HD5870 and HD5850 Reviews

[RaZz!]

Anand, bit-tech, Hcanucks,....., and it can´t beat easy the 4870x2.
With same specs, better clocks, and without de handicap of crossfire, i expected better performance in all games. So, yes, the results with the drivers used is a bit dissapointing.

even though the benchmarks show a pretty strong performance i tend to agree, atleast for now. if you consider that the 5870 is basically a 4870x2 in one chip (without crossfire which adds some sort of overhead etc) i still can't really think of a reason why the 5870 isn't faster than a 4870x2 most of the time

i really hope performance increases with more mature drivers later on. or are we really looking at a bottleneck regarding the 256bit interface some people "predicted" prior the launch?

two 5870 in cf however look impressive. pretty good scaling considering the fact these use release drivers, especially compared to the scaling of the gts295 or two 285 in sli.

however, i'm really tempted to replace my 4850 with a 5870 - even if i don't know why, as everything i play atm runs fine on the 4850 i think it's the charm of new hardware and dx11 (including compute shader etc...)

[Wesker]

dont know why, but i am not that impressed. Will def. wait for nvidia's response.

Sums up my feelings, exactly.

The performance isn't bad or anything (unlike the R600, for example), but it just isn't the "WOW!" that I was expecting. Perhaps, my expectations were set too high. But with the specs being double of RV790 (with the exception of bandwidth), I was expecting at least 1.9-2x the performance of RV790. Instead the performance is just under that figure, on average. Perhaps it's the early drivers or the memory bandwidth holding it back--But then again, RV770 was never bandwidth limited...

I may perhaps wait for GT300 or Hemlock. This HD4890 is still strong.

[jmke]

or am I missing something here?

that very few games require more than 2gb ?
HD 5850 is listed at €200 in Europe... 88% performance of the Radeon HD 5870 AND it outperforms the GTX 285 easily;
if you don't have a 30" screen and don't play on triple monitor carnage, the HD 5850 seems to be the card to go; all the other cards in the €150-200 bracket will have to reduce their prices to remain valuable as an option; any card over €200 which is not faster than HD 5850... don't buy.

[saaya]

I wonder if 5870CF is going to be faster than the 5870X2?

first of all, there seem to be serious issues with 5800 xfire, so those numbers anandtech showed will def improve...
but overall, yes, 5870x2 will DEF be slower than 5870CF, always...
with 5870x2 your limited by power and heat, with 5870CF your not...

Anand, bit-tech, Hcanucks,....., and it can´t beat easy the 4870x2.
With same specs, better clocks, and without de handicap of crossfire, i expected better performance in all games. So, yes, the results with the drivers used is a bit dissapointing.

well, in some tests its 5% or 10% in the lead... but yes, its not clearly faster than the 4870x2... its more of a replacement for the 4870x2, and the fact that its single gpu makes it a good replacement even if it ONLY offers about the same perf... at least in my opinion... and xbit and anandtech seem to agree...

but yes, i also expected more... and im confused why the 5870 doesnt do better...

Unlike some people, I am of the opinion that history has shown that new drivers do not increase the performance of a single chip card unless there are issues to begin with. Dual GPU cards? Yes, new drivers definately increase performance in existing games.

Where single GPU cards get performance increases is with new games that hadn't been optimized in past drivers or fixing errors with performance in current games. One example of an issue is the HD 5870's HawX DX10.1 performance. I am sure that will be fixed in future revisions.

Otherwise, I will put money on the fact that the HD 5870 will never beat the HD 4870X2 on a regular basis even with driver revisions. There is a substantial bandwidth gap between the two making equalized performance next to impossible.

check the xbitlabs article about this, called driver scaling with 4870 iirc... from half a year back... they showed that in several games perf had improved around 10%... thats not much, but if youd compare 8.7 to 9.9 im sure youd see more than a 10% boost for a 4870... probably around 15-20% in several games... and not just av fps, min fps has increased as well...

its unusual for this to happen though, and its nowhere near the perf jumps dual gpu setups make with driver updates, def agree on that...

looking at all results of the reviews ive seen so far, id say 4870x2 is roughly equal to a 5870, and 5870 is single gpu so thats making it the winner...
even if it will never beat the 4870x2 accorss the board, i dont care, its single gpu and thats worth more than 5% perf imo... especially if you look at what fps numbers we are talking about here...

my current impression after having read several reviews is this:
5870: fails to beat 295 and 4870x2, but does about the same with only 1 gpu... price needs to be reduced to 300-350$ then itll be a very sweet deal

5850: notably beats 285 at almost 100$ lower price, good deal for people who dont have a 285 or faster atm...

im really curious about 5740 or whatever juniper is called?
4890 perf at 99$ with dx11?

saaya: It's hard to see you base your conclusions on a techpowerup review that used 6 gb of ram on a 32 bit system - or am I missing something here?

what does that have to do with anything
he probably has a 64bit partition but hes doing 32bit for now cause thats what he built his db with and he doesnt wanna mix 32 and 64bit...

that very few games require more than 2gb ?

require? 0
benefit? i only know of one, gta4, and that seems to only happen at mega res and in some scenarios...

HD 5850 is listed at €200 in Europe... 88% performance of the Radeon HD 5870 AND it outperforms the GTX 285 easily;
if you don't have a 30" screen and don't play on triple monitor carnage, the HD 5850 seems to be the card to go; all the other cards in the €150-200 bracket will have to reduce their prices to remain valuable as an option; any card over €200 which is not faster than HD 5850... don't buy.

IF you dont have a gtx285 already... and even if you have a 4890 id rather wait than upgrade... i mean really, what for? what can you play with a 5850 that you cant play with a 4890 or even 4870?

[jmke]

what can you play with a 5850 that you cant play with a 4890 or even 4870?

only 2, Crysis, Crysis Warhead, with high IQ

[Calmatory]

But with the specs being double of RV790 (with the exception of bandwidth), I was expecting at least 1.9-2x the performance of RV790. Instead the performance is just under that figure, on average.

Welcome to the real world. How could it possibly be "at least 2x" faster when NOT ANY of the specs go OVER 2x when compared to RV790?

Indeed, no way. You expected something which is not even possible, and now feel disappointed. Disappointed because of your overly-optimistic expectations.

Good luck next round, will be disappointed again, and again.

[SKYMTL]

check the xbitlabs article about this, called driver scaling with 4870 iirc... from half a year back... they showed that in several games perf had improved around 10%... thats not much, but if youd compare 8.7 to 9.9 im sure youd see more than a 10% boost for a 4870... probably around 15-20% in several games... and not just av fps, min fps has increased as well...

While I haven't read it, I am sure the games showing the most improvement are the ones that were either new releases or not even released when the HD 4870 was brought to market.

[kiwi]

you talk as if those days are over
what do you think the 5870x2 and gt300 are going to cost when they come out?

Back then there were no x2 solutions cause x2 is just a cheat Unless they drastically reduce power consumption.

With cpu cores they did that. More cores but +/- same power. But with GPUs they can't (yet). You get the point OK, going a bit offtopic here

[flippin_waffles]

Has anyone seen a review on anything other than an i7 cranked up to the absolute limit of the chip? Not sure about anyone else, but I don't have one, nor do I intend to get one. I'm also equally confident that not all users of i7 run them overclocked, in fact i'll bet that the vast majority don't. So while it's all good and nice to say that this is the only way to see what the potential of the card is, what about writing a review that caters to actual users? You know, with hardware settings that people can relate to. I don't give a siht how this card performs with an i7 trying to hit 4GHz, when I have a 955 running at stock, or a couple hundred MHz higher. How does all that do me any good if i'm trying to decide what kind of improvement i'll get in my system? How about some REAL WORLD performance guys? I keep seeing that bloody term being thrown about, and then see review after review trying to clock there cpu's to the stratosphere. Would some variety be so bad? Oh I know, "we are trying to eliminate the cpu bottleneck". Well most of us have a cpu bottneck! What kind of bizarre thinking is that?? Ok, sure it's nice to see the max potential of new hardware, but is it that critical that nobody can write a review that is helpful to the other 90% of the online population? Seems like kind of a no brainer to me. How about some stock settings guys?

Anyway, it's the same principal and failed logic that goes into deciding to review cpu's with game benchmarks at ultra low resolutions that would be run on 10 inch monitors 15 years ago. Why not just stick to cpu benchmarks to benchmark a cpu? JHC Who cares how fast a cpu will run a game at 640x480. Yeah yeah, I know, if you don't like it don't read it, or write your own review. But wouldn't a more practical approach smarter in the long term?

More to the point though, looks like this is a great card.

[Boissez]

source: http://www.madshrimps.be/vbulletin/f...e-chart-66465/

enjoy

Well either the 8800GT is 20% faster than the 9800GT... or the numbers are somehow wrong.

[kgk]

huh?
4870 and 4890 both oced well...
what were def clocks for 4890? 850mhz?
and they clocked to close to 1ghz on air... hows that bad ocing?

Close to?

They clocked well over 1Ghz. I have a Sapphire Toxic with a stock clock of 960mhz that easily runs 1030Mhz 24/7 on air. There are plenty of 4890's that do 1050Mhz on air 24/7 as well with good cooling.

Personally I'm not jumping to the 5000 series until we see a 5890 or better overclocks/cooling and more memory on the 5870. Almost always better to jump in after the revisions. Did the same when the 4870 was released. The initial 512MB cards were a joke, although quickly replaced with the 1GB versions. Still glad I held out for the 4890's, especially considering the cheap price and the performance jump.

The lack of any significant DX11 titles until next year, at which time we'll see a mature DX11 card selection make the purchase of the current 5800 lineup rather pointless for anyone with a decent card now.

All in all, still a very impressive offering. But I'm not jumping in until I see a top title like BF3, Rage, etc. that can bury a 4890 or GTX285 @ 1920x1200 or 2560x1600.

[kgk]

only 2, Crysis, Crysis Warhead, with high IQ

I have yet to meet a single person who actually plays these titles. Everyone tried Crysis when it came out to see the new graphics. Then we all moved back to playing our multiplayer games like CoD, TF2, and L4D.

[Wesker]

Welcome to the real world. How could it possibly be "at least 2x" faster when NOT ANY of the specs go OVER 2x when compared to RV790?

Indeed, no way. You expected something which is not even possible, and now feel disappointed. Disappointed because of your overly-optimistic expectations.

Good luck next round, will be disappointed again, and again.

I was expecting further improvements efficiency. Not only does it share similar paper specifications as R700, but it also share's RV790's core clock speed.

R600 -> RV770 was impossible? RV770's performance over R600 was greater than the 2.5x increase in the ALU count. Most of RV770's performance wins over R600 came from internal improvements in the chip design (most notable being the inclusion of hardware based MSAA resolve).

Years ago, were you also around saying that the performance of chips like Conroe, Hammer, Nehalem, R580, R300 and G80 were impossible?

...oh, and thanks for the attitude.

[kgk]

require? 0
benefit? i only know of one, gta4, and that seems to only happen at mega res and in some scenarios...

It starts to become noticeable at 1920x1200 with everything maxed. That is not a mega res. That's a very common gaming resolution today. At 2560x1600 it becomes a very noticeable problem in GTAIV. It's only going to become more noticeable with newer games that push the envelope further. So if I were to pick up a 5870, I'd much prefer it to have 2GB of memory.

IF you dont have a gtx285 already... and even if you have a 4890 id rather wait than upgrade... i mean really, what for? what can you play with a 5850 that you cant play with a 4890 or even 4870?

Exactly. Even if I were on a 3870 or 9800gtx, I'd still wait to see what's available in the spring. Then we'll actually have some selection of good games that will take advantage of DX11, more cards to choose from, and better prices thanks to increased competition.

[SKYMTL]

Has anyone seen a review on anything other than an i7 cranked up to the absolute limit of the chip?

I can get mine to 4.6Ghz but in-game stability gets thrown out the window and stability is what's needed for benchmarking.

[Mech0z]

i doubt 2gb will be faster, why should it? its not any faster for current gpus...
and i hope we will see better 3rd party heatsink soon as well...
?

That might be my next card

http://www.fudzilla.com/content/view/15648/1/

[flopper]

Has anyone seen a review on anything other than an i7 cranked up to the absolute limit of the chip?
More to the point though, looks like this is a great card.

sami ran the single-GPU X11364 score with CPU aircooled at 4GHz and a single 5870 at 1250MHz engine and 1390MHz memory.
http://www.legitreviews.com/article/1080/15/

always someone pushing stuff....

[jmke]

I have yet to meet a single person who actually plays these titles. Everyone tried Crysis when it came out to see the new graphics. Then we all moved back to playing our multiplayer games like CoD, TF2, and L4D.

well you met me, so now you can no longer claim that
I've played through Crysis several times now;
Crysis Warhead was slightly less entertaining;

[jmke]

Well either the 8800GT is 20% faster than the 9800GT... or the numbers are somehow wrong.

very few samples of 8800 GT (only from Anand in comparison TO the HD 5870)
see the numbers always in reference to the HD 5870 results, as that card was ALWAYS available in all charts, however not all cards were in all result charts; the 8800 GT popped up, and I did find it useful to include as this card was quite popular


----------

I've not gone through all reviews out there, anybody encounter one where they actually run two/three monitors ?

[Boissez]

very few samples of 8800 GT (only from Anand in comparison TO the HD 5870)
see the numbers always in reference to the HD 5870 results, as that card was ALWAYS available in all charts, however not all cards were in all result charts; the 8800 GT popped up, and I did find it useful to include as this card was quite popular

I see - Wouldn't it be better though to merge the 8800 and 9800 numbers then? they're basicly the same cards anyway. Anyways many thanks for compiling this list - saves lazy guys like me a lot of time riffling through reviews

[flippin_waffles]

I can get mine to 4.6Ghz but in-game stability gets thrown out the window and stability is what's needed for benchmarking.

I'm just saying, in a long winded way, what can I expect if i'm running a Phenom II 955 or 945 at stock, if I upgrade to this GPU? Or an Inel CPU at stock for that matter. Obviously it'll be faster, but then everyone already knew that. How much faster, is the benefit worth it? I haven't seen a review that answers that question. You'd think reviewers would clue in to the fact that not all people reading reviews, by a long shot, are users that run there equipment out of spec. So given the 20+ reviews out there, if you had a review that catered to the other 90% of the population, which one of the 20+ would get more views when those people search for answers?
And personally, i'd also like to see which platform is more efficient.

[strange|ife]

will probably be getting the 5850 after prices come down.

4870 512meg is decent, but struggles @ 1920 in some games

[Olivon]

OTH roxxx !

Thanks again

[Jamesrt2004]

I can get mine to 4.6Ghz but in-game stability gets thrown out the window and stability is what's needed for benchmarking.

meh you know what he means.. compare the percentage of people with normal C2D and C2Q's etc at normal 2.4~3.2 ghz compared to people with i7 @ like 4.2 lol

I like him and im sure a lot of other people would like to see results on say a Q6600 @ 3.2 or 3.6 or something a little more.. realistic for a lot of us

[HuffPCair]

wow very impressed. Seems like a great card from what 3Dguru is saying. Can't wait to see some XSers throw out some benchmarks.