At what price difference is a 5850 a justifiable upgrade over a 4890?

[cbn]

A 5850 is no where near 53% faster then a 4890.
Show me 5 benchies where a 5850 is 53% faster then a 4890 and I'll show you 25 that say it's not.

The video card itself is 53% faster than HD4890.

1440 stream processors @ 725 Mhz is 53% more TFLOPs compared to 800 stream processors @ 850 Mhz.

 

[happy medium]

Unless you OC the 5850, which it does much better than the 4890, for obvious reasons. A 1GHz 5850 crushes an OCed 4890 and is a much better value.

A 4890 also can hit 1000 core and has better shaders to boot.
Compared to a 150$ 4890 the 5850 is not worth one penny more then 225$
I only say that because of its better feature set. Performance wise it isn't worth 200$

 

[toyota]

A 4890 also can hit 1000 core and has better shaders to boot.
Compared to a 150$ 4890 the 5850 is not worth one penny more then 225$
I only say that because of its better feature set. Performance wise it isn't worth 200$

the "average" 4890 could not hit 1000mhz unless you are talking volt mod.

 

[happy medium]

The video card itself is 53% faster than HD4890.

1440 stream processors @ 725 Mhz is 53% more TFLOPs compared to 800 stream processors @ 850 Mhz.

I showed you the benchmarks and they speak for themselves.16/26% faster at most.

It's also well known that the shaders on the 58xx series are weaker then the 48xx series.

Why do you think the 5830 with 1120 shaders is beaten by the 800 shader 4890?

I don't care if it has a zillion tflops it's not more then 25% faster then a 4890 when gaming.

 

[MagickMan]

A 5850 is no where near 53% faster then a 4890.
Show me 5 benchies where a 5850 is 53% faster then a 4890 and I'll show you 25 that say it's not.

Unless they're both overclocked, which most of us generally do.

 

[happy medium]

the "average" 4890 could not hit 1000mhz unless you are talking volt mod.

4890 can be overvolted in MSI afterburner also.

And the average 5850 dosen't hit 1000 core either.

 

[cbn]

thats not the cpu limiting it though. you could probably lower that i7 down to 2.3 and still be basically 100% gpu limited in Crysis at 2560 with single 5850.

Well something is preventing HD5850 from delivering the full increase in playable frame rates. I would expect boost greater than 16% from a product with 53% more processing power.

Two HD5870s (in comparison) deliver an improvement of 65% over one HD5870. (Major difference here is that we are comparing cards of the same generation, not HD58xx to HD48xx)

 

[MagickMan]

A 4890 also can hit 1000 core and has better shaders to boot.
Compared to a 150$ 4890 the 5850 is not worth one penny more then 225$
I only say that because of its better feature set. Performance wise it isn't worth 200$

A 4890 @ 1GHz isn't the same as a 5850 @ 1GHz, by any stretch of the imagination. Nor is it as easy to reach as on the 5850. Got 5 minutes? Bam, you have a 5850@1GHz.

 

[happy medium]

Unless they're both overclocked, which most of us generally do.

Wrong again. A 4890 is a good overclocker 15%. and the 5850 20%.
So you gain 5%. Look at the link I provided the 5850 does not destroy the 4890 or gtx 285.

 

[happy medium]

A 4890 @ 1GHz isn't the same as a 5850 @ 1GHz, by any stretch of the imagination. Nor is it as easy to reach as on the 5850. Got 5 minutes? Bam, you have a 5850@1GHz.

Bam if you overvolt. It's not bam if you don't. it's 950 core. and if your lucky 1300 mem.

 

[cbn]

It's also well known that the shaders on the 58xx series are weaker then the 48xx series.

I don't care if it has a zillion tflops it's not more then 25% faster then a 4890 when gaming.

Why do you think its the shaders on HD58xx that are weaker?

P.S. I agree with you that something is wrong with how HD58xx scales in comparison to HD48xx.

 

[MagickMan]

And the average 5850 dosen't hit 1000 core either.

Yes they will. Not seen one yet that wouldn't do 1GHz @ 1.35v. Some will run it for less, but that's just luck of the draw.

 

[MagickMan]

Bam if you overvolt. It's not bam if you don't. it's 950 core. and if your lucky 1300 mem.

all can be overvolted easily. Got any more?

 

[happy medium]

Why do you think its the shaders on HD58xx that are weaker?

P.S. I agree with you that something is wrong with how HD58xx scales in comparison to HD48xx.

They are weaker. It was proven with the 5830 series with over 200 more shaders and looses to the 4890.

I'm no video card tech....but there was something cut off the 58xx cards to make the die smaller that probrobly effected it's shaders.

 

[cbn]

They are weaker. It was proven with the 5830 series with over 200 more shaders and looses to the 4890.

You are right

Hasn't ATI ever offered an official explanation for this?

 

[happy medium]

Yes they will. Not seen one yet that wouldn't do 1GHz @ 1.35v. Some will run it for less, but that's just luck of the draw.

You just proved my point I said"without being overvolted" stock is 1.2v.
They do not all hit 1000 core with no overvolt. They avg about 930/940.
Most 4890's do the same.

I'll give you 1000 core and 950 for the 4890 and it still will not beat a 4890 by 25%.
My link proves it. Please read the whole thread.

 

[happy medium]

You are right

Hasn't ATI ever offered an official explanation for this?

No, not that I have read.?

We had big conversation about it in the 5830 thread last month or a few weeks back.

It might also explain why a 1600 shader 5870 can't beat a 4870x2 with 1600 shaders and thats with crossfire overhead!.

2 4870's in perfect crossfire would smoke a 5870. Whats the overhead on crossfire 20%?

 

[MagickMan]

You just proved my point I said"without being overvolted" stock is 1.2v.
They do not all hit 1000 core with no overvolt. They avg about 930/940.
Most 4890's do the same.

I'll give you 1000 core and 950 for the 4890 and it still will not beat a 4890 by 25%.
My link proves it. Please read the whole thread.

No, stock is 1.0875. Have you even worked with these cards or are you going by what other people write? All these cards require more juice to excel, be it a 4890 or a 5850. The difference is how far they can be pushed before you reach diminishing returns. Yes, I read the article, Anand doesn't address that. Take a 4890 to it's limits and do the same with a 5850, you'll see the >25% difference and then some.

 

[happy medium]

No, stock is 1.0875. Have you even worked with these cards or are you going by what other people write? All these cards require more juice to excel, be it a 4890 or a 5850. The difference is how far they can be pushed before you reach diminishing returns. Yes, I read the article, Anand doesn't address that. Take a 4890 to it's limits and do the same with a 5850, you'll see the >25% difference and then some.

Yea, ok your right man a 5850 @ 1000 core beats a 4890 @ 950 (5%) core by 53% when gaming but a when at stock 5850 cant beat it by 25%. ok whatever.

Plaese read this review and get back to me.

http://www.anandtech.com/video/showdoc.aspx?i=3650

Quote
"Onething that’s very clear in these benchmarks is that as things currently stand, the 5850 has made the GTX 285 irrelevant (again). The 5850 is anywhere between 9% and 16% faster depending on the resolution."

From the review..............I go by math and Anantech reviews

 

[MagickMan]

Yea, ok your right man a 5850 @ 1000 core beats a 4890 @ 950 (5&#37 core by 53% when gaming but a when at stock 5850 cant beat it by 25%. ok whatever.

Plaese read this review and get back to me.

I didn't say 53%, WTF are you talking about? I'm saying at their avg OCing sweet spot the 5850 will crush a 4890 by >25% in most benchmarks and games.

Again, I read the review, it doesn't address OCing at all.

 

[cbn]

No, not that I have read.?

We had big conversation about it in the 5830 thread last month or a few weeks back.

It might also explain why a 1600 shader 5870 can't beat a 4870x2 with 1600 shaders and thats with crossfire overhead!.

2 4870's in perfect crossfire would smoke a 5870. Whats the overhead on crossfire 20%?

I was just looking at the HD5870 Crysis 2560x1600 Benchmark.

Cypress with 33% die area along with being 40nm vs 55nm only scores 36% increase in FPS when compared to HD4890.

This must be why ATI is changing architectures for N. Islands. (They are at the point of diminishing returns)

It would be interesting to go back and compare HD3870 to HD4870 using an i7 CPU? How much of an improvement occured with 2.5 times increase in stream processors?

 

[happy medium]

So you are saying if I overclock a 4890 by 15% and you overclock a 5850 by 20% ,you somehow get a 45% boost and the 4890 gets 15%?

I don't need a overclock bench to do simple math.

If you have a stock 4890 that is beaten by a stock 5850 by 15% to 25% and you overclock tham both by 15% 4890 and 5850 20% my math tells me the 5850 only gains 5% from the original 15/25%. so 20/30% right? Not 53% not even close.

 

[happy medium]

I was just looking at the HD5870 Crysis 2560x1600 Benchmark.

Cypress with 33% die area along with being 40nm vs 55nm only scores 36% increase in FPS when compared to HD4890.

This must be why ATI is changing architectures for N. Islands. (They are at the point of diminishing returns)

It would be interesting to go back and compare HD3870 to HD4870 using an i7 CPU? How much of an improvement occured with 2.5 times increase in stream processors?

I'm also surprized this hasn't came up more.
Some are saying its the drivers and they will unleash more power when Fermi arrives.

Who knows?

Edit:it's also wierd that a 5770 in crossfire beats a 5870.
or 2 5750 beats a 5850. There must be some crossfire overhead yet the single cards loose?
Don't make sense.

 

[RussianSensation]

I would say if you are upgrading from 4890 and can sell it used for $150, then 5850 for $225 is a decent deal.

For the most part, the $300 price of 5850 right now is simply overpriced. Check a compilation of games at 1920x1200 comparing 5850 to 4890:

http://www.xbitlabs.com/articles/video/display/radeon-hd5830_6.html#sect0

However, in more modern games, I expect 5850 will start to pull away.
BF:BC2 - http://www.techspot.com/article/255-battlefield-bad-company2-performance/page5.html

Metro 2033 - http://www.pcgameshardware.de/aid,7...tX-11-und-GPU-PhysX/Action-Spiel/Test/?page=2

However, I personally would upgrade to at least a 5870 / GTX 470+

 

[happy medium]

I would say if you are upgrading from 4890 and can sell it used for $150, then 5850 for $225 is a decent deal.

For the most part, the $300 price of 5850 right now is simply overpriced. Check a compilation of games at 1920x1200 comparing 5850 to 4890:

http://www.xbitlabs.com/articles/video/display/radeon-hd5830_6.html#sect0

However, in more modern games, I expect 5850 will start to pull away.
BF:BC2 - http://www.techspot.com/article/255-battlefield-bad-company2-performance/page5.html


Metro 2033 - http://www.pcgameshardware.de/aid,7...tX-11-und-GPU-PhysX/Action-Spiel/Test/?page=2

However, I personally would upgrade to at least a 5870 / GTX 470+

I agree...+1

Edit: I would love to see how a 5830 does vs the gtx 285 in Metro 2030. Seems as though it would beat it and kill a 4890. Even a 5770 beat a 4890.

Edit 2: looks like the 5850 beats a 4890 by about 20% in BC2.