512MB ATI HD 4870 ve 1GB GeForce 9800 GX2 Crysis Tests

[Schmitzer]

Test System

• Intel Core 2 Extreme QX9650 @ 3.00GHz

• ASUS P5E3 Deluxe

• 2×1GB OCZ DDR3 1333MHz 9-9-9-26

• Western Digital 250GB 16MB SATA2 7200RPM

• 1GB GeForce 9800 GX2 - 600/1500/2000mhz

• 512MB ATI HD 4870 - 800/3400mhz

Source 1

Source 2

ATI

[SiM]

Awesome... I guess it falls behind in 1920x1200 since it will be held back by the ram...

[staffsMike]

Thats good, but you already know whats going to happen when the new nvidia is released.

[this_is_gav]

Thats good, but you already know whats going to happen when the new nvidia is released.

It's going to be called the 10800GTX and will be based on the 8800GTS?

As ever, I'll reserve judgement until NDAs have lifted. Some Inq-led rumour (so lies then) suggests the HD4870 will be released after the HD4850, for whatever reason. They are apparently continuing to get very good yields, as they did with with the HD3000s.

[MadduckUK]

this may even work with folding @ home in the year 2013.. you never know!

[badass]

Fingers x'd this is a true representation of the 4870's performace with both cards running up to date drivers etc.
If this is true, the 4870 is sub £200 and is released in the next quarter, I might just buy one.

[badass]

Awesome... I guess it falls behind in 1920x1200 since it will be held back by the ram...

How come? They both have 512MB per GPU, 256 Bit memory interfaces and memory running over an effective 2GHz.

[SiM]

How come? They both have 512MB per GPU, 256 Bit memory interfaces and memory running over an effective 2GHz.

This doesn't say that...

Test System

• Intel Core 2 Extreme QX9650 @ 3.00GHz

• ASUS P5E3 Deluxe

• 2×1GB OCZ DDR3 1333MHz 9-9-9-26

• Western Digital 250GB 16MB SATA2 7200RPM

• 1GB GeForce 9800 GX2 - 600/1500/2000mhz

• 512MB ATI HD 4870 - 800/3400mhz

[staffsMike]

Yeah but since it's SLi The GX2 has 512mb per card, they dont share memory.

[Hicks12]

the only true 1gb card is the 3870x2 really (well dual card solutions) as they are on 1 pcb so i assume they share it.

Looks good but as much as i want amd/ati to get back on top for pricing etc i bet it will be killed instantly. Lets hope nvidia's 9900 series is a major flop like this 9800 series , and finally crysis is playable but they should trey more aa on it .

[badass]

Yeah but since it's SLi The GX2 has 512mb per card, they dont share memory.

exactily Its effectively a 512MB card

[badass]

the only true 1gb card is the 3870x2 really (well dual card solutions) as they are on 1 pcb so i assume they share it.

Looks good but as much as i want amd/ati to get back on top for pricing etc i bet it will be killed instantly. Lets hope nvidia's 9900 series is a major flop like this 9800 series , and finally crysis is playable but they should trey more aa on it .

The 3870 x2 is the same as the 9800 GX2. 1GB=512MB of un sharable memory per GPU.
Otherwise known as a poor, wasteful design.

[SiM]

Fair enough... I didn't know that... so it can't use both sets of vram at the same time?

[badass]

Fair enough... I didn't know that... so it can't use both sets of vram at the same time?

Correct. Making both the 3870X2 and the 9800 GX2 a pikey bodge. If they were designed properly, then they would have either designed the Chips in the first place with a Hypertransport-esque interconnect running much faster than any Hypertransport currently available* or modified ATi's 512 bit ringbus to run off die but in the same module effectively sharing the memory controllers on each die.
Imagine that - Crossfire on a single GPU, except without crossfire and with near linear performance scaling!
The problem with doing it properly is they would need a different core for the X2/GX2 cards or have lots of useless silicon on every single GPU card bumping up the cost of single GPU cards.

Before anyone lays into me, if the cards are designed right, the copper paths between the GPU's could be short and noise free enough to easily run as fast as the memory interfaces themselves. There also wouldn't need to be much of a latency hit for accessing memory connected to the other GPU.

[kalniel]

Before anyone lays into me, if the cards are designed right, the copper paths between the GPU's could be short and noise free enough to easily run as fast as the memory interfaces themselves. There also wouldn't need to be much of a latency hit for accessing memory connected to the other GPU.

But there's no point, because the other GPU would be using it first

Unless, of course, you have one GPU idle while the other is doing all the work, that'd be efficient

I love the OP comparison of current card vs yet to be released card. We love them here

Now get the nvidia yet to be released card vs ATI yet to be released card.. that'd be a nice comparison.

[aidanjt]

To coin a term from someone on a popular TV show.. *points at nV* "HaHa!".

4870 X2 anyone?.. Might actually get min. 30fps in Crysis :/