Sapphire Radeon HD 4850 X2 2048 MB

[Binge]

There are a lot of reviews on the front page. Some seem a little rushed (specifically Benchmarker's), but one thing I'm noticing is that reviews which are getting better results for the 4870x2 than the 4850x2 have faster ram. I have no authority to say that's why the results vary in such a way, but I've seen crazy performance changes in my machine when I beef up the memory bandwidth. Can't wait to hear what ATi has to say about this.

 

[douglatins]

Now here is a review i believe

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=244&Itemid=1
http://www.hardwaresecrets.com/article/645/7

Read and wheep 4850X2 lovers, 4870X2 FTW
LoL

Now one thing 4850X2 has in its favours, dual fan and 4 DVI, but i would rather see cooper heatsinks they are better and look the best

 

[TRIPTEX_CAN]

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=244&Itemid=1
http://www.hardwaresecrets.com/article/645/7

Read and wheep 4850X2 lovers, 4870X2 FTW
LoL

Now one thing 4850X2 has in its favours, dual fan and 4 DVI, but i would rather see cooper heatsinks they are better and look the best

Are you trying to rub it in people faces that ATI's second most expensive card is slightly slower than the most expensive card?


This just in.!!!!! The GTX280 is faster than the GTX260!!!!

lol im just kidding.

 

[W1zzard]

AMD is looking into it and they confirm that they have seen similar tendences in other reviews

 

[jbunch07]

This makes me think that if the 4870 X2 used GDDR4 or GDDR3 then it might actually perform better?

 

[niko084]

I very well may have one of these in a few weeks...
Can't wait to find out for sure, be a pretty fast machine if he gets it

 

[Darkrealms]

A thought on the 4870 vs 4850, is there any way the drivers could be optimized for the ddr3 memory? Something that no one has mentioned yet (drivers and memory type specifically) so I figured I'd throw it out there.

W1zz, again thanks for a great review! I especially like the new F@H addition.
A question though, with the GTX260 are you now using the 216sp or is there a mix in there? And in the future would it be very hard to specify that like you do with memory quantities?
Thanks for all the hard work.

1 negative shouldnt go against them and that is the CUDA Support, because of fact that Nvidia wont share the arch, just like SLI at the time. This Card is an excellent choice when pitted against the top dog cards (odd the slower chip is very comparable to both high end parts) And the average power draw isnt bad either.

If you think about this, its part of the cards performance. Consider it Nvidia drivers vs ATI drivers. If this was a third party utility I would agree with you.

I really think going with the GTX260(216) would be the best bet for buying a video card right now though. Overclocking to match a GTX280, and only $300 can't really be beat.

I'm tending to agree with you on that, and even cheaper than that with some of the rebates and sales out there!


This is just a personal frustration for me, even being an Nvidia fan, I want them to be challenged and beaten to make them produce not rehash (8000/9000 example).
Why is it that these 4870x2's and 4850x2's don't utterly destroy the GTX280's and 9800x2's. They are current tech (9800x2 last gen) and have double the memory . . . They should be KILLING Nvidia right now (across they board but the performance doesn't show that).
Again just a personal frustration.
2¢

 

[W1zzard]

A question though, with the GTX260 are you now using the 216sp or is there a mix in there?

it's the 196 SP card. should we ever add a 216 SP card to the test group it will be indicated properly

 

[v-zero]

it's nice to see that memory bandwidth becomes unimportant at high resolutions as the shaders become the bottleneck, at which point the lower latency of DDR3 compared to DDR5 gives it the upper hand

I'm fairly certain of this...

 

[douglatins]

Someone needs to explain to me in detail how the hell is this doing better than the 4870x2, it has to be a mistake...

Im in denial to be honest, i feel kind of bad for getting the 4870x2 , someone tell me it wasn't a bad choice....

Don't be, there's obviously something that need further investigation, just look for other reviews, even in their database, it should set your mind at ease. This is phenomenon present in this review and somewhat in overclockers

 

[Mussels]

for all we know, the test system here just happens to be the one in 3 million chance, of being perfect for the 4850, whilst holding back the 4870. it'll be interesting to see if a cause is ever found.

 

[sdedalus83]

Most of the tests are CPU limited. Adding CPU limited tests into the average performance is pointless. Why don't you test at 2560 x 1600 and give the averages for just that resolution?
Your 3.6 Ghz dual core processor doesn't help any either. These aren't cards that people are going to get for 24" or smaller monitors and moderately overclocked, air cooled dual core processors.

Preface your review with "Our conclusions are only meaningful if you have a mid range computer with screen smaller than 24"." Although in that situation, having anything more than an HD4870 512 or GTX260 192 is extreme overkill. You'd probably be better off with a 4850 or 9800GTX for around $150.

 

[TRIPTEX_CAN]

Most of the tests are CPU limited. Adding CPU limited tests into the average performance is pointless. Why don't you test at 2560 x 1600 and give the averages for just that resolution?
Your 3.6 Ghz dual core processor doesn't help any either. These aren't cards that people are going to get for 24" or smaller monitors and moderately overclocked, air cooled dual core processors.

Preface your review with "Our conclusions are only meaningful if you have a mid range computer with screen smaller than 24"." Although in that situation, having anything more than an HD4870 512 or GTX260 192 is extreme overkill. You'd probably be better off with a 4850 or 9800GTX for around $150.

You signed up just to post that? Gimme a break. :shadedshu

Go make your own review if you dont like W1zz's work.

 

[AsRock]

Most of the tests are CPU limited. Adding CPU limited tests into the average performance is pointless. Why don't you test at 2560 x 1600 and give the averages for just that resolution?
Your 3.6 Ghz dual core processor doesn't help any either. These aren't cards that people are going to get for 24" or smaller monitors and moderately overclocked, air cooled dual core processors.

Preface your review with "Our conclusions are only meaningful if you have a mid range computer with screen smaller than 24"." Although in that situation, having anything more than an HD4870 512 or GTX260 192 is extreme overkill. You'd probably be better off with a 4850 or 9800GTX for around $150.

Dam right W1z gives honest as possible reviews if there is a issue he does his best to correct if there is one. Which he already trying to get sorted out and lets face it he's not the only one with reviews with a review pretty much the same.

If any thing point your finger else were.

You signed up just to post that? Gimme a break. :shadedshu

Go make your own review if you dont like W1zz's work.

hehe yeah.

 

[zithe]

This makes me think that if the 4870 X2 used GDDR4 or GDDR3 then it might actually perform better?

Only at first due to driver development.


Can you clock GDDR3 that high? Wouldn't that mean a higher failure rate?

 

[W1zzard]

Why don't you test at 2560 x 1600

you people need to click the ads more so i can buy a big screen

 

[mlee49]

you people need to click the ads more so i can buy a big screen

Do you seriously make a profit from us simply clicking on ad's? If so I have no problem with clicking my way to help TPU!

 

[Ourasi]

you people need to click the ads more so i can buy a big screen

The first thing you should get is a new motherboard, the PCI 1.1 port on the current one is holding the HD4870x2 cards back where the GPU is the limiting factor, as we discussed in the feeedback thread for the HD4870x2. Try crossfire with 2 HD4870x2 on 8/8x PCIe 2.0 ports and see what gets lost, even 2 vanilla HD4870 in crossfire looses form on this config and remember a 16x PCIe 1.1 have the same bandwidth as 8x PCIe 2.0, and worse tech speccs...

We all know this by now, and even if ATI states that 16x PCIe 1.1 should be enough for HD4870x2, it does not mean it performs at it's best. And there is an economical insentive behind this statement aswell, as many still uses P35 boards and are potensial buyers.

The HD4850x2 is probably not hitting the PCIe 1.1 bandwidth bottleneck at all, atleast not as much as it's big brother, and will therefore seem closer to HD4870x2 then it really is. The GDDR3 is far behind GDDR5 in every way since the factor here is GB/s, not timings, and the speculation from some about the GDDR3 being better then GDDR5 is way off the mark, at best.....

 

[v-zero]

The first thing you should get is a new motherboard, the PCI 1.1 port on the current one is holding the HD4870x2 cards back where the GPU is the limiting factor, as we discussed in the feeedback thread for the HD4870x2. Try crossfire with 2 HD4870x2 on 8/8x PCIe 2.0 ports and see what gets lost, even 2 vanilla HD4870 in crossfire looses form on this config and remember a 16x PCIe 1.1 have the same bandwidth as 8x PCIe 2.0, and worse tech speccs...

We all know this by now, and even if ATI states that 16x PCIe 1.1 should be enough for HD4870x2, it does not mean it performs at it's best. And there is an economical insentive behind this statement aswell, as many still uses P35 boards and are potensial buyers.

The HD4850x2 is probably not hitting the PCIe 1.1 bandwidth bottleneck at all, atleast not as much as it's big brother, and will therefore seem closer to HD4870x2 then it really is. The GDDR3 is far behind GDDR5 in every way since the factor here is GB/s, not timings, and the speculation from some about the GDDR3 being better then GDDR5 is way off the mark, at best.....

That's all absolute rubbish. The reason crossfired cards get bottlenecked in 8x PCI-E 2.0 systems is because they have to transfer data between cards through the PCI-E lanes (yes , they do that), and hence they are hitting a wall there. In the 4870X2 and 4850X2 this bottleneck doesn't exist, and PCI-E 1.1 offers plenty of bandwidth for what is essentially a single card...

 

[Ourasi]

That's all absolute rubbish. The reason crossfired cards get bottlenecked in 8x PCI-E 2.0 systems is because they have to transfer data between cards through the PCI-E lanes (yes , they do that), and hence they are hitting a wall there. In the 4870X2 and 4850X2 this bottleneck doesn't exist, and PCI-E 1.1 offers plenty of bandwidth for what is essentially a single card...

You'd better wake up dude... Depending on the game, even a 9800GX2 can be limited on a 16x v1/8x v2 PCIe slot http://www.tomshardware.com/reviews/pci-express-2-0,1915-12.html
Some games are hurt badly in terms of performance, and can easily show the limits of the PCIe on a P35...

The next I might hear from you might be somthing like "but most games have no loss", and I'll just answer you this: Some games get hurt badly, and are proof that "plenty of bandwidth on PCI 1.1 16x" might be true on older hardware, but not on the latest high-end hardware..

COD4 are one of the games that get hit by the bottleneck, and thats in even lower resolution then wizz's highest one, and as I have stated before, some of the results in wizz's HD4870x2 review does not show the cards true potensial...

 

[eidairaman1]

thats why you use tweakguides to get the most out of your hardware, i mean you can always increase the PCIE clock to compensate for certain games, seems like intel should of never released the P35 and should of jumped to the P65.

 

[W1zzard]

The HD4850x2 is probably not hitting the PCIe 1.1 bandwidth bottleneck at all,

so you are saying that the slower card does not hit the bottleneck, while the fast card hits the bottleneck. that should make the faster card run at the bottleneck speed and the slower card below bottleneck speed. by definition of this the 4850 x2 MUST perform worse than the hd 4870 x2 which it does not do in many tests

 

[W1zzard]

dear ourasi, after giving you one infraction a few minutes ago i'll give you another one now because you apparently just registered here to crap on the 4870 x2 and 4850 x2 reviews (i knew i remembered your name from somewhere). if you are unhappy with our testing then start your own review site

 

[TRIPTEX_CAN]

you people need to click the ads more so i can buy a big screen

Why dont you just write horribly biased reviews for specific manufacturers to get free hardware? <--- Sarcasm.

Seriously though I'll start clicking adds just to help out. You're going to need an i7 (or Daneb) test bed eventually.

 

[W1zzard]

Why dont you just write horribly biased reviews for specific manufacturers to get free hardware? <--- Sarcasm.

Seriously though I'll start clicking adds just to help out. You're going to need an i7 (or Daneb) test bed eventually.

i7 hardware is coming. but i'm not sure if it will make much of a difference for the vga reviews. if i remember correctly i actually saw some i7 reviews that said games performance will go down. remember, most games use one (1) core