The official Fermi benchmark thread

[GheeTsar]

i've been looking at it from a none bias none fanboy prospective and it seems worth it.

its faster, more features and doesnt go over 80c on the "user reviews" i have seen. the 5870 goes around 75 on load i think so not much difference.

I presume fan noise doesn't bother you then?

[kalniel]

i've been looking at it from a none bias none fanboy prospective.

its faster, more features and doesnt go over 80c on the "user reviews"

These two statements seem mutually exclusive

[Domestic_Ginger]

Speaking of features; nostrodamus has surfaced on bit tech. Is he humbled my arse still the same record.

[j.o.s.h.1408]

These two statements seem mutually exclusive

hehe but just like reviewing games, would you trust what IGN was paid to say about a game compared to users own reviews of a game?

Anyways across all benchmarks from hexus etc all indicate this card being the fastest and most feature richest card out and that is a FACT.

[this_is_gav]

Anyways across all benchmarks from hexus etc all indicate this card being the fastest and most feature richest card out and that is a FACT.

Fastest GPU, not card.

[kalniel]

Anyways across all benchmarks from hexus etc all indicate this card being the fastest and most feature richest card out and that is a FACT.

The 5970 is faster. That is a FACT

hehe but just like reviewing games, would you trust what IGN was paid to say about a game compared to users own reviews of a game?

What users? This card isn't even out yet.

[j.o.s.h.1408]

here is an interesting post made comparing teh two based on what figures and features that have been mentioned across various benchmark results:

I have been wanting to see what sort of power load a 5870 has once overclocked and volt modded to a max of 1.35v. I also wanted to see how much of a performance boost a 5870 has with a volt mod and overclocked.

Taken from these two Hardocp reviews.

OC - Power - Temp - ASUS EAH5870 Video Card Review | [H]ard|OCP

Power - Temp - Sound - NVIDIA Fermi - GTX 470 - GTX 480 - GTX 480 SLI Review | [H]ard|OCP

Asus 5870

5870 idle = 139w
5870 load = 341w

5870 oc idle = 139w
5870 oc load = 355w

5870 oc + volt mod idle = 142w
5870 oc + volt mod load = 426w

This was on a 775 socket setup.

As you can see there is a massive 85w difference between the volt modded load and stock load.

The overclock and volt mod boost gave the following clock speeds.

5870 core = 850Mhz
5870 mem = 1200Mhz

5870 oc core = 950Mhz
5870 oc mem = 1300Mhz

5870 oc + volt mod core = 1010Mhz
5870 oc + volt mod mem = 1300Mhz



The temps are as follows.

5870 idle = 45c
5870 load = 87c

5870 oc idle = 45c
5870 oc load = 88c

5870 oc + volt mod idle = 37c
5870 oc + volt mod load = 90c


Apparently the fan was at 100% for the volt modded card and it is mentioned how worrying it was to have the fan at 100% and be at 90c. The reviewer (Kyle Bennett) said that to get higher clocks better cooling would be needed.

Ok so this shows us that the 5870 volt modded and overclocked to 1010Mhz on the gpu and 1300 on the memory uses 85w more than a stock card and runs at 90c with the fans on 100%. They then bench test 2 games RE5 and need for speed shift. Both games get a boost with Re5 getting 10% from volt modding and need for speed getting 11% from the volt mod. This is from a 17% overclock on the gpu and 8.3% overclock on the memory.

As this test was done on a 775 socket and hardocps next test was done on a i7 rig I looked at the wattage increase from stock to volt modded and added this 85w figure to the GTX 480 review which also has a 5870 in it.

This review shows the following:

5870 idle = 179w
5870 load = 367w

5870 oc + volt mod idle (+3w) = 182w
5870 oc + volt mod load (+85w) = 452w

480 idle = 175w
480 load = 480w


As you can see a volt modded 5870 is 28w less than a stock 480.


Now if we compare the 2 temps of 480 and a 5870 with volt modding you will see there is only 3c difference between the two cards at full load.

5870 oc + volt mod load = 90c
480 load = 93c

Ok so we have a 3c difference and a 28w difference from both cards at load.

Now it has been said many times that the 480 is about 15% on average better than the 5870 when both are stock speeds. Now if we volt mod the 5870 it gives it a 10-11% boost in performance which closes the gap between the two cards to say 4-5%. So even with a volt mod a 5870 isnt quiet as fast as a 480 gtx at stock and its running temps and power usage are very close to the 480.

Now I have been reading a fair bit on the 480 and it seems that you can overclock the 480 by about 14-15% which gives a 804Mhz gpu speed from 700Mhz and a boost to 1058Mhz on the memory from 924Mhz. It also seems that the frame rate boost is closer to the jump in clock speeds. ie 15% would give a bigger percentage boost on the 480 than the 5870 in raw performance in terms of frames.


http://www.overclockersclub.com/vima...480/480_oc.jpg

Overclocking GTX-480 / EVGA Precision 1.9.2 - Bjorn3D.com

This review gives no clock speeds but shows approx a 15% boost in frames from overclocking at 1920x1200.

NVIDIA GeForce GTX 480/GTX 470 Performance Preview




So to round it all up in terms of frames and taking a stock 5870 as the base line 100%.

5870 = 100%
5870 + volt mod = 111%

480 = 115%
480 overclocked (15% boost) = 132.5%


As we are also on virgin beta drivers it is reasonable to assume that the performance of the 480 will increase with mature drivers. A 10-15% boost IMO is conservative for any card from launch to mature drivers. Most of the time its allot more than that. But lets work with 5,10 & 15% boost. Please note that this next part is purely speculation only and only time will tell how much maturer drivers will help boost performance.

480 virgin beta drivers = 132.5%

480 overclocked (5% boost) mature drivers = 139.1%
480 overclocked (10% boost) mature drivers = 145.75%
480 overclocked (15% boost) mature drivers = 152.4%

So potentially if both the 5870 and 480 were overclocked with mature drivers there could be a approx 28-40% difference between the two cards. If you accept that the 480 will see increases in performance with newer drivers I think it only fair to assume that the 5870 will also see some increases in performance with newer drivers. However bare in mind the 5870 drivers are allot more maturer than the 480's and this is the first set of beta drivers for the 480 which as history shows us is always improved upon by a fair bit, especially on new architecture.

If there is zero boost from drivers on the 480 which is totally unlikely or you totally ignore the influence of future drivers there is on average 21.5% difference in performance between the two cards once overclocked as it stands today.

If you compare a stock 480 with a volt modded 5870 there is about 4% separating them in terms of performance in favour of the 480 with 28w less power use and 3c less core temp on full load with the fan @ 100% on the 5870.


I would also like to add that the 480 is also capable of volt modding just like the 5870 but no data is available yet on how it boosts the clocking performance of the 480.


So what does everyone think of the above?

And here is the guy who has a 480

http://www.evga.com/forums/tm.aspx?&m=285823&mpage=2

[kalniel]

here is an interesting post made comparing teh two based on what figures and features that have been mentioned across various benchmark results:

Nope, that's comparing a modded 5870. The 5970 is far faster, and easily beats a 480.

And here is the guy who has a 480

Where are the user reviews? I just see one user who isn't comparing it to a 5970 in the same environment so I'm not sure where you get your comparative comments from.

[j.o.s.h.1408]

Nope, that's comparing a modded 5870. The 5970 is far faster, and easily beats a 480.


Where are the user reviews? I just see one user who isn't comparing it to a 5970 in the same environment so I'm not sure where you get your comparative comments from.

As someone pointed out and corrected me, the 480 is the fastest SINGLE GPU card out at the moment.

Aint the almighty 5970 cost more and produces more heat then a 480?

[GheeTsar]

I just see one user ....

It's rumoured to be Rollo....

[kalniel]

Aint the almighty 5970 cost more and produces more heat then a 480?

I can walk out of a shop today with the 5970 for £517. I can spend £1500 and still not be able to get a 480, as you can't buy one.

The 5970 produces less heat than a 480, and uses less power (which is how it produces less heat, obviously). Come on j.o.s.h.1408, you said you were going to look at this without bias or fanboi-ism - look at the facts for yourself!

[j.o.s.h.1408]

I can walk out of a shop today with the 5970 for £517. I can spend £1500 and still not be able to get a 480, as you can't buy one.

The 5970 produces less heat than a 480, and uses less power (which is how it produces less heat, obviously). Come on j.o.s.h.1408, you said you were going to look at this without bias or fanboi-ism - look at the facts for yourself!

if thats the case then **** it. ATI here i come

[kalniel]

if thats the case then **** it. ATI here i come

If you really want best value for money, with 480 beating performance, go for a pair of 5850s - see the Hexus review here for why:
http://www.hexus.net/content/item.php?item=24061

[Fraz]

If you really want best value for money, with 480 beating performance, go for a pair of 5850s - see the Hexus review here for why:
http://www.hexus.net/content/item.php?item=24061

Problem with that solution - as ever - are the instances of crossfire/sli fail, where you end up getting way worse frame-rates than if you just had a single 5850.

[j.o.s.h.1408]

I' BUY A 5870. JOB done lads. screw NV and their power hungry money grabbing gits

[kalniel]

Problem with that solution - as ever - are the instances of crossfire/sli fail, where you end up getting way worse frame-rates than if you just had a single 5850.

They are few and far between these days, and with the XF set up it's easy enough to disable XF for a particular game.

But it's not like most people really need that kind of performance yet anyway