGA
Board configuration Specification Description
Graphics Chip D10M2-20 (GT216-200-A2)
Core clock 625 MHz
Memory clock 790 MHz
Frame buffer 1GB DDR3, 128-bit

Core power 52 W (Max board power)
NVIDIA GeForce 315 (1GB) PCIe x16 Graphics Controller display resolutions and refresh rates
NOTE: Other resolutions may be available but are not recommended as they may not have been tested and qualified by HP.
[/b]
m
0
l
December 2, 2009 8:26:10 PM

On that HD 5970 pic in the second tech report link I'm just assuming that the two identical chips are the GPU processors but what is the chip between them?
m
0
l
a b U Graphics card
December 2, 2009 8:29:06 PM

yeah , what is the chip between those cypress dies ?
m
0
l
a b U Graphics card
December 2, 2009 8:53:47 PM

rawsteel said:
http://techreport.com/articles.x/17815/4
http://techreport.com/articles.x/17986/2

take a look at both tables of comparisson

taken the numbers from both and merging them I get this

| 295 | GF100 | 5870
-----------------------------------------
GTexels |92.2 | 83.2 | 68.0
GPixels |32.3 | 31.2 | 27.2
Memory GB/s|223.9| 201.6 | 153.6


It seems Fermi is going to be slower than 295. And that is if they hit 650/1700 Mhz.
Boy they will be in trouble if they hit only 600/1500.

But even if they hit their mark and looking at current numbers 295 is about 10% faster than 5870 (or even less), and taking into consideration of few more months of maturing drivers for 5870, if Fermi is slower than 295 then it will be total disapointment as it will not be able to beat 5870 at all, may be at most equal = failure

I am rooting for nVidia at the moment as I dont want to see another iteration of monopoly leaving it all to ATI while we wait next nVidia/Larabee cards to try to compete.


while the 295 has good theoretical numbers , in practice it suffers from scaling issues and depends upon cpu too much (i7 beats q9550 by a countable margin) . i am hoping to see even more driver optimizations for fermi because its rather new architecture . hoping because i dont know how this architecure will exactly affect graphics performance in shaders and stuff . cypress is more similar to rv770 than the fermi is to gt200 .

a lot remains to be seen from gf100 after the gf100 itself is seen in stores .
m
0
l
December 2, 2009 9:15:36 PM

cyberkuberiah said:
yeah , what is the chip between those cypress dies ?

It's the PLX PCIe bridge chip

Reading the article help :D 

"The 5970's dimensions and peak power draw are so ample because the card has to accommodate two copies of what is currently the fastest GPU on the market. This really is "CrossFire on a stick," as we like to say, and the performance potential from such a beast is naturally quite considerable. Situated between the two GPUs in the picture above, under a metal cap, is a PCI Express switch chip from PLX, the same model used in the Radeon HD 4870 X2. This chip can support a trio of PCIe x16 links: one to each of the GPUs and a third to the PCIe x16 slot in the host system."

I know reading long papers suck I perfer to hold a book or mag in my hands then just stare at a screen so i use a text to speech program for long reads. It's always fun for GPU to sounds like GaPoo
m
0
l
a b U Graphics card
December 3, 2009 5:14:37 AM

thanks , and nice humor !
m
0
l
December 10, 2009 6:43:15 AM

Based on records:
GeForce 8800GT = GeForce 9800GT = GeForce 9800GTX+ = GeForce GT 250
& Gt210= Gt 310.
From the above
I'm not surprised that the nvidiots are going sell a revison of the Gt 200 series as the Gt 300 series.. Damn u nividiots... we want new chips!!
m
0
l
a b U Graphics card
February 4, 2010 8:26:09 PM

Looks like the 5450 just killed it.
m
0
l
!