ATI HD4800 Review Thread

[Truckchase!]

The question now is how does the added bandwidth of the GDDR5 directly affect performance? Based off the initial reviews I've read, it doesn't look to be as much of a factor as I had suspected it would be based off of the relatively low performance gaps at high resolutions....

It seems that without AA, the gap is larger, showing the increase in bandwidth. As soon as you introduce AA, however, the gap shrinks as the bottleneck shifts back to math.

Keep in mind that 4870's default clock is 20% higher than the 50s... Thoughts?

[Miwo]

jimmyz will be testing more with the HD4850 @ 750MHz, so stay tuned on that. Check his thread in the BIOS modification section. I'm curious of the same thing as well.
keep in mind that the HD4870 has much beefier vregs on it that will make it better suited for higher clocks and higher volts. Higher core speeds might love the extra bandwidth once you start pushing the core, but we really can't tell for certain until the software tools are updated.

[SKYMTL]

Keep in mind that 4870's default clock is 20% higher than the 50s... Thoughts?

In some situations as you mentioned the memory bandwidth comes into play and you see those huge performance gaps between it and the HD4850. However, we have to remember that an increase of 20% in the core speed does not equate a 20% increase in performance. That is why I think the bandwidth comes into play more than we may have originally thought since this card beats the HD4850 by 20% AND MORE in many cases.

[sephiroth8748]

dang the 4870 rocks. now I just got to wait to buy one...

[Truckchase!]

jimmyz will be testing more with the HD4850 @ 750MHz, so stay tuned on that. Check his thread in the BIOS modification section. I'm curious of the same thing as well.
keep in mind that the HD4870 has much beefier vregs on it that will make it better suited for higher clocks and higher volts. Higher core speeds might love the extra bandwidth once you start pushing the core, but we really can't tell for certain until the software tools are updated.

Good point... I've got a 4850 coming tomorrow and I'm debating not even taking it out of the box now; I may have to make that run to Microcenter.

In some situations as you mentioned the memory bandwidth comes into play and you see those huge performance gaps between it and the HD4850. However, we have to remember that an increase of 20% in the core speed does not equate a 20% increase in performance. That is why I think the bandwidth comes into play more than we may have originally thought since this card beats the HD4850 by 20% AND MORE in many cases.

Yeah that's what I was getting at... hopefully I can get my hands on one of these cards and pull one of these numbers on it:

http://www.xtremesystems.org/forums/...5&postcount=31

Awesome site btw, (hwcanucks) it's on the regular rotation now....

[SKYMTL]

Thanks for the compliment. This is what we get for 24 hours of reviewing since I went to AMD personally yesterday to pick up the card or it would have never got to me in time....

[grimREEFER]

well, i looked at the 4870 specs and assumed that it was gonna perform like a 4850.
i was wrong, and i think im gonna buy a visiontek 4870 now.
has anyone overclocked the card yet?

[Miwo]

HR-03GT is a great cooler
I'll be sticking to my Accelero S1 though, since I can cover the entire card with airflow much easier on that. Less heat for those oh-so-toasty vregs

[Tonucci]

Wow, the higher clocks made it much faster than i thought.

[Dragy2k]

sky .. flipping that HR03 around the other way......it still fit ok yeah ??

[AriciU]

Finally. I love the CF performance shown by Anand

[iTravis]

So which cooler should I go for? HR-03GT or the Accelero S1?

[adpr_02]

Is it me, or is Anand WAY off? If those results are real, Nvidia is about to get destroyed.

[GoldenTiger]

http://arstechnica.com/reviews/hardw...ies-review.ars

Woah... it performs within ~10% of the 280, and sometimes has HIGHER minimum framerates which I consider more important than a bloated average with spikes from some high maximums, so long as the average remains good (which it does). This review solely used AA/AF, which are what I and most other gamers play with (no aa/af is pretty much the benchers' domain).

So which cooler should I go for? HR-03GT or the Accelero S1?

Don't know, I'm trying to decide that too !

SKYMTL, is it the S2 or the S1 rev. 2 that fits on the card? On SVC.com I only see those two options, so I'm not sure if there was a typo in the review on the page about non-stock cooling with the HR-03 GT? I want to clarify as they are much cheaper than the HR-03 GT and I'd like to buy one right away for my incoming 4870 .

P.S. That review was incredible, your work really showed .

[Battle_Rattle]

I just wanted to post this little pic since I think it is oh-so-sexy:

Canucks says the old 3870 Thermalright HR-03's FIT!!!! Yeh!!!!!!

Welcome to quick and faster overclocking!!

[eleeter]

Canucks says the old 3870 Thermalright HR-03's FIT!!!! Yeh!!!!!!

Sweet. Add another stock cooler to my growing pile lol.

[SKYMTL]

sky .. flipping that HR03 around the other way......it still fit ok yeah ??

May bump into the inductors at the back of the card so I will have to check.

[IvanAndreevich]

Looks like the Inquirer was right this time around.

[AriciU]

May bump into the inductors at the back of the card so I will have to check.

Please do. It's the only way two of them would fit in a crossfire setup.

[SKYMTL]

SKYMTL, is it the S2 or the S1 rev. 2 that fits on the card? On SVC.com I only see those two options, so I'm not sure if there was a typo in the review on the page about non-stock cooling with the HR-03 GT? I want to clarify as they are much cheaper than the HR-03 GT and I'd like to buy one right away for my incoming 4870 .

36 hours without sleep will do that to ya. S1 Rev.2. It may not have th appropriate mount of VRM sinks though.

[GoldenTiger]

Thanks Sky!

EDIT: Apparrently because SVC doesn't carry the turbo module as well, it would be a waste to buy the S1 Rev. 2 from them and the turbo module from newegg due to shipping. So, it's really a $13 difference total for the HR-03 GT + a fan, and the S1 Rev. 2 + the turbo module for me. I like that the S1 Rev. 2 is smaller in terms of using only 3 slots total instead of the HR-03 GT's 4 slot profile, but would the HR-03 GT outperform it by much? It doesn't look to but not sure if I'm missing something . I have a few random spare RAMsinks I could use to cover up any extra VRMs that would need to be cooled.

[Spawne32]

sky that review is down right awesome, and the addition of the HR-03 test was fantastic

[AriciU]

All that was missing was two 4870's in crossfire, to put the hurtin on the GTX 280 at the same/lower price I'm sure it'll come. Would've been interesting too as you clocked the CPU at ~3.9ghz vs 3.2ghz Anand.

[Jowy Atreides]

very nice temps with an aftermarket cooler idle+load down significantly

http://img262.imageshack.us/img262/2...d487084fh1.jpg

.........wow

is the stock cooler really that bad?
or is it that monkeys are applying the thermal compound/pads?

[GoldenTiger]

.........wow

is the stock cooler really that bad?
or is it that monkeys are applying the thermal compound/pads?

Kinda wondering that myself... not sure whether just re-applying the thermal compound would help those numbers a lot, or if an aftermarket really is the way to go for this card (i.e. truly a huge difference)...