Monday, November 16th 2009

Radeon HD 5970 Offers Massive Overclocking Headroom

AMD's dual-GPU flagship graphics accelerator, the Radeon HD 5970, is closer than you think it is. Slated for 18th Nov, it includes every feature that allows AMD to reclaim the performance leadership it yearned for since the beginning of this year. In a series of company slides sourced from XtremeSystems Forums, it is learned that this could be one of the first accelerators which AMD "openly" markets as having a "Massive Headroom" for overclocking. While the clock speeds on the HD 5970 are lower than those on the single-GPU HD 5870, AMD lifted limits on what the driver-level ATI Overdrive software can offer in terms of clock speeds. While the engine (core) and memory speeds are set at 720/1000 MHz, the unlocked ATI Overdrive lets users take the clock speeds all the way up to 1000/1500 MHz. That's 30% for the core, and a stellar 50% for the memory.

To back such speeds, AMD seems to have splurged heavily on top-notch components on the PCB. To begin with, the PCB holds two high-grade AMD Cypress GPUs, each with all its 1600 stream processors enabled. The GDDR5 memory, while clocked at 1000 MHz or 4 GT/s, is technically rated by its manufacturer to run at 1250 MHz or 5 GT/s. All systems are powered by high-grade digital PWM voltage regulators, with independent Volterra VRM controllers that allow real-time monitoring, and software voltage control. Barring the five-odd cylindrical solid-state capacitors, Japanese pure ceramic surface-mount capacitors are extensively made use of.

The card's cooling assembly isn't any less descriptive either. It consists of a back-plate that cools memory chips on the reverse side of the PCB, while its obverse side is cooled by a large, consistent vapor-chamber plate, which covers the main components such as GPUs, the PCI-Express bridge chip, and the VRM chips. This is a design change compared to the R700 and R680, in which each GPU had its own heatsink, and one of the two would end up with second-hand (pre-heated) air from the other. Instead, the vapor-chamber plate conveys heat directly to a large, monolithic heatsink, which from the looks of it, features aluminum-fabbed air-channels. AMD's workhorse leaf-blower is still around, though this time, it is controlled by an SMSC EMC-2103 multi-point programmable PWM fan controller. The specs sheet shows the card's idle and maximum board power draws to be 42W and 294W, respectively.


Source: XtremeSystems Forums
Add your own comment

90 Comments on Radeon HD 5970 Offers Massive Overclocking Headroom

#1
TheMailMan78
Big Member
You know what would be more awesome than this?! If there were actual 58XX series cards on the F*#KING market to be bought!
Posted on Reply
#2

Nice! I'm curious about nVIDIA response. When it will come... Next year, or next millennium...:banghead:
#3
btarunr
Editor & Senior Moderator
by: Kenshai
Btarunr said NDA lifts on the 19th, so I assume the card will be available around then as well at the new drivers.
Make that 18th :)
Posted on Reply
#4
Izliecies
Hey, I thought that videocards are OCed with ATI Catalyst, not AMD OverDrive! Isn't that right?
Posted on Reply
#5
TheMailMan78
Big Member
by: Izliecies
Hey, I thought that videocards are OCed with ATI Catalyst, not AMD OverDrive! Isn't that right?
Both from what I remember. Ether or.
Posted on Reply
#6
Imsochobo
Amd clock tool is VERY reccomended download.

You can monitor min, avg, max temp of EACH core, and clock.

Its from AMD, and is widely available.

Then set your clocks in CCC :)
Posted on Reply
#7
HossHuge
Anyone know how long this it is?
Posted on Reply
#8
FreedomEclipse
~Technological Technocrat~
If the price is good - I might be getting one of these instead of a 5870
Posted on Reply
#9
sideeffect
by: W1zzard
those numbers are no hard limit.. given a decent psu (any) you can easily overdraw the external power connectors without any harm. 150w from a single 6 pin is no problem.
if an intel cpu is specced to 3 ghz and you can run it at 4.5g you dont ask "where do the mhz come from?"
No offence but that analogy sucks :)

The extra MHz on the CPU comes from giving it more power. You could perhaps lower the Vcore and add MHz and keep at the same power draw but you are not getting something for nothing. You are limited by the amount of power + how efficiently the device can utilise the available power.

You can draw 150w from a single 6 pin no problem? Evidence please because from what I have seen with my ATI 4850 which had 1 x 6 pin this was not the case and it was power starved. I was using a 1KW OCZ ProXtreme to power it as well.
Posted on Reply
#10
TheMailMan78
Big Member
by: sideeffect
No offence but that analogy sucks :)

The extra MHz on the CPU comes from giving it more power. You could perhaps lower the Vcore and add MHz and keep at the same power draw but you are not getting something for nothing. You are limited by the amount of power + how efficiently the device can utilise the available power.

You can draw 150w from a single 6 pin no problem? Evidence please because from what I have seen with my ATI 4850 which had 1 x 6 pin this was not the case and it was power starved. I was using a 1KW OCZ ProXtreme to power it as well.
:laugh: Oh man you just pwnd yourself nasty!
Posted on Reply
#11
sideeffect
by: TheMailMan78
:laugh: Oh man you just pwnd yourself nasty!
Do you even know what we were talking about?
Posted on Reply
#12
mechtech
by: HossHuge
Anyone know how long this it is?
loooonnnnngggg lol
Posted on Reply
#13
AddSub
by: sideeffect
I will believe it when I see it.
Same here.

Source? "XtremeSystems Forums"? I'm reserving my judgment for after this thing is benchmarked by some respected sites like iXBT Labs or similar.
Posted on Reply
#14
TheMailMan78
Big Member
by: sideeffect
Do you even know what we were talking about?
Yeah I know what your talking about. I also know W1zz is the owner of the website and has a vast amount of experience and wouldn't be talking out of his ass. I am far more inclined to take his word over yours.

Saying his "analogy sucks" isnt the smartest thing to do.
Posted on Reply
#15
phanbuey
by: TheMailMan78
Yeah I know what your talking about. I also know W1zz is the owner of the website and has a vast amount of experience and wouldn't be talking out of his ass. I am far more inclined to take his word over yours.

Saying his "analogy sucks" isnt the smartest thing to do.
+1 to that...

Sideeffect, Just how was that 4850 was "power starved"? did you volt mod it? or just because you couldn't overclock it past whatever MHz you assumed that the chip was not getting enough power?

Ive seen 4pin cpu plugs melt with too much power going through them and I can attest to the fact that a plug will give as much power to a device as it draws until the PSU doesnt give it anymore/plug melts...

Maybe your PSU had a hard limit on the 6 pin of 75W... iDK, I know mine doesn't.
Posted on Reply
#16
Zubasa
by: Imsochobo
Amd clock tool is VERY reccomended download.

You can monitor min, avg, max temp of EACH core, and clock.

Its from AMD, and is widely available.

Then set your clocks in CCC :)
If there is no "OC limit" in CCC, I might as well use MSI Afterburner which is the tool of choice for me. :toast:
Posted on Reply
#17
wolf
Performance Enthusiast
by: Zubasa
If there is no "OC limit" in CCC, I might as well use MSI Afterburner which is the tool of choice for me. :toast:
Seconded, if there is voltage control, clocks up to 1ghz/6ghz, then Afterburner is already perfect to take care of that. tool of choice f'sho'
Posted on Reply
#18
W1zzard
by: sideeffect
No offence but that analogy sucks :)

The extra MHz on the CPU comes from giving it more power. You could perhaps lower the Vcore and add MHz and keep at the same power draw but you are not getting something for nothing. You are limited by the amount of power + how efficiently the device can utilise the available power.

You can draw 150w from a single 6 pin no problem? Evidence please because from what I have seen with my ATI 4850 which had 1 x 6 pin this was not the case and it was power starved. I was using a 1KW OCZ ProXtreme to power it as well.
power starved doesn't even exist .. power is a function of current and voltage, when you increase the current drawn, the voltage at your endpoint drops due to the resistance in the wire. if you draw too much power either the voltage drops below a value that the card's voltage regulations circuitry need to work reliable, or your wires will heat up and burn. if you want more detailed physics formulas let me know.

when i have a few minutes i'll try to power tomorrow's card through a single 6-pin molex
Posted on Reply
#19
TheMailMan78
Big Member
by: W1zzard
power starved doesn't even exist .. power is a function of current and voltage, when you increase the current drawn, the voltage at your endpoint drops due to the resistance in the wire. if you draw too much power either the voltage drops below a value that the card's voltage regulations circuitry need to work reliable, or your wires will heat up and burn. if you want more detailed physics formulas let me know.

when i have a few minutes i'll try to power tomorrow's card through a single 6-pin molex
Exactly. If the 4850 was truly "starved" for power it wouldn't run reliable OR wouldn't work at all. Also adding voltage can have the reverse effect on circuits. Maybe THIS is why it only had one PCI-E connector? :D

If your PCI-E power cord wasnt supplying enough power I would look at your PSU. Also it may have been supplying the power but the power may have been "dirty". Watch them ripples son ;)
Posted on Reply
#20
[I.R.A]_FBi
by: TheMailMan78
Exactly. If the 4850 was truly "starved" for power it wouldn't run reliable OR wouldn't work at all. Also adding voltage can have the reverse effect on circuits. Maybe THIS is why it only had one PCI-E connector? :D

If your PCI-E power cord wasnt supplying enough power I would look at your PSU. Also it may have been supplying the power but the power may have been "dirty". Watch them ripples son ;)
oh, ripples :(
Posted on Reply
#21
W1zzard
by: TheMailMan78
Exactly. If the 4850 was truly "starved" for power it wouldn't run reliable OR wouldn't work at all. Also adding voltage can have the reverse effect on circuits. Maybe THIS is why it only had one PCI-E connector? :D

If your PCI-E power cord wasnt supplying enough power I would look at your PSU. Also it may have been supplying the power but the power may have been "dirty". Watch them ripples son ;)
adding extra connectors does not increase the voltage (parallel circuit). all it does is add more wire diameter which reduces resistance which allows more current per °C temp increase which allows more current to flow before your cables melt
Posted on Reply
#22
TheMailMan78
Big Member
by: W1zzard
adding extra connectors does not increase the voltage (parallel circuit). all it does is add more wire diameter which reduces resistance which allows more current per °C temp increase which allows more current to flow before your cables melt
It safety increases the flow of power correct? As each connection is rated to a maximum voltage/current?

What I meant by "only having one PCI-E" connector wasn't to increase voltage but to maintain the integrity of the power connection.
Posted on Reply
#23
W1zzard
yes, it is specified to a maximum current (google for awg 18), specified means you can run it outside of spec, a realistic limit is when the cables get hot to the touch, the physical limit is when the insulation starts to burn or melt->the wire touches some ground bzzzt
Posted on Reply
#24
IceCreamBarr
Does anyone find it odd that with the total power from the pci-e slot (75W), the 6 pin (75w) and the 8 pin (150w) that there is only an additional 6w left for overclocking (using their 294w max)? I'm not sure what kind of overclock 6w will bring you but then again, does taking the card to the max power draw even sound intelligent? By design, this card sounds like an overclocking dud.

Barr

EDIT: Oh, I see this is already being discussed - I guess reading only the first page leads to dupe posts :)
Posted on Reply
#25
Robert-The-Rambler
As Far As Clock Speed Goes

I think it is a way to save a whole bunch of money and avoid unnecessary warranty claims. The card already has 3200 shaders and whether or not AMD/ATI clocked the chip higher would not offer significant gains to almost any gamer at any resolution. I believe it is a logical business decision based on simple math. The card is so damn fast that if they underclock it a bit to avoid almost any cooling issues and power consumption issues then offer headroom for tweakers who realize they can't RMA it if it explodes then it is a win situation for everybody. :toast:
Posted on Reply
Add your own comment