HD5970 Reviews

[Monstru]

Well, mine goes up to 810Mhz GPU with stock volts, I guess that is like 10%. Anyway, do the same experiment with 4870X2, GTX 295, GTX 280, 9800GX2, 3870X2, 8800 Ultra, etc. How much do they OC with the stock bios and stock volts (no Asus bios with higher latencies for HD 4870X2, no extra volts, no nothing). What do you think that is the average OC % on those, stock?

Then, bump everything in all of them (volts, special bios, fan to 100%, vmod, etc), but keep the stock cooler and the regular air temp, and see how much do they OC in %.

That si what I am talking about

[W1zzard]

4870 x2 = 6.5%, gtx 295 = 20%, gtx 280 = 13%, 9800 gx2 = 18%, 3870 x2 = 12%

[D749]

overvolt tool at tpu downloads section

Thanks W1zzard.

I have three retail HD 5970 cards sitting here and not one of them will OC more than 5% without a SIGNIFICANT bump to the voltages. One of them is even slightly unstable at reference speeds.

From my experience the overclocking potential with stock voltage is extremely limited. Much more so than any other high-end card released in the last few years.

Three? I'll PM you my address.

[Monstru]

Thx W1zzard, so we have an average ~ 13%, so the HD5970 comes a little bit under that when it comes to stock volts (5 % HC's case and ~10 % our case - I tested 2 retail and 2 BBA's). Now, what about that "bios/volts/all you can throw except stock cooler" part? 'Cause there is where the 5970 shines

[W1zzard]

every card will shine if you give it +25% voltage. retail and bba, all cards are the same, same factory, same production run, technically bba is not even built by ati

[Monstru]

A....no, it won't

I worked pretty hard with GTX 295 or HD 4870X2....Yeah, the GPU's would sure shine, and they do as long as you give them a little bit of cooler air. But with regular room temps, the GTX 295 stops at one point because the cooling can't keep up. At one point, it can't take that voltage anymore, and it is normal. You can't give it 25% more volts with the stock cooler and normal temps, that's the whole ideea here. The 5970's cooler allows a little bit more headroom on that part. And yes, once you put them under LN2 the things get "back to normal"

And no, BBA and retail is not allways the same thing. The "BBA" HD 5870 I have has much more extra parts on it compared to the retail versions. It's no "secret plan" as it overclocks worse then the other cards I have, but it is obviously an earlier sample in the production line.

[jaredpace]

I noticed that some of the vapor chambers on the 5970 are painted black with copper showing only where the dies touch. But, on some of them copper is visible on the entire surface underneath. Just a small change in production runs I guess

[initialised]

Well in that case yes sire, you are right. BUT, that has allways been the case with high-end cards and PSU's built on a ship between Taiwan and China. A PSU like that pops with GTX 295, 4870X2 too

Well the last 850W unit I was testing didn't go bang when I tried it with 3.85GHz i7 and 4870x2/GTX295/285/5870. it just dipped down below 11.4V and politely reset the machine so that's a slight improvement. The same crap with over-current protection. But the number of people who will try (and hopefully fail) to RMA their cards as faulty when this happens to their shiny new system can only hurt their availability and costs, it'll be like the HP i7 fiasco all over again for system integrators that don't do their testing right.

[onethreehill]

ATI reveals HD 5970 overlocking details
ATI has revealed a critical engineering decision that led to the successful production of its Radeon HD 5970 graphics card.

"When we were first designing the 5970, we came to a crossroads surrounding the board's power consumption. We had two major options available to us," explained Hemlock product manager Devon Nekechuk.

"On one hand, we could cap ourselves at 300W and stay within the PCI Express specification. In doing so, we would stay compatible with the majority of the high-end power supplies in the market with 6-pin and 8-pin power connectors. On the other hand, we had the potential to fill a 400W power budget with performance, so we were considering using two 8-pin connectors and pushing this thing to its limits."

http://www.tgdaily.com/hardware-feat...ocking-details

[Splave]

so do you think we will see a dual 8 pin version soon?

[W1zzard]

so do you think we will see a dual 8 pin version soon?

amd says not before q2 2010

[SocketMan]

However, having reviews published using a tool that ATI didn't intend to be available to the general public leaves a bad taste in my mouth.

Why would a company "provide" a tool capable of frying it's flagship product - to the "general public"?
People who really needed - will find it.General public is safer
with out it imo.
I mean you don't see family minivans that come with nitrous oxide as a standard option - it's just bad for business.
Could be something simple like not have a disclaimer pop-up about voiding the warranty if the tool is used.

Besides it's not like without this tool - certain features that you pay for - get disabled (after the fact) when/if
"öther" hardware is present (cough,cough,coughhhhphysX)


As for all the reviews (no matter how great/detailed they are) - none are complete without the actual DX11 games.

edit*
Can someone please tell the ATI guys that AMD also "makes" (designs) CPUs,and a very few people out there are interested how these new cards would perform on systems other then i7 920.In other words send a couple to the sites that would run these (cards) on Phenom/Athlon 2''s.

[Hell Hound]

Wait for x6 then we can compare x4 to x6 amd only platform.What I want to know is why are they not using 8x aa when available,also why 32bit os with 6gb mem.

[Helloworld_98]

Wait for x6 then we can compare x4 to x6 amd only platform.What I want to know is why are they not using 8x aa when available,also why 32bit os with 6gb mem.

they don't use 8xAA because they know the nvidia cards will get owned

for the 32b OS, they could have done the registry hack which allows it to use more than that.

[SKYMTL]

Why would a company "provide" a tool capable of frying it's flagship product - to the "general public"?

I am not saying they SHOULD. What I was saying is that AMD provided it to reviewers, reviewers used it in their articles and based part of their conclusions off of software that would not be available to the general public. Indeed, ATI dedicated a number of marketing slides to Hemlock's overclocking ability but the only way one can achieve those overclocks is to bump voltage into the chips.

[Behemot]

so do you think we will see a dual 8 pin version soon?

I personally don't understand this pin kind of thing. The best PSU's are that with one 12V rail. But I think most of the multi-rail PSU's have all wires in single PCIe power connector solded to the same 12V rail. So why is everybody messing with a milion-wire-PCIe power connector? All the wires are solded together on the PCB of the card, though. Really really stupid specification for me. If I had even 2-pin connector, I'd use dongle to 8-pin anyway. The vendors just have to be so intelligent to use enough thick wires. If they not, I'll change them.

[Solus Corvus]

I am not saying they SHOULD. What I was saying is that AMD provided it to reviewers, reviewers used it in their articles and based part of their conclusions off of software that would not be available to the general public. Indeed, ATI dedicated a number of marketing slides to Hemlock's overclocking ability but the only way one can achieve those overclocks is to bump voltage into the chips.

Just because they aren't providing end users with the official ATI made tool doesn't mean they aren't trying to provide a way for its users to tweak voltages. Since the warranty is up to each AIB, it's probably up to each one to decide if they will provide an "official" voltage tweak tool or not. Essentially it would be a legal CYA for ATI while still providing the functionality you'd need to extract as much possible from the card. The tgdaily article reinforces that impression:

"We've built all of this headroom into the board itself, and we're also doing some special stuff in the software as well. The bar in the ATI Overdrive utility goes higher than ever before, and we're working with some of our AIB partners on software that will unlock even more overclocking potential through voltage tweaking applications," added Nekechuk.

[Monstru]

I am not saying they SHOULD. What I was saying is that AMD provided it to reviewers, reviewers used it in their articles and based part of their conclusions off of software that would not be available to the general public. Indeed, ATI dedicated a number of marketing slides to Hemlock's overclocking ability but the only way one can achieve those overclocks is to bump voltage into the chips.

Andpeople who really overclock really mind this...The test with non-stock volts are for overclockers (at least from my point of view). No software tool, no problem, you can allways vmod the card, just like allways.

[Helloworld_98]

I personally don't understand this pin kind of thing. The best PSU's are that with one 12V rail. But I think most of the multi-rail PSU's have all wires in single PCIe power connector solded to the same 12V rail. So why is everybody messing with a milion-wire-PCIe power connector? All the wires are solded together on the PCB of the card, though. Really really stupid specification for me. If I had even 2-pin connector, I'd use dongle to 8-pin anyway. The vendors just have to be so intelligent to use enough thick wires. If they not, I'll change them.

it's because of heat, if you have all that energy going down one wire, it's going to get hot, and that means more resistance which means lower efficiency.

[Behemot]

it's because of heat, if you have all that energy going down one wire, it's going to get hot, and that means more resistance which means lower efficiency.

Thats why I said intelligent makers use thick wires. I know that some wires of some PSU's would melt a sec before the PSU would burn in fire

[Shadov]

Wait for x6 then we can compare x4 to x6 amd only platform.What I want to know is why are they not using 8x aa when available,also why 32bit os with 6gb mem.

Some AAx8 for you then... :-)

[safan80]

Some AAx8 for you then... :-)

lol
I would like to see more reviews use SLI 285s instead of the 295(which is a joke except for the asus mars). The one thing that is missing when crossfire 5870 is thrown in is SLI 285s and vice versa. Only a few reviews have both crossfire 5870s and SLI 285s. Reviewers are getting lazy and care more about getting the review out the door than having quality attachjed to their review.

[Hell Hound]

Thats still 4x aa

[STEvil]

I personally don't understand this pin kind of thing. The best PSU's are that with one 12V rail. But I think most of the multi-rail PSU's have all wires in single PCIe power connector solded to the same 12V rail. So why is everybody messing with a milion-wire-PCIe power connector? All the wires are solded together on the PCB of the card, though. Really really stupid specification for me. If I had even 2-pin connector, I'd use dongle to 8-pin anyway. The vendors just have to be so intelligent to use enough thick wires. If they not, I'll change them.

it's because of heat, if you have all that energy going down one wire, it's going to get hot, and that means more resistance which means lower efficiency.

Thats why I said intelligent makers use thick wires. I know that some wires of some PSU's would melt a sec before the PSU would burn in fire

Thickness of the wires doesnt really matter because the connectors are where most of the inefficiency (resistance) occurs.

That said most mutli-rail PSU's are in reality single rail units with just current limiting devices placed on each virtual rail.

[Sam_oslo]

Finally GPU-OCing is getting more like CPU-OCing, with adding more volts.

Everybody changes the stock cooler on the CPU to get a better OC, maybe we should get used to the idea of changing coolers on GPUs too.