5770 is NOT good as 4870 or GTX260

[dug777]

I don’t think you looked at the benchmarks properly, specifically the percentages.

That and architectural differences would make no difference in that situation if the card is as starved for bandwidth like you claim


Sure you can, because minimums affect averages by definition.


If I show minimums that prove you wrong, will you retract your claims and apologize to the entire forum for being wrong?

Or will you simply continue to tip-toe around any results that don’t back your incorrect notions?

I mean you don’t accept [H]’s benchmarks that clearly show plot points putting minimums into a context that don’t back your claims. Why should I waste my time if you’ll do the same to my results?

Ha! Pigs will fly before that happens I suspect.

I share the concerns about discounting games that don't fit the 'theory', ie. the ones I have highlighted from the xbit review.

In other threads Azn is claiming that 5XXX is a very minor tweak to 4xxx (I believe the comparison he uses are two different core2due processors ), yet here he is happy to suggest they are so significantly different as to obliterate any memory bandwidth advantages the 4890 has in some games, such that the minimum fps actually improves relative to the 4890 as you increase resolution...in defiance of all logic if memory bandwidth was an issue.

 

[jvroig]

I hope this is not too off-topic since this is about the 128-bit memory bandwidth. In what resolution is the 128-bit bandwidth too small for? Is this more than enough for 1680x1050? How about 1920x1080? A post above mentioned something alluding to a 128-bit bandwidth being not ideal only for >1920x1080, is this correct?

 

[Meaker10]

128 bit SDR, 128 bit DDR, 128 bit GDDR3, 128 bit GDDR5

133mhz SDR, 200mhz DDR, 800mhz GDDR3, 1200mhz GDDR5?

128bit is only part of it, I assume you mean 128bit GDDR5 on the 5770 running at 1.2ghz? Then it depends on the details and amount of AA/AF you want to use.

 

[Number1]

I am happy with my Sapphire HD5770 so far. I bought it for the HDMI connectivity, power consumption, hardware video decoding, low noise and games. I can overclocked the card to the maximum in the ATI control panel and it runs fine providing I turn the fan to the max. The fan is really noisy at max.

I did not base my purchase decision on gaming performance alone.

My previous card was a Sapphire HD3600 so there is a big difference there.

Here is a DX 11 benchmarking program called Heaven Benchmark available at

http://unigine.com/download/

The DX 11 effects are gorgeous but my frame rate is disappointingly low. About 15 FPS at 1280X1024 with the card overclocked to the max and the CPU running at 4 GHz.

 

[Highmodulus]

One thing I would add about the 5770 is to buy one now if you are interested, as the newer 5770's are coming with the less expensive cooler from the 5750 rather then the nice full shroud one from the reference design.

Also to the poster who tried to support his argument by stating the Radeon drivers are "mature". Fail to you sir. Driver support for the 5xxx series is still problematic at best, and they are not getting as much out of them as they could. Additionally, the rather buggy nature is a drawback to 5xxx series.

Like Number1 above I went from a 3xxx series to a 5xxx series card and the performace is night and day, but the drivers seem a bit buggier (ATI + Microsoft Security Essentials = hard crashes on TF2 for me).

"Ha! Pigs will fly before that happens I suspect." I suspect they would evolve rockets and fly to the moon for space based bacon delivery before such an admission would be forthcoming.

 

[Daedalus685]

One thing I would add about the 5770 is to buy one now if you are interested, as the newer 5770's are coming with the less expensive cooler from the 5750 rather then the nice full shroud one from the reference design.

Also to the poster who tried to support his argument by stating the Radeon drivers are "mature". Fail to you sir. Driver support for the 5xxx series is still problematic at best, and they are not getting as much out of them as they could. Additionally, the rather buggy nature is a drawback to 5xxx series.

Like Number1 above I went from a 3xxx series to a 5xxx series card and the performace is night and day, but the drivers seem a bit buggier (ATI + Microsoft Security Essentials = hard crashes on TF2 for me).

"Ha! Pigs will fly before that happens I suspect." I suspect they would evolve rockets and fly to the moon for space based bacon delivery before such an admission would be forthcoming.

The new cooler only looks like the 5750, it is not the same under the hood. I have seen reports that it actually cools better than the shroud, fudzilla is the only one that comes to mind.. not exactly reliable... though there is no exaust.

I think that it is a fine card.. I value performacne more than dx11 as I update rather regularly and don't see a reason to replace my 4890's yet (obviosuly). But people have to understand that some people like the idea of a smaller lower power card. It is pretty simple, if raw perforamnce doesn't matter to you then don't get the card.. but understand that others arelookign for something else..

 

[jvroig]

The DX 11 effects are gorgeous but my frame rate is disappointingly low. About 15 FPS at 1280X1024 with the card overclocked to the max and the CPU running at 4 GHz.

Can anybody comment on this performance? It seems too low to me (but my opinion is worthless, which is why I'm asking), given that the resolution is low, plus the card and CPU are both overclocked?

Or is this Unigine Heaven Benchmark far slower and demanding than any real app because it needs to be very demanding in order to be a useful benchmark at all?

 

[Avalon]

Probably going to buy one for the lady, since she games at 1440x900. Will be a nice upgrade to her 3850. Waiting for a deal, though.

 

[Dribble]

Can anybody comment on this performance? It seems too low to me (but my opinion is worthless, which is why I'm asking), given that the resolution is low, plus the card and CPU are both overclocked?

Or is this Unigine Heaven Benchmark far slower and demanding than any real app because it needs to be very demanding in order to be a useful benchmark at all?

The DX11 mode makes lots of use of tessellation which means the graphics card generates a lot of extra triangles to make it look prettier, this obviously hits performance really hard. While the benchmark itself is silly (DX9 mode is so low poly the stairs are a slope!) the DX11 performance drop is to be expected.

DX11 means tessellation (big performance hit), higher quality lighting (performance hit) and shadows (big performance hit). Really you're going to want at least a 5850 if you are buying a card because you want to play DX11 games.

 

[T2k]

Unigine bench IS demanding. It's perfectly fine to get a rather low fps with a mid-level card like 5770 - does not mean anything about gaming, I used to play Clear Sky with a single 4850 in 19x12 (no AA and details & effects adjusted, of course.)

 

[AzN]

I don’t think you looked at the benchmarks properly, specifically the percentages.

That and architectural differences would make no difference in that situation if the card is as starved for bandwidth like you claim


Sure you can, because minimums affect averages by definition.

Different architecture act differently to games. It's the same reason why on radeon cards a game is faster and not on Nvidia and vice versa.

If I show minimums that prove you wrong, will you retract your claims and apologize to the entire forum for being wrong?

Or will you simply continue to tip-toe around any results that don’t back your incorrect notions?

Did you apologize to the entire forum when you were arguing SP made the biggest difference or did you just tip-toe around any results that didn't back your incorrect notion?

I remember when you did your ultra investigation you basically denied everything. You still deny it today when we have 20 page long of you arguing saying SP made the biggest difference.

As for me I would apologize but will the entire forum apologize to me when my notion is correct? Like I said investigate further into minimum fps and I have no problems setting the record straight.

I mean you don’t accept [H]’s benchmarks that clearly show plot points putting minimums into a context that don’t back your claims. Why should I waste my time if you’ll do the same to my results?

What about Hardocp? They're benchmarks aren't consistent. I wish they were. They're minimum fps are erratic even on their best to worst cards.

 

[Daedalus685]

What about Hardocp? They're benchmarks aren't consistent. I wish they were. They're minimum fps are erratic even on their best to worst cards.

Minimum FPS is always erratic. It is by definition a single point, thus it can be an outlier due to any number of things. Sustained minimum you gain from looking at a plot,the Stdev over a section of a plot, or the average minimum of certain time samples are useful values. The isntantaneous minimum does not mean anything unless it is adjusted to ignore ouliers, or processed in some way.

A sampling of 100 FPS values, 99 of them 60, and one of them 1 will show up as ave: 59.4fps, min:1. What exactly is that supposed to tell you?

 

[soccerballtux]

Bandwidth is fine it's call GDDR5 look it up
Twice the bandwidth of GDDR4, so basically, 128bit 5 bus = 256bit 4 bus.

 

[AzN]

Minimum FPS is always erratic. It is by definition a single point, thus it can be an outlier due to any number of things. Sustained minimum you gain from looking at a plot,the Stdev over a section of a plot, or the average minimum of certain time samples are useful values. The isntantaneous minimum does not mean anything unless it is adjusted to ignore ouliers, or processed in some way.

A sampling of 100 FPS values, 99 of them 60, and one of them 1 will show up as ave: 59.4fps, min:1. What exactly is that supposed to tell you?

That's why you get rid of point of errors which hardocp doesn't do it seems like. Even in hardocp's "graph" the hovering frame rates is lower on the 5770. I have no idea why GTX260 is getting crappy results with NFS shift as I have no problems running the game with full detail.

 

[Daedalus685]

That's why you get rid of point of errors which hardocp doesn't do it seems like. Even in hardocp's "graph" the hovering frame rates is lower on the 5770. I have no idea why GTX260 is getting crappy results with NFS shift as I have no problems running the game with full detail.

But how shoudl they choose which points are erratic.. I can filter a plot down to a flat line if I wanted.. it doesn't say much. I'd much rather them leave teh data un altered than filter it mysteriously and not tell anyone how.

It is still useful to see an unaltered plot if you ahve any dea what you are looking at. It is not hard to remove the erratic poitns yourself, it is at least complete to leave them in. If you start manually removing outliers you get into a whole nasty subjective area.. They could just remove every point ouside of the stdev, still result in a good looking and meaningful average, have a nice and consistend minimum... yet have the minimum mean even less than it did before.

 

[Wreckage]

Well the Tech Report still recommends the GTX260 in their system guide. I would say it's easily the best card of the last generation. The 4870 is right up there. The 5770 is maybe a bad 5870 chip.

 

[evolucion8]

I don't usually follow system recommendations from pages like Toms Hardware because the best way to determine a choice in videocard is doing multiple researches through many reviews. For example, the HardOCP Powercolor HD 5770 review, just a simple review of that card, but can be used to check how well the HD 4870 and the GTX 260 have aged in current and previous games, and shows that the HD 4870 was the fastest of the three cards, smoking the GTX 260 in Arma II, or being slighly defeated by the GTX 260 in Batman AA, but in the end at average, all is a wash anyways.

No, the HD 5770 is not a harvested HD 5870, the chip used in the HD 5770 is much smaller and cheaper to manufacture. The HD 5770 is a good upgrade from HD 3x00 owners, HD 4600 or below, 8800GTS or below, etc. But from the GTS 250 and above, or HD 4850 and above, is more of a sidegrade than anything, not worth it.

 

[soccerballtux]

Well the Tech Report still recommends the GTX260 in their system guide. I would say it's easily the best card of the last generation. The 4870 is right up there. The 5770 is maybe a bad 5870 chip.

4890 was best card last gen we all know that. 4870 was were the price/performance was at
5770 = low power 4870

 

[Wreckage]

I suppose.

 

[dajeepster]

5770 = low power 4870

NO IT ISN'T

the 5770 is only single precision where as the 4870 is double precision.

but the only program i know of where it matters is milkyway@home

 

[edplayer]

I suppose.

if the 4870 is "right up there", where is the 4890?

Maybe it wouldn't be a good buy if the premium was very large but I think its msrp is around $200 and you can buy a XFX version right now for about $155 (after rebate and bing)

 

[lothar]

I agree about the 5770, but the 6600GT was a better card than the 9700 and 9800 all around.

AFAIK, they were pretty much about equal.

 

[lloydxd]

and here I thought it was safe to get the 5770 and use eyefinity for WoW at 5700x1080 resolution--good thing I saw this thread.

 

[AzN]

NO IT ISN'T

the 5770 is only single precision where as the 4870 is double precision.

but the only program i know of where it matters is milkyway@home

Perhaps in the future games will use double precision. nevermind it has 50% lower bandwidth compared to 4870.

 

[BFG10K]

Different architecture act differently to games.

Different architectures? So now the 4850 is different to the 5770, is it?

So which is it, Azn? Are they different or not?

Now, if you discount it on the basis of two games, the burden of proof is with you to demonstrate that the 5770’s drivers are 100% optimal with the rest of them. Go ahead, show us.

It's the same reason why on radeon cards a game is faster and not on Nvidia and vice versa.

But how can that be given you claimed the 5770 is castrated by bandwidth? A superior architecture doesn’t matter if bandwidth holds it back, right?

Did you apologize to the entire forum when you were arguing SP made the biggest difference or did you just tip-toe around any results that didn't back your incorrect notion?

I had no trouble admitting the 8800 Ultra’s results surprised me.

Now, are you going to admit the GTX260+ is a balanced part, given core/mem/shader affected performance equally overall?

Are you going to admit that the 5770’s primary limitation is the core?

Go ahead, I’m waiting.

I suspect you’ll say those two don’t count because they used averages, but the 8800 Ultra’s do count because they use averages too, but you agree with them, so that’s okay.

I remember when you did your ultra investigation you basically denied everything. You still deny it today when we have 20 page long of you arguing saying SP made the biggest difference.

It did, but the 8800 Ultra had ample amounts of it, more than I thought. But I have no trouble admitting to learning things, unlike you who seems incapable of doing so.

Like I said investigate further into minimum fps and I have no problems setting the record straight.

I still didn’t get answer from you. If I show the minimums having the same effect as averages, will you tell everyone you were wrong and retract your claims about bandwidth?

What about Hardocp? They're benchmarks aren't consistent. I wish they were. They're minimum fps are erratic even on their best to worst cards.

And how do you think a minimum is derived? Again, do you understand the concept of what a minimum is? Do you understand the notion of a single data point?