GTX 460 Reviews

Hardocp shows more of their poor benchmarking. After much complaining they finally rebenchmarked with newer drives.



BF BC 2
5870 x2 gtx 460 super overclock x 2
10.7a 10.8
11/68/44.1 24/92/63.6 25/73/51.8


They could have of course come to the conclusion that the driver broke crossfire for bf bc 2 by simply using 10.6 drivers but hardocp has been a bunch of tools for a long time.

 

Indeed, so did I. Are you sure there are not other factors at play that makes RV770 do one thing an Evergreen another?

 

For RV770, apart from clocks the differentiator was that 4850 used GDDR3 and 4870 used GDDR5, the difference in memory b/w alone was worth quite a bit of performance.

For Cypress and Juniper, due to need for higher b/w, GDDR5 was used for all variants. So they had to differentiate the variants by some other means other than just clocks. Im guessing thats why they disabled some units. Though yield might have been a factor as well

Ps: Why haven't we seen an updated version of the 5830(say a 5840 with more units enabled) to counter GTX 460?

 

I think HD5830 with 32 ROPs enabled would be enough to compete GTX460. Anyway, Bart is on the way and I expect it will be much better competitor than HD5830 (faster, cheaper - I expect almost 100mm² smaller die than GF104, cooler and more consitent in performance).

 

Nope, just making use of the information available to the unwashed masses. If you're implying that there are extenuating cirumstances that make Cypress more reliant than GF100 on disabling units to differentiate parts then I don't know what they are. Is Evergreen so easily overclockable that simply lower clocked parts don't make sense?

 

ATi wants to prevent users from buing a cheaper card and overclocking it to the level of more expensive part. HD3800 series was scaled by different memory modules, HD4800 series was scaled by different memory modules, but the entire HD5800 series is based on GDDR5…

 

Anyone?!

 

Colorful Unveils GeForce GTX 460 iGame Graphics Card.

http://www.techpowerup.com/129947/Colorful_Unveils_GeForce_GTX_460_iGame_Graphics_Card.html



Apart from 900 MHz (core), the CUDA cores are clocked at 1800 MHz, and the memory at 1050 MHz (4200 MHz effective).

 

Who is colorful? But that's a nice looking cooling system.

 

What's up with mini-HDMI? Afaik everyone but Palit uses it on their 460's.

HDTV's for gaming are becoming increasingly popular and if an adapter is needed to use HDMI then nVidia is needlessly annoying lots of customers IMO.

I'm glad to see MSI refining their Hawk line. I'm hoping to see such a version of the replacement of the 5770.

That might fit my upgrade schedule.

 

Thats exactly what i said in the last page

Well DP performance of GF104 is already much lower to begin with, so i dunno if they will cripple it in consumer cards. Where did u read about these cracked drivers? Soft modding consumer cards to Quadro's hasnt been done since the Geforce 6800 days

Using mini HDMI they can fit dual DVI + HDMI in the lower slot leaving the entire upper slot for venting out hot air. ATI's cards have only half the top slot for venting hot air thanks to the dual DVI+DP+HDMI ports on their cards. Nvidia dont bundle mini HDMI to HDMI adapters tho which is very annoying

 

GF104, GF100 Core Architecture Comparison
http://news.mydrivers.com/Img/20100730/02501268.jpg
GF104 SM architecture (part of the speculation)
http://news.mydrivers.com/Img/20100730/02503995.jpg
GF100 SM architecture
http://news.mydrivers.com/Img/20100730/02504021.jpg
NVIDIA graphics core in recent years, the evolution diagram
http://news.mydrivers.com/Img/20100730/02521912.jpg
G80, GT200, GF100, GF104 contrast the core memory and multithreading
http://news.mydrivers.com/Img/20100730/02521937.jpg

 

My gigabyte 460 came with an adapter.

 

I think Dave is implying and others are chiming in with, is that disabling units in 5850 has less to do with yields and more to do with product differentiation.

Sure 5850 is going to allow for the use of some salvage parts, but it's also quite likely that demand for 5850 is far greater than the amount of salvage chips of 5850 class available. 5830 is where things get murky. I doubt ATI would deliberately cut down a chip capable of being a 5870/5850 down to 5830 levels. Then again, demand for 5830 isn't exactly great, IMO, so they don't need a lot of them. Amusingly enough reviews questioning the existence of 5830 at it's price point helps somewhat in keeping it low demand.

Now, whether you believe that to be true or not is up to you. But I believe that's what Dave is implying. And disabling parts of the chips would be the only way to prevent enthusiasts from overclocking or otherwise turning a 5850 into a cheap 5870. Well other than using GDDR 3 for 5850, but I have a feeling that would have crippled the card far too much.

Just thought of another reason, that's only semi-related to yield. It could also be that Cypress chips that can't meet 5870 power requirements are relegated to 5850 class cards. Disabled units would help to lower power use of the chip that in all ways would opperate just fine as a 5870 except with higher power consumption. Although I think the power use reduction is probably minor at best. IE - nowhere near the difference we see between GTX 480 and GTX 470. As well I think one of the things that was pointed out about 5830 is that even with all those units disabled it still uses more power than 5870. I wonder how much those chips would use if the disabled units were functional and enabled? Perhaps near Fermi levels?

Regards,
SB

 

Sure, if you subscribe to the general notion that Cypress yields are awesome I only know what I see, and what I see are parts that are more cut down than they were last generation. The 5830 really throws a wrench into any theories of yield nirvana.

For the 5850 I could see them cutting back on units if downclocking would be too easy for end users to circumvent (by overclocking).

 

This is wrong, says 128bit for GF106, while the PCB has spots for 6 memory chips, aka 192bit

 

http://forum.beyond3d.com/showthread.php?p=1450418#post1450418

 

Excellent explanation, thank you! I'm surprised I hadn't stumbled on this in a review. Is it so obvious?