5850 just as fast as a 5870?

[dug777]

There's nothing I can do for you to accept proof with benchmarks about bandwidth limiting minimum fps with 5770 vs 4870/4890 GTX260 or 4870x2 beating out 5870. You can whine about core making the biggest difference but that's not even what I implied from the very start as I told you that about your ULTRA long ago which you still deny till this day. Calling me troll or telling me retract something I've never said. That's it.

Now you say bandwidth makes difference in minimum fps. Okay. That's all I wanted to hear. lol

For pete's sake, stop it already!

You haven't proved that the 5770's deficient min fps scores in many (but, crucially, not all) games is due to memory bandwidth.

All you have done (and I am going to explain this incredibly slowly and simply) is as follows.

1) post benchmarks that show the 5770's minimum fps are worse than the 4870/4890 (something nobody has disputed).

2) claimed that this MUST mean the 5770 is crippled by its memory bandwidth, and it couldn't be anything else (especially not drivers!).

Can't you see (and I have to admit that it's a tricky one) that correlation does not equal causation? You haven't eliminated any other possibilities (other than to suggest that because the 5xxx cores are derived from heavily tweaked 4xxx cores driver problems or architectural problems are impossible, without any proof other than you breathlessly claiming that).

Finally, I don't remember anyone claiming that bandwidth wouldn't make a difference to minimum fps. Kinda ironic and hypocritcal that you are claiming other people have put words in your mouth

As I understand it, what BFG suggested was that his minimum FPS data was consistent with his avg fps data from his testing.

This means that his core speed reductions still had a greater impact than equivalent memory speed reductions. This unequivocally doesn't fit with your argument that the 5770 is 'crippled' by its bandwidth; if it was, you'd expect reductions in memory speed to have a lockstepped reduction in fps (or very close to), as well as having far greater reductions in fps than an equivalent reduction in core speed.

BFG's testing suggested that the 5770 was a relatively well balanced part, with a slight but significant bias towards its core.

I don't know how much more simply I can break it down for you...

 

[God Mode]

For some reason, reading this thread made me angry towards the 5770.

 

[T2k]

For some reason, reading this thread made me angry towards the 5770.

 

[Highmodulus]

*beats head against wall*

Let me explain this once and for all.

The comparison between the 4890 and 5770 doesn't in any way shape or form demonstrate that memory bandwidth is the issue behind the lower minimum fps of the 5770 compared to the 4890.

All it does is show you that clearly something is different, which from that example, COULD be memory bandwidth.

So, having posited that the 5770 is memory bandwidth limited, you then test the theory (just like BFG did), by changing first core, then memory speeds, and observing the impact on fps.

If the 5770 was primarily bandwidth limited, you would expect that the results of that testing would show that changes to core speed had relatively limited impact, while memory speed changes had a significant impact (so if the 5770 was extremely memory bandwidth limited, you would expect performance to fall away in very close correlation to reductions in memory speed, while you would see more limited reductions in FPS from core speed reductions).

BFG found that there was a greater change in FPS (both min and avg) for a change in core speed than for a change in memory speed.

This means that the 5770 is more limited by its core than its memory bandwidth. No ifs, no buts, it is that simple. The theory that the 5770 is entirely or even primarily memory bandwidth limited has been disproven by that test.

What that doesn't explain is where the bottleneck or issue lies, but we know that there have been significant architectural changes to the core from 4XXX. There could be a hardware bottleneck anywhere in those changes and changed interconnects, or there could simply be a driver issue that is creating the bottleneck by not most efficently using the architecture.

*cries*

Thanks for the well written post.

 

[Meaker10]

I get my second HD 5770 tomorrow. You know what? It's going to be overclocked along with my other one to 1000/1400 and I am going to enjoy my gaming bliss.

 

[Highmodulus]

Also to continue the 5770 hijack- I own one and the drivers have been less then stellar so far- and a review of the ATI forums confirms this. Not conclusive evidence, but a solid trend.

Also comparing the 4890 to the 5770 often ignores issues with the 4890 for many users attracted to the 5770 and the 5750- power use, heat and card size.

http://www.anandtech.com/video/showdoc.aspx?i=3658&p=13

Especially look at the load power for the WOW test- the 5770 in CF uses only 8 more watts than a single 4890.

Similarly with temps, the 5770 is clearly better.

And with load noise- it is a MASSIVE difference.

So the 5770 is a great tool (except the drivers so far) for what it is supposed to do. Compete with the EOL, loud, hot, big and power hungry 4890 isn't it.

If noise, heat and power are not concerns- don't get a 5770 (or a 4890): get a 58XX series or wait for Fermi.

 

[God Mode]

^ You can also turn on the AC to balance out the heater for perfect temperature bliss.

 

[Meaker10]

Yeah but you will need ear muffs lol.

 

[AzN]

For pete's sake, stop it already!

You haven't proved that the 5770's deficient min fps scores in many (but, crucially, not all) games is due to memory bandwidth.

All you have done (and I am going to explain this incredibly slowly and simply) is as follows.

1) post benchmarks that show the 5770's minimum fps are worse than the 4870/4890 (something nobody has disputed).

2) claimed that this MUST mean the 5770 is crippled by its memory bandwidth, and it couldn't be anything else (especially not drivers!).

Can't you see (and I have to admit that it's a tricky one) that correlation does not equal causation? You haven't eliminated any other possibilities (other than to suggest that because the 5xxx cores are derived from heavily tweaked 4xxx cores driver problems or architectural problems are impossible, without any proof other than you breathlessly claiming that).

Finally, I don't remember anyone claiming that bandwidth wouldn't make a difference to minimum fps. Kinda ironic and hypocritcal that you are claiming other people have put words in your mouth

As I understand it, what BFG suggested was that his minimum FPS data was consistent with his avg fps data from his testing.

This means that his core speed reductions still had a greater impact than equivalent memory speed reductions. This unequivocally doesn't fit with your argument that the 5770 is 'crippled' by its bandwidth; if it was, you'd expect reductions in memory speed to have a lockstepped reduction in fps (or very close to), as well as having far greater reductions in fps than an equivalent reduction in core speed.

BFG's testing suggested that the 5770 was a relatively well balanced part, with a slight but significant bias towards its core.

I don't know how much more simply I can break it down for you...

Stop what? Even BFG AGREES "much later" after showing more evidence that minimum fps makes difference with more bandwidth yet he was arguing this whole time saying it was drivers like you. lol The whole basis of my other thread about 5770 was just that yet BFG says drivers you mindlessly been agreeing this whole time.

I never said that bandwidth doesn’t make a difference to the minimum, so stop lying

I'll say this again. All I told BFG was to test minimum fps relevance to bandwidth and compare to cards like 4870/4890 gtx260 before coming up with conclusion. Broader conclusion if you will. I never implied Juniper or Cypress was limited to bandwidth which makes biggest performance difference but limited to bandwidth compared to cards before it. If you don't understand that that's you. Don't blame your lack of understanding and try to put words in my mouth. :/

 

[hans030390]

Stop what? Even BFG AGREES "much later" after showing more evidence that minimum fps makes difference with more bandwidth yet he was arguing this whole time saying it was drivers like you. lol The whole basis of my other thread about 5770 was just that yet BFG says drivers you mindlessly been agreeing this whole time.

Wow, can you read? He never disagreed with you. He was only saying that the core is making a much bigger difference here than the bandwidth. His testing proves that, minimum framerate or not. Based on that, you can come to the conclusion that it's likely not the memory bandwidth that is causing most of the lower performance. It's either something screwed up with the 5xxx series or possibly the drivers.

 

[BFG10K]

Stop what? Even BFG AGREES "much later" after showing more evidence that minimum fps makes difference with more bandwidth yet he was arguing this whole time saying it was drivers like you. lol The whole basis of my other thread about 5770 was just that yet BFG says drivers you mindlessly been agreeing this whole time.

No, stop lying and twisting things. That wasn’t the issue at all. You know you’re wrong now so you’re playing semantic games to try to weasel out of it.

Right from the start I was arguing that the core makes the biggest difference, and I stated that I believe it comes from the drivers. You denied this and were harping on about bandwidth. I never argued bandwidth made no difference to the 5770 given my own results showed it did.

You also claimed that the core can’t impact minimums as much as bandwidth can, and that my 5770 benchmarks were invalid because they didn’t have minimums.

I have evidence that disproves both of your claims and instead of asking to see it, you’re now changing the story to imply that you only claimed there’d be some difference to the minimums from bandwidth. Well duh, nobody ever argued memory can’t impact minimums in some way. That’s a strawman you’ve set up to try to cover yourself.

I’m still waiting for an admission from you that the GTX260+ is a balanced part. After all, you accepted my 8800 results, and the GTX260+ results were done exactly the same way.

I’m also still waiting for an admission from you that after I produce minimum scores for the 5770 tests, you’ll retract your two claims above. The requirement for me to contact other websites is bullshit on your part.

 

[MrK6]

Man this thread is lulz. Let's see if Azn tries to pick himself up again *grabs popcorn*

 

[AzN]

Obviously some people here have comprehension problems.

My first reply in this thread

With all video cards bandwidth plays the equalizer to the core. Of course bandwidth is the limitation of the 5870 considering all the core doesn't have the breathing room it used to have with rv770. If it didn't it would be faster than 4870x2 by 15% or more in all cases but it's not.

The link above by OP only strengthen that concept as 5850 with 5870 clocks is only 2% slower than 5870.

What you are proposing is that 5850 have better drivers than 5870. We all know this isn't true.

It's quite unbelievable you can't comprehend this. I'm simply saying it doesn't have the bandwidth saturation it used to have with RV770 and is the reason why 5870 loses to 4870x2. Does this mean 5870 is primarily bottlenecked by bandwidth? Of course not although it could mean it's bottlenecked by bandwidth compared to RV770 but here you are putting words in mouth with insults and what not.

Your bottleneck investigation on the 5770 was missing minimum fps. You can't conclude on just on avg. fps alone. You should update your investigation further and show minimum fps of the 5770 where bandwidth has the most effects on instead of only showing avg fps that is swayed by core clocks to come up with that conclusion.

When I mention minimum I'm simply asking you to test further with cards before it to get a broader conclusion instead of just testing average fps to come up with your conclusion.

I've seen his article.

I agree when you "only" compare avg. frame rates but what about minimum frame rates?

When you compare the card to say GTX 260 or 4890 minimum frame rates isn't anywhere nar it.. It's fine for 1680x1050 but not more than that or barely get by 1920.

http://www.xbitlabs.com/articles/video/display/radeon-hd5770-hd5750_8.html#sect0

I even agreed when you only compare avg frame rates but it's much more than that as other tests show big discrepancy in minimum fps with 5770 compared to 4890 or GTX260.

Some people should learn how to read or at least try to understand where someone is coming from instead of derailing threads with insults.

As for you BFG I simply implied these things and yet you were rigorously talking about drivers is the reason why 5870 does not beat 4870x2, 5770 does not beat 4890 and how core made the biggest impact. You don't seem to take critical comment about your articles and get upset. I suggest you stop writing articles if you can't.

The basis of my other thread is just that. I wasn't comparing avg frame rates but minimum frame rates to show examples of this but now you agree minimum fps makes difference with bandwidth after you've been saying drivers when that's what I was arguing this whole time.

 

[dug777]

Never mind, water under the bridge.

 

[Voo]

Never mind, water under the bridge.

Yeah just give it up, you can't get anyone to agree if they just don't want to.. I mean it was funny for some time, but now it's just boring.

 

[hans030390]

It's quite unbelievable you can't comprehend this. I'm simply saying it doesn't have the bandwidth saturation it used to have with RV770 and is the reason why 5870 loses to 4870x2. Does this mean 5870 is primarily bottlenecked by bandwidth? Of course not although it could mean it's bottlenecked by bandwidth compared to RV770 but here you are putting words in mouth with insults and what not.

Here is what you were referring to:

"Of course bandwidth is the limitation of the 5870 considering all the core doesn't have the breathing room it used to have with rv770. If it didn't it would be faster than 4870x2 by 15% or more in all cases but it's not."

Let me repeat something, emphasis added:

"Of course bandwidth is THE limitation of the 5870..."

And we're the ones that can't comprehend?

 

[Highmodulus]

This thread is heading towards Epic. . .

 

[BFG10K]

Azn, I'm done wasting my time with you.

 

[AzN]

This thread is heading towards Epic. . .

Yes quite. I can believe most people in this forum can't even read. It's quite unbelievable.

 

[MrK6]

Yes quite. I can believe most people in this forum can't even read. It's quite unbelievable.

lol, simply as a bystander viewpoint - when about five people tell you you have no idea what you're talking about and prove to you, with evidence and logical progression, just how wrong you are, "they" aren't the problem .

 

[hans030390]

It's quite unbelievable.

Tell me about it.

 

[Meaker10]

Well using the leaked drivers my vantage score certanly increased.

 

[yacoub]

The bottom line is the 5770 would benefit from having a more reasonable bus width. 128-bit is pathetic in this day-and-age. We were going somewhere with 384-bit and 512-bit but then, because the new tech required more traces and whatnot, GPU makers either couldn't fit the same size bus widths on the card with the new GPU, or they could but it cut too much into their per-card profits so they didn't. And logically their marketing followed suit to try to make it seem like 128-bit is enough, even though more is indeed merrier when it comes to bandwidth between modern GPUs and DDR5 VRAM. Either way, we'll eventually see AMD release a refresh card down the road with a better bus width and you'll see the performance increase nicely.

 

[Meaker10]

No you wont.

You are all focusing on the 128bit rather than the actual bandwidth.

It's got 20-21% more bandwidth than the 4850.

 

[Voo]

lol, simply as a bystander viewpoint - when about five people tell you you have no idea what you're talking about and prove to you, with evidence and logical progression, just how wrong you are, "they" aren't the problem .

Hear hear, but I fear he won't understand that