Curse

Lord Sojar · Oct 02, 2009, 06:43 PM // 18:43

OpenCL + Windows7 + your PC = screaming performance. DirectX Compute = godly too. G300 won't just make games run at their highest settings, but also Windows7 and many many related apps.

Improvavel · Oct 03, 2009, 02:52 PM // 14:52

Quote:

Originally Posted by Rahja the Thief

OpenCL + Windows7 + your PC = screaming performance. DirectX Compute = godly too. G300 won't just make games run at their highest settings, but also Windows7 and many many related apps.

Isn't the RV870 capable of that as well?

KZaske · Oct 03, 2009, 09:26 PM // 21:26

Quote:

Originally Posted by Rahja the Thief

OpenCL + Windows7 + your PC = screaming performance. DirectX Compute = godly too. G300 won't just make games run at their highest settings, but also Windows7 and many many related apps.

Sounds like I will be staying with nVidia when I replace this computer next summer.

Lord Sojar · Oct 04, 2009, 05:16 PM // 17:16

Quote:

Originally Posted by Improvavel

Isn't the RV870 capable of that as well?

Not to the levels that GT300 (and G300) are capable of. You are probably looking at a 1000%+ performance difference due to the nature of the GT300 chip and its hardware bound technologies versus the software driven parts of ATi's RV870. Mind you, that number is pulled out of my ass, but I am trying to suggest the difference will be very noticeable.

Also remember, GT300 can natively run C++, and CUDA will see a much larger section of the market lean towards its use with Nexus.

If you are a programmer/software junkie, you might find this bit of info interesting to say the least (though you won't be able to access the really juicy developer logs of Nexus and its abilities, you will find some very cool stuff here)
http://developer.nvidia.com/object/nexus.html

Enjoy.

Improvavel · Oct 07, 2009, 01:56 AM // 01:56

Quote:

Originally Posted by Rahja the Thief

Also remember, GT300 can natively run C++, and CUDA will see a much larger section of the market lean towards its use with Nexus.

"Natively".

That would be interesting to see. It has a new unified to access memory (from what I get from the white paper). So apparently it can use OOP language. It still needs to be compiled to Femri, though.

From Femri white paper:

Quote:

Fermi is the first architecture to support the new Parallel Thread eXecution (PTX) 2.0 instruction set. PTX is a low level virtual machine and ISA designed to support the operations of a parallel thread processor. At program install time, PTX instructions are translated to machine instructions by the GPU driver.

Quote:

Fermi and the PTX 2.0 ISA implement a unified address space that unifies the three separate address spaces (thread private local, block shared, and global) for load and store operations. In PTX 1.0, load/store instructions were specific to one of the three address spaces; programs could load or store values in a specific target address space known at compile time. It was difficult to fully implement C and C++ pointers since a pointer’s target address space may not be known at compile time, and may only be determined dynamically at run time.

With PTX 2.0, a unified address space unifies all three address spaces into a single, continuous address space. A single set of unified load/store instructions operate on this address space, augmenting the three separate sets of load/store instructions for local, shared and global. The 40-bit unified address space supports a Terabyte of addressable memory, and the load/store ISA supports 64-bit addressing for future growth.

Quote:

Conclusion

The importance of data locality is recognized through Fermi’s two level cache hierarchy and its combined load/store memory path. Double precision performance is elevated to supercomputing levels, while atomic operations execute up to twenty times faster. Lastly, Fermi’s comprehensive ECC support strongly demonstrates our commitment to the high-performance computing market.

On the software side, the architecture brings forward support for C++, the world’s most ubiquitous object-orientated programming language, and Nexus, the world’s first integrated development environment designed for massively parallel GPU computing applications.

moriz · Oct 07, 2009, 02:08 AM // 02:08

sounds like something my dad would be interested in. he does ocean current modelling, and currently rely on expensive mainframes/computing clusters for his simulations, which also takes a lot of time. a single $2000 fermi tegra might potentially reduce the computational time significantly while cutting back on costs.

Improvavel · Oct 07, 2009, 02:17 AM // 02:17

http://www.xbitlabs.com/news/video/d...cs_Cards.html#

Quote:

Nvidia Corp. has confirmed in a brief media interview that it will be able to cut-down the Fermi-G300 graphics processors in order to address certain specific markets and price-points. The move is natural for all graphics chips designers, but this time Nvidia openly admits that many of implemented capabilities will hardly offer benefits for the consumer right away.

Quote:

“We're not talking about other (chips) at this point in time but you can imagine that we can scale this part by having fewer than the 512 cores and by having these cores have fewer of the features, for example less double-precision,” said Mr. Dally, who did not explain how it is possible to reduce double-precision floating point performance without decreasing single-precision point speed, something which is needed by video games. In fact, Mr. Dally’s comment may imply that non-flagship Fermi derivatives will have not only be slower in terms of performance, but will be seriously different in terms of implementation.

Now, this, if true, seems much more realistic.

Lord Sojar · Oct 07, 2009, 06:07 AM // 06:07

Yes, Fermi (GT300) is a Tesla unit. G300 will be the graphics derivative, but will still have many of the cGPU features needed for OpenCL and DirectX Compute. Double precision isn't needed for much in relation to gaming (minus a few new forms of AA and AF that haven't seen major markets yet), and ECC isn't needed for GPU driven work in a graphics environment.

Fermi's scaling allows us to be much more mobile in marketplace segments than we were with GT200. You will see a much larger and more robust lineup of GPUs spanning many customer segments and price points, which is a great thing for the consumer in the end in regards to performance per dollar.

Fermi is what GT200 should have been, but we needed to see what a very large chip design could do in real world environments first. GT200 was successful where it needed to be, and G300 will fill in the areas of GT200's weakness.

Improvavel · Oct 07, 2009, 04:54 PM // 16:54

Quote:

Originally Posted by Rahja the Thief

Fermi's scaling allows us to be much more mobile in marketplace segments than we were with GT200. You will see a much larger and more robust lineup of GPUs spanning many customer segments and price points, which is a great thing for the consumer in the end in regards to performance per dollar.

I hope so. I don't like to have to pay more than €250 for a graphics card

Gotta love the Ti4200 and the ATI 9500 Pro that would unlock to 9700 Pro

And before the 4850, the last good midrange card was the 6600 GT (although the 7600 GT was quite decent too). I was starting to miss those times.

KZaske · Oct 08, 2009, 06:21 AM // 06:21

Quote:

Originally Posted by Improvavel

I hope so. I don't like to have to pay more than €250 for a graphics card

Gotta love the Ti4200 and the ATI 9500 Pro that would unlock to 9700 Pro

And before the 4850, the last good midrange card was the 6600 GT (although the 7600 GT was quite decent too). I was starting to miss those times.

I really miss the drivers (pre 100). Small, clean and they actually worked well. These new drivers have too many dependancies and are a prime example of broken bloatware.
Why can't they break the driver down into chunks allowing you to install only the parts you want?

Zomgyogi · Oct 11, 2009, 06:43 PM // 18:43

Quote:

Originally Posted by Rahja the Thief

Though, the HD5850 is a great card for the money, hands down. An excellent buy, especially if you plan to run 2 in Crossfire! Big kudos to AMD for the 5850, and a golf clap for the 5870.

So buying a 5850 is a good buy over nVidia at the price point?

Also how many "gaming" DX11 cards will nVidia be releasing?

Ec]-[oMaN · Oct 12, 2009, 05:19 AM // 05:19

Quote:

Originally Posted by Zomgyogi

So buying a 5850 is a good buy over nVidia at the price point?

Also how many "gaming" DX11 cards will nVidia be releasing?

Based on previous releases I'm skeptical Nvidia will or can offer a part that out performs anything ATI is offering or will currently offer for the next 6 months below the 300$ mark.

5850 doesn't seem like a bad choice at all. You can also buy the 5850 now where's we don't know when we can even get our hands on a sub 300$ dx11 Nvidia part. If you are just a tad up to date with upcoming releases you always sell the 5850 just before the competition releases something assuming it's comparable or better at that price point.

Zomgyogi · Oct 13, 2009, 08:43 AM // 08:43

Well I already bought one and they are beasts! They run very cool, but the fan is still pretty loud above 45%. It running load temps about 5c higher then my 4890 idle temps.

Highest so far...850/1200

moriz · Oct 13, 2009, 12:27 PM // 12:27

wow, very impressive overclock there. that should give you almost 5870 performance except in shader heavy instances.

Elder III · Oct 13, 2009, 03:29 PM // 15:29

nice OC and nice temps my stock cooled 4850 often hits 80 degrees, and is about half as powerful.

Lord Sojar · Oct 14, 2009, 08:58 PM // 20:58

The 5850 is a great buy, no doubts there. G300 will be for higher end computers, to be perfectly honest. The release will be intended to smash the 5870 and 5870X2 into the ground. Since, by current numbers, G300 should be able to at least match the 5870X2, the dual derivative card should be the absolute pinnacle of performance, but as usual, don't expect a low price on a dual card solution. If you are thinking about buying a 5870, wait. If your budget only allows a 5850, invest now and get one. Hope that helps.

Elder III · Oct 14, 2009, 09:57 PM // 21:57

Quote:

Originally Posted by Rahja the Thief

The 5850 is a great buy, no doubts there. G300 will be for higher end computers, to be perfectly honest. The release will be intended to smash the 5870 and 5870X2 into the ground. Since, by current numbers, G300 should be able to at least match the 5870X2, the dual derivative card should be the absolute pinnacle of performance, but as usual, don't expect a low price on a dual card solution. If you are thinking about buying a 5870, wait. If your budget only allows a 5850, invest now and get one. Hope that helps.

That sounds about like what I expected...... frankly though, with the games that are out on the market now or that are likely to come out this fall season, I don't see any need to get a 5870 over a 5850 - nothing out there that the cheaper card can't handle and if you have a Crossfire capable mobo you are well set for 2-3 yrs of gaming in my ever so humble opinion.

I am curious to see how much the G300x2 (or whatever it will be called) can beat a 5870x2 by and what the price difference is.... time will tell.

Improvavel · Oct 14, 2009, 10:42 PM // 22:42

Quote:

Originally Posted by Rahja the Thief

The 5850 is a great buy, no doubts there. G300 will be for higher end computers, to be perfectly honest. The release will be intended to smash the 5870 and 5870X2 into the ground. Since, by current numbers, G300 should be able to at least match the 5870X2, the dual derivative card should be the absolute pinnacle of performance, but as usual, don't expect a low price on a dual card solution. If you are thinking about buying a 5870, wait. If your budget only allows a 5850, invest now and get one. Hope that helps.

So shouldn't we expect smaller G300 variants?

By the time G300 arrives the 5850 should be sub $200 and the 5870 around the $250 with the 5870x2 on the $400 market.

And then rumours have the 6xxx series by the end of 2010, so if nVidia takes as long to get a 2xG300 as it did to have the GTX295 it might have to compete with the 6xxx series.

Grabbing popcorns to enjoy engineers teams go on the fight

Lord Sojar · Oct 15, 2009, 03:08 AM // 03:08

I doubt ATi will release the 6000 series by the end of 2010. That is assuming it is a new architecture, and it better be, or they have no chance of keeping this momentum going. They will need a brand new architecture to continue this.

The 5770 is a great card too, btw. Though, the GTX260 is a better choice, but... the 5750 certain takes the cake on budget cards, seeing as how it is priced the same as the GTS250 and kicks its ass. However, the GTX260 can be had for the same cost as a 5770, and it beats it, hands down.

KZaske · Oct 15, 2009, 06:30 AM // 06:30

This is interesting, not a video card but the first Tesla on the market? What do you think of this thing Rahja?
http://www.tigerdirect.com/applicati...7BBTkwCjCECjCE