Anand's 5830 Review!! Benches and all!! Anand is disappoint with this card!!

[heyheybooboo]

Gosh, you video card guys are tough.

It looks to me that AMD dropped this card at a specific price point midway between the 5770 and 5850 (and what remains of the GTX 2XXs).

As the GT100s come on the market AMD is clearly prepared to position the product line to compete on price.

If Fermi variants blow, AMD leaves nothing on the table. If GT100s are stout AMD can drop prices $20-60 across the range.

It is tough on AMD to have to compete with themselves (and what remains of the GTX 2XXs).





--

 

[dzoner]

Buzz killing prices = questionable marketing IMO.

Putting short term profits ahead of keeping the positive buzz growing and expanding ... it's always a trade off when a competitor stumbles badly ... but a slightly sour taste is starting to grow in my mouth when I think Radeon. I think AMD is messing up a golden longer range marketing opportunity to fully supplant Nvidia.

Why generate ANY negative name buzz at this crucial marketing juncture.

 

[dguy6789]

How puzzling. The 5830 has 4Ghz 256 bit memory, the 4890 has 3.9Ghz 256 bit memory. 5830 has more memory bandwidth. The 5830 has 1120 stream processors at 800Mhz. The 4890 has 800 stream processors at 850Mhz. 1120x800 = 896,000. 800x850 = 680,000. 5830 has more general GPU power. The ONLY thing the 4890 has over the 5830 is 16 rops at 850Mhz versus the 5830's 16 rops at 800Mhz. Given the above specs, there should be virtually no scenario where the 4890 outperforms the 5830. Are higher clocked rops really worth that much?

This means that either 1. Overclock the 5830 by 50Mhz and it will outperform the 4890 in 100% of all scenarios, or 2. This card is held back by drivers, possibly deliberately.

 

[n7]

I am disappoint.

I am disappoint too.

Makes the <$200 CND i spent on my HD 4890 so long ago seem like an amazing deal, which it apparently was.

 

[exar333]

When this card was annaounced a few months back, I said the price should be $180-200. That seems to be about right for the performance. Very disapointing performance for the specs, there has to be something artifically limiting the performance. I just can't see more specs (and 50mhz less core) being ~20% slower than the previous gen.

 

[thilanliyan]

This gpu actually costs them MORE than a 5850 if they're using a 5870 pcb for it

Apparently ATI sent reviewers 5830 cards with 5870 pcbs so they OC better but retail cards will most likely not have that. It was mentioned in the hardwarecanucks review.

 

[Piotrsama]

I read another review and they did non AA tests and here it beats the 4890. They mentioned that the 5830 hurts from AA more than any other card...

Curious.
Based on actual price and performance, this should've been called 5810.

Hopefully future drivers improve performance, or AMD reduces the price. (or a bit of both).

 

[RussianSensation]

It's got 320 (40&#37😉 more shaders than the 4890 (and only 50Mhz, or 6%, slower), 16 (40%) more texture units, same number of ROPs, slightly faster memory, and is on the 40nm manufacturing process (vs. 55nm for the 4890).

Ok, but 5850's performance advantage over 4890 also isn't as pronounced as its specifications would lead you to believe. The bottom line is that it appears that HD4000 series is more efficient per clock than HD5000. This is why 5870 is not 2x faster than 4870 is (and it has been shown many times that memory bandwidth doesn't really starve the HD5870 either).

dguy6789, I doubt drivers will increase 5830s performance that much more since it's essentially the same architecture as 4000 series is. So driver increases will affect both generations. Looks like this round of gen wars is a *yawn* until Fermi arrives. Hopefully, it will finally put the pressure on prices.

 

[SlowSpyder]

How puzzling. The 5830 has 4Ghz 256 bit memory, the 4890 has 3.9Ghz 256 bit memory. 5830 has more memory bandwidth. The 5830 has 1120 stream processors at 800Mhz. The 4890 has 800 stream processors at 850Mhz. 1120x800 = 896,000. 800x850 = 680,000. 5830 has more general GPU power. The ONLY thing the 4890 has over the 5830 is 16 rops at 850Mhz versus the 5830's 16 rops at 800Mhz. Given the above specs, there should be virtually no scenario where the 4890 outperforms the 5830. Are higher clocked rops really worth that much?

This means that either 1. Overclock the 5830 by 50Mhz and it will outperform the 4890 in 100% of all scenarios, or 2. This card is held back by drivers, possibly deliberately.

I could be wrong on this, but I thought it was discussed to some degree in another thread that AMD had to remove some things from the core of the 5xxx cards to keep the die around the size they wanted. To make up for those removed pieces, the shaders had to do those jobs. So, the shaders may be doing additional work with the 5xxx cards compared to the 4xxx cards. I don't believe that they're weaker, just that there is more for them to handle.

 

[Stoneburner]

This card is technically the most 'expensive' for AMD to make. Once yields improve, this card will dissapear or perhaps be updated at some point, otherwise AMD probably doesn't see much problem with relegating this to the Geforce 4MX croud of buyers (hell I was almost one of them!) at present.

The cost of production should be irrelevant on such a product. Since these are basically failed chips AMD does not necessarily need to profit. By selling them for a lower price they'd be recouping some costs. They should not price it relative to the cost for producing each chip they should price it for what the market dictates in order to lose less money.

 

[SunnyD]

Something is obviously wrong with this sample set/review. As was mentioned, given all things, this card should be at least as fast as a 4890 if not faster. I'm guessing this might be a case of something turned off in the video BIOS, similar to what happened with the original 4830 reviews where they inadvertently turned off an extra set of shaders on the first batch of review cards.

 

[dguy6789]

I could be wrong on this, but I thought it was discussed to some degree in another thread that AMD had to remove some things from the core of the 5xxx cards to keep the die around the size they wanted. To make up for those removed pieces, the shaders had to do those jobs. So, the shaders may be doing additional work with the 5xxx cards compared to the 4xxx cards. I don't believe that they're weaker, just that there is more for them to handle.

That's interesting. I vaguely recall reading something about that a long time ago but it has mostly escaped my memory. That would definitely explain the card's performance.

As for people that say this card is too much money, just wait for Fermi then everything in ATI's lineup will probably drop in price. ATI is merely taking advantage of a lack of competition to sell their parts for a lot of money, the 5830 is no exception to this.

 

[Piotrsama]

Ok, but 5850's performance advantage over 4890 also isn't as pronounced as its specifications would lead you to believe. The bottom line is that it appears that HD4000 series is more efficient per clock than HD5000.

Or maybe the architecture doesn't scale that good after 800 shaders.

Something is obviously wrong with this sample set/review. As was mentioned, given all things, this card should be at least as fast as a 4890 if not faster. I'm guessing this might be a case of something turned off in the video BIOS, similar to what happened with the original 4830 reviews where they inadvertently turned off an extra set of shaders on the first batch of review cards.

Might be, good memory there, SunnyD.

 

[Kuzi]

Yes, the quote below is from Anand's 5450 review:

"With this card (5450) we can finally pin down something we couldn&#8217;t quite do with the 5770: clock-for-clock, the 5000-series is slower than the 4000-series.

This is especially evident on the 5450, where the 5450 has a 50MHz core speed advantage over the 4550, and yet with everything else being held equal it is still losing to the 4550 by upwards of 10&#37;."

http://www.anandtech.com/video/showdoc.aspx?i=3734&p=15

 

[dguy6789]

Yes, the quote below is from Anand's 5450 review:

"With this card (5450) we can finally pin down something we couldn’t quite do with the 5770: clock-for-clock, the 5000-series is slower than the 4000-series.

This is especially evident on the 5450, where the 5450 has a 50MHz core speed advantage over the 4550, and yet with everything else being held equal it is still losing to the 4550 by upwards of 10%."

http://www.anandtech.com/video/showdoc.aspx?i=3734&p=15

Nice work finding this article. This proves it. Slowspyder is correct.

 

[deimos3428]

Between 4870 and 4890 performance + DX11, Eyefinity, & lower idle power/temps.

Can't see anyone buying that for $240, but it'll be interesting to see how far it drops when Fermi arrives.

 

[Piotrsama]

Interesting data, Kuzi...
RussianSensation also said that, and he was right.

 

[Lonyo]

Apparently ATI sent reviewers 5830 cards with 5870 pcbs so they OC better but retail cards will most likely not have that. It was mentioned in the hardwarecanucks review.

http://www.xbitlabs.com/articles/video/display/radeon-hd5830_2.html

The XFX PCB is absolutely tiny compared to the 5870 that some others do seem to be using (it's certainly not required).
Almost looks like they sent an HD5770 or something with a restickered heatsink on it.

 

[hans030390]

Yes, the quote below is from Anand's 5450 review:

"With this card (5450) we can finally pin down something we couldn’t quite do with the 5770: clock-for-clock, the 5000-series is slower than the 4000-series.

This is especially evident on the 5450, where the 5450 has a 50MHz core speed advantage over the 4550, and yet with everything else being held equal it is still losing to the 4550 by upwards of 10%."

http://www.anandtech.com/video/showdoc.aspx?i=3734&p=15

Let's just hope this is more driver/software related rather than hardware...

Even if it is hardware, I'll still be happy with my 5770.

 

[Kuzi]

It's hardware changes ATI decided to do with the 5000 series to decrease the die size.

 

[pmv]

The 5850/5870 are not that expensive from what we're used to. With brand new top of the line cards, I can't remember the last time they were released at $300 or $400.

But they aren't 'brand new', are they? And 'top of the line' rather begs the question, that's sort-of the point - normally cards of that age with that performance gain over older cards would not be 'top of the line' at this point.

Put it this way, going from an x1950xt to a 8800gt gave me an improvement akin to going from the 8800gt to a 5850 would now, but cost less than a 5770 does now. Cards are just expensive at the moment, _and_ in short supply, _and_ on top of that there's the ATI GSOD to worry about (I'd be furious if I paid that kind of money for something that turned out to have a fault).

I'm curious as to whether this is all nvidia's fault (i.e. that AMD are resting on their laurels purely because there's no pressure on them to do better) or whether there's a more general issue, i.e. that AMD wouldn't be doing any better even if nvidia were still in the game, i.e. that there's a general bottleneck in card development now.

 

[dguy6789]

The 8800GTX launched at a higher price than the 5870. And it is likely that the 5870 will maintain its current "high" price for less time than the 8800GTX had kept its own.

 

[konakona]

how much were the 4890s again when they were at their lowest? sub $120 after rebate and all IIRC. Soooo should have grabbed one back then 🙁

 

[Xcobra]

I think people got spoiled with the HD4000s pricing. Just stop bringing that up. Just because you bought a 4890 for $175 doesn't mean you'll get the same performance per dollar next time around. Don't be greedy. Now that said, I am a bit disappointed about its performance. Hopefully its some kind of anomaly in drivers or something.

 

[konakona]

no I am not saying that at all. It just sucks there isn't something that fits my bill atm, but not really pointing finger at AMD or anything; they are a business entity and expected to do whatever to be profitable after all.