Which ATI 5 series card to replace 4870 1GB

[gcwebbyuk]

After modifying my Zalman GV1000 cooler and being very happy with the sound levels, I have come across a snag.

I bought an Asus Essence STX sound card, and was unable to install it due to the thickness of the modified cooler, and the screw mounts on the back of the card - this ruled out the small PCI-E slot above the GPU were my old Xonar DX was installed, the small PCI-E slot below the GPU and the full size PCI-E slot.

I have have had to switch back to the standard Zalman fan and put up with the extra noise for the moment.

I "could" just replace the cooler with something else instead, but while I am in upgrade mode, I am considering buying a 5 series card.

Which is the best card to get to better the 4870 - not really by much as am pretty content with the performace of the 4870. What I am most interested in though is the noise level of the stock cooler.

SO basically which card will be better than the 4870 and make it worth the upgrade, but also nice and quiet.

[gcwebbyuk]

Curently thinking Saphire Vapor-X 5770, even though it's a similar performance to the 4870.

Does anyone have any experience of this card compared to a Zalman GV1000? The Zalman is quoted as having a noise level of 17 dBA

[dhanson865]

If you want to take the easy route and not have to mod the card you can get a passive 5750 and take a slight hit in performance for a completely silent card.

No muss, no fuss, bam you have silent gaming card.

a typical 5750 card will idle about 13W, 54 watts in a game and peak about 71W in furmark or similar.

4870 is about 60 watts idle and 150 watts in a game and higher in furmark

that's probably a 15 or 20% hit in performance for 1/3 the power and none of the noise.

The 5770 is about a wash vs the 4870 within a couple of percent. Sometimes slower sometimes faster depending on the game an screen resolution.

Unfortunately the 5770 generates too much heat for a passive solution to be safe for all cases. You can get away with that in a cool room with good airflow in the case but if you are pushing warm ambient temps or are doing the zero or one case fan attitude you'll probably have to put a fan on the 5770 heatsink even if you mod an accellero S2

To me it isn't worth the extra 15% performance to have to deal with the heatsink + fan issues on the 5770.

The 5850 would be noticeably faster than all 3 of the cards mentioned above (something like 50% faster than a 4870) it also puts out half again more heat than the 5770 so again going fanless isn't an option.

However if you like modding a 5850 with an accellero S1 and cutting/bending fins doesn't scare you you can make a nice 5850 with effort and a fan or two of your choice.

Call me lazy, call me a wimp, but I'd take the 5750 fanless and avoid the hassle of modding my own 5850.

Here is a quick recap of power numbers

Code: Select all

card idle load furmark
4870  60   150  185

5750  13    54   71
5770  18    78  100
5850  19   112  150

[gcwebbyuk]

Thanks for your reply.

A month or two ago I probably wouldn't have worried about a performance hit as the only games I played worked 100% fine with the 4870 at full settings (Dragon Age).

I have started playing Borderlands and occasionally Far Cry 2, and these can't manage full settings on the 4870, and am worried what they would run like on the 5750.

I don't think I do enough gaming to warrant a 5850 either.

It's a tricky one...

[psyopper]

5770 then, or the 5830, depending on your budget. As pointed out, the higher numerical value the card is, the more heat it needs to dissipate.

[Cryoburner]

According to the benchmarks at Tom's, the stock 5770 is often up to 10% slower than a 4870, and is probably meant more as an upgrade to the 4770 or lower. The Vapor-X version might fare slightly better, but probably not enough to match the 4870's performance. I think you'd have to move up to at least a 5830 to actually improve overall performance by a small margin. The only advantages I see the 5770 having over the 4870 are things like DX11, Eyefinity and lower power consumption. Performance-wise it's a slight downgrade.

If I were in your situation, I'd probably wait on replacing the card.

[dhanson865]

According to the benchmarks at Tom's

According to the benchmarks at http://www.techpowerup.com/reviews/Powe ... en/28.html the 5770 is faster than the 4870 at higher resolutions and slower than the 4870 at lower resolutions. On average out of all their tests it was even.

Maybe it's relevant that tom's tested about half as many games as techpowerup did.

[gcwebbyuk]

Well I try and play all games at the native resolution of the screen 1920x1080 and aim for 60fps

Just checked borderlands and at all settings high and options switched on it does get 60fps, it was a v-sync issue before that was causing me problems.

Looks like a Vapor-X 5770 is on the shopping list then

[gcwebbyuk]

Just checked, I can get £60 trade in for my 4870, so a 5770 is only going to be:
Vapor-X: £79.99
Powercolor: £56.29

or a 5830:
Powercolor: 137.99

5850:
Powercolor: £174.99 inc CODMW2 - card would be £144.99
Vapor-X: £187.98

Am thinking either the Powercolor 5770 or 5850

I think I read that the powercolor cards are quieter than the Vapor-X am I right?

[gcwebbyuk]

http://www.techpowerup.com/reviews/Powe ... us/28.html

24.3dbA idle - 32.4dbA load

But will that be similar noise level to the Zalman?

[psyopper]

I would opt for the 5850 - you are getting a considerable upgrade in both performance and pwoer management for the same price as you would have normally had marginal (at best) performance increase.

[Cryoburner]

According to the benchmarks at Tom's

According to the benchmarks at http://www.techpowerup.com/reviews/Powe ... en/28.html the 5770 is faster than the 4870 at higher resolutions and slower than the 4870 at lower resolutions. On average out of all their tests it was even.

Maybe it's relevant that tom's tested about half as many games as techpowerup did.

Err... according to that summary page you linked to, the 5770 had slightly lower average performance at every single resolution tested aside from 2560x1600.

More importantly, note that the 4870 tested there was only a 512 MB model, which is why the 5770 was able to nearly keep up, and why it managed to pull ahead at the highest resolution. With a gigabyte of video memory, the 4870 would have likely been a bit faster at the higher resolutions, and would have outperformed the 5770 at 2560x1600 as well. The amount of memory on the cards is probably more relevant than the number of games tested. : )

The 5770 does certainly win on the matter of heat and power consumption though, especially under load. If I were to choose between the two right now, I'd definitely go with a 5770, as the small performance difference wouldn't matter as much as its superior thermal performance. From a quiet PC standpoint, the 5770 is certainly a better card. If I already owned a 4870 though, I really couldn't see myself replacing it on that matter alone.

Of course, you could probably sell the 4870 fairly easily to recoup much of the cost of switching cards. The performance difference between the two is probably not going to be very perceptible either, and it sounds like you would still need to buy a new cooler if you wanted to keep your current card. I guess it could be considered a reasonable upgrade in that case, though not so much from a performance standpoint.

[dhanson865]

According to the benchmarks at Tom's

According to the benchmarks at http://www.techpowerup.com/reviews/Powe ... en/28.html the 5770 is faster than the 4870 at higher resolutions and slower than the 4870 at lower resolutions. On average out of all their tests it was even.

Maybe it's relevant that tom's tested about half as many games as techpowerup did.

Err... according to that summary page you linked to, the 5770 had slightly lower average performance at every single resolution tested aside from 2560x1600.

Who games at 1024x768 (.7MP) or 1280x1024 (1.3MP) anymore on a serious gaming card? 1680x1050 (1.7MP), 1920x1080 (2MP) monitors are cheap. If you are playing Crysis you might choose a low enough res (1440x900 (1.3MP) or 1600x900 (1.4MP) depending on your aspect ratio) that the 1.3MP bench helps but the .7MP test is totally worthless to me. I just assumed you and any serious gamer looking at those benches would focus on the tests at the 3 highest resolutions.

At 1.7MP it was 3%, at 2.3MP it was 3%. 3% is a far cry from 10% The difference being I'm talking average numbers and you used the phrase "up to 10%" meaning you are focusing on the worst case for the 5770 while not giving it credit for where it does better.

Every game and resolution tells a different story and picking worst, average, or best puts a different spin on it as well.

It's up to each of us to decide which benchmarks in an article speak to us as relevant to how we use a PC. If we go by gcwebbyuk's target of 60FPS and avoid 2560x1600 due to the ram difference we get :

Code: Select all

I left out FC1, Prey, Quake4, UT3 as they are insanely high frame rates so the difference hardly matters and I got tired of doing math.

techpowerup                    Toms
Game    4870  5770 %gain/loss  4870  5770 %gain/loss
BF      72.0  60.6 -16%
COD4    71.8  71.9   0%
COJ2    72.3  72.0   0%
COH    101.9 119.5  17%
Crysis  47.4  44.2  -7%        44.5  40.9 -8%
DOW2    58.8  60.6   3%
DiRT2   62.9  64.7   3%
ET:QW   81.3  71.3 -12%
FC2     58.7  62.6   7%        69.5  68.7 -1%
GTA4                           56.5  46.7 -13%
HAWX    67.2  61.3  -9%        63.0  53.0 -16%
L4D                            66.7  63.7 -4%
R:DA    65.8  57.0 -13%
RE5                            70.2  62.7 -11%
STALKER 69.8  67.7  -3%
S-CS    61.4  59.2  -3%        46.1  43.3
WiC     51.0  45.0 -12%        64.0  56.0 -12%

Avg                -3.2%                  -9.3%

It varies widely but none of the games Tom's tested favor the 5770 at ~60FPS. Techpowerup tells a different story. Four games favor the 5770.

There's 3 kinds of lies.

1 Lies
2 Damn Lies
3 Statistics

btw thanks for making me reexamine these. It was an interesting diversion for a lazy afternoon.

[Cryoburner]

That still ignores the fact that the 4870 TechPowerUp tested was only a 512MB model though. It's not just at 2560x1600 that the additional memory will make a difference. Games that utilize a lot of video memory can see better performance with a 1GB card, even at more moderate resolutions. My point is that TechPowerUp is comparing a lower memory version of one card against a higher memory version of another, making the results skewed for anyone comparing two cards of equal memory.

For proof, we'll look at one of the tests where the 1GB 5770 substantially beat the 512MB 4870, Far Cry 2. For these tests, TechPowerUp used the highest image quality settings and 4x AA at 1680x1050, 1920x1200 and 2560x1600, which are the resolutions we'll look at. In the tests, the 1GB 5770 averaged 45%, 62% and 248% higher framerates than the 512MB 4870, for each respective resolution. That's a substantial win, but one entirely due to the amount of on-board video memory.

To prove this, we can simply look at another test at Tom's where they compared the effect of the amount of video memory on the performance of various games, using none other than a set of HD 4870s. One of their tests was Far Cry 2, where much like TechPowerUp, they used maximum settings at the same three resolutions, both with and without 4x AA. The results with AA enabled on the lower half of the page look much like the results at TechPowerUp, only with a 1GB 4870 in place of the 1GB 5770. In these tests, the 1GB card offered 12%, 32% and 250% better performance at each of the three resolutions. The exact percentages can't be directly compared, since these are entirely different systems and different tests being performed, but it's obvious that in this game at these settings, 1GB of video memory offers a significant performance boost over 512MB, all else being equal.

In Tom's own tests comparing the 5770 against the 4870, where each card had equal video memory, the 4870 came out on top in Far Cry 2. It only performed around 1 to 2 percent better than the 5770 at each of those resolutions, but that's a 'far cry' from the 512MB version's loss in the tests at TechPowerUp. That's not to say the results where significantly affected for all tests performed there, but for those utilizing more than 512MB of video memory, they were skewed in favor of the 1GB cards. While Tom's didn't test the other few games where the 5770 managed to pull ahead at TechPowerUp, it's highly likely that they faced a similar situation of filling all available memory on the 512MB cards as well. In short, you can't accurately compare a 1GB 4870 against a 1GB 5770 based on the tests at TechPowerUp, since they didn't test one of those cards. The card they did test is similar, but produces wildly different results under certain situations, which affect the overall results.

Anyway, as I mentioned before, the 5770 is certainly a good card, and one that provides similar performance to the 4870, while offering some new features and much lower power requirements. It probably wouldn't be hard to overclock the card to perform better than a 4870 while still retaining its low power draw and quiet operation either. : )

[alecmg]

I got the 5770 and quite satisfied with it, doing fair amount of gaming and happy with idle power. But if my case allowed, I would get 5850 with some headroom in performance and same good idle power.

Whereas to original poster problem, perhaps change of motherboard will be less costly? Get one with suitable arrangement of pci-e slots

[gcwebbyuk]

Graphics card is on hold for the time being as I bought a resorator, so thats cooling the card nicely at the mo

Will be buying one next month probably though, but will need to be a card that will work with the Zalman GPU and GPU RAM blocks.