Radeon HD 4670: A perfect balance?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
@Vicotnik: I wish you'll have sucess with RBE tweaking.
Btw, you have excellent power consumption. Is that measured from the wall? Sure PSU is helping a lot.
But how do you handle 3D load?
Did you measure how much you have without 4670, or how much HIS cards adds to the power consumption?
@Matija: No, I'd say that's because Sapphire has no voltage regulation programmed in Bios.
Btw, you have excellent power consumption. Is that measured from the wall? Sure PSU is helping a lot.
But how do you handle 3D load?
Did you measure how much you have without 4670, or how much HIS cards adds to the power consumption?
@Matija: No, I'd say that's because Sapphire has no voltage regulation programmed in Bios.
-
- Posts: 353
- Joined: Sat Oct 18, 2008 6:30 am
- Location: Moldova, exUSSR
3D load is no problem now when I can modify the bios using RBE. I'm experimenting with underclocking and undervolting the 3D mode to have some headroom.lechuck wrote:But how do you handle 3D load?
Did you measure how much you have without 4670, or how much HIS cards adds to the power consumption?
I haven't measured power consumption without the discreet graphics card.
I use a kill-a-watt type of meter that plugs into the wallKsanderash wrote:Vicotnik
How do you measure AC wattage? I personally don't have TrueRMS multimeter and I wonder how you guys are getting with AC power
-
- Posts: 10
- Joined: Sat Nov 29, 2008 11:31 pm
- Location: Finland
Hi,Kriz wrote: According to RBE 1.15...
HD4670 idle clocks: 165MHz GPU / 250MHz RAM / 0.900v
HD4670 full speed: 780MHz GPU / 1000MHz RAM / 1.250v
Now for some more power testing with the Seasonic SS-300SFD 300W PSU...
Using onboard GPU: 41W Idle / 69W Load (Prime95 Small FFTs)
Using HIS HD4670: 54W Idle / 81W Load (Prime95 Small FFTs)
So, it's still around 12-13W extra for my particular configuration.
More to come next week.
I have Sapphire Radeon 4670 512Mb. With default settings the card has quite high idle power usage, 20W.
4850e + 780g on idle 49W
4850e + 4670, 780g disabled = 69W
So, 4670 was +20W on idle, nowhere near the SPRC's tested +4W.
I made a profile, and was able to set the idle settings at:
GPU 165Mz, Memory 500Mhz, Voltage 0,9V. The idle power usage was 63w = +14W. (I cannot verify that the voltage was actually lowered )
So, it seems I cannot get better idle power by switching to HIS 4670, because your card seems to take also +13W idle.
Is there anyone who has a card that takes that +4W that SPCR tested? Or is this +4W vs. my +14W only a result of lower idle power usage of my 780g vs. SPCR's i945Ga? Main reason for me buying this card was SPCR's test...
-
- Posts: 353
- Joined: Sat Oct 18, 2008 6:30 am
- Location: Moldova, exUSSR
My Force3D 4670 idles at the same 13.3W (1.29V on GPU, 1.8V on RAM). Maybe (I'm again with this version) the card simply don't switch to VID that RBE is providing? Hm, I'll do some tests on this, to be shure.TheSilentOne wrote: So, it seems I cannot get better idle power by switching to HIS 4670, because your card seems to take also +13W idle.
Oh yeah, the magic card SPRC were tested... I like it tooIs there anyone who has a card that takes that +4W that SPCR tested? Or is this +4W vs. my +14W only a result of lower idle power usage of my 780g vs. SPCR's i945Ga? Main reason for me buying this card was SPCR's test...
There can't be so much difference, furthermore the 780G has a lower idle consumption than 945G.
I've also been wondering about that 4W idle reading, it sounds a little too low.
If our cards are almost the same as the one used in the review, and our numbers are correct, then I can see a SPCR review where a HD4550 or HD4650 appears to use no power, or has the ability to generate it's own.
After doing a little digging, I've found this Intel PDF document (link) stating that with the G965 chipset, it is able to disable it's IGP when an external graphics card is installed, resulting in a TDP reduction of 9W, bringing it to the same TDP of the P965 (Page 14, 2.4.3 - Note 3.).
Also, with the 945G I found in this PDF document (link) that the IGD (Integrated Graphics Device) cannot operate with an external PCI Express graphics device (Page 257, 10.6.3 - Note).
With these facts in mind, perhaps the 945G also disables it's IGP, resulting in a TDP reduction from 22.2W to the same level as the 945P TDP of 15.2W? And following that same line of reasoning, maybe the TDP difference is very close to the IDLE difference of 7W?
I'm just making guesses here, I know that TDP has nothing to do with IDLE watts, but in this case it could partially explain those low numbers.
I wonder if Lawrence Lee still has that HD4670 from the review? It would be interesting to see it's idle power on a non-Intel chipset motherboard.
If our cards are almost the same as the one used in the review, and our numbers are correct, then I can see a SPCR review where a HD4550 or HD4650 appears to use no power, or has the ability to generate it's own.
After doing a little digging, I've found this Intel PDF document (link) stating that with the G965 chipset, it is able to disable it's IGP when an external graphics card is installed, resulting in a TDP reduction of 9W, bringing it to the same TDP of the P965 (Page 14, 2.4.3 - Note 3.).
Also, with the 945G I found in this PDF document (link) that the IGD (Integrated Graphics Device) cannot operate with an external PCI Express graphics device (Page 257, 10.6.3 - Note).
With these facts in mind, perhaps the 945G also disables it's IGP, resulting in a TDP reduction from 22.2W to the same level as the 945P TDP of 15.2W? And following that same line of reasoning, maybe the TDP difference is very close to the IDLE difference of 7W?
I'm just making guesses here, I know that TDP has nothing to do with IDLE watts, but in this case it could partially explain those low numbers.
I wonder if Lawrence Lee still has that HD4670 from the review? It would be interesting to see it's idle power on a non-Intel chipset motherboard.
-
- Posts: 1
- Joined: Mon Dec 01, 2008 4:16 am
- Location: UK
Hi All
Ive recently bought the ASUS 4670 to fit into my new shuttle sx48p2 deluxe.
The card fits ok but because of the big heatsink and fan it means that it takes up the other pcie slot.Also cables keep getting jammed into the fan, so I thought id buy the zalman vf900 and fit that.When I took of the stock heatsink there was all this white gluey stuff which was still fluid.How can I get this off? Ive ordered some articlean which is coming with my vf900.The exposed GPU doesnt seem to have a metal cover on it so Im bit worried about how to clean it properly.Does anyone have any tips or advice with cleaning the GPU , plus any experience with a shuttle SFF box, an ASUS 4670 and a zalman vf900.
Sorry about the lengthy post, but its my first go and im a total noob when it comes to modding video cards.
Ive recently bought the ASUS 4670 to fit into my new shuttle sx48p2 deluxe.
The card fits ok but because of the big heatsink and fan it means that it takes up the other pcie slot.Also cables keep getting jammed into the fan, so I thought id buy the zalman vf900 and fit that.When I took of the stock heatsink there was all this white gluey stuff which was still fluid.How can I get this off? Ive ordered some articlean which is coming with my vf900.The exposed GPU doesnt seem to have a metal cover on it so Im bit worried about how to clean it properly.Does anyone have any tips or advice with cleaning the GPU , plus any experience with a shuttle SFF box, an ASUS 4670 and a zalman vf900.
Sorry about the lengthy post, but its my first go and im a total noob when it comes to modding video cards.
1. Different way to measure power drawMatija wrote:X-bit Labs measured the PowerColor card at 8W in idle...
Why all the differences?
2. Different cards with different power management
Notes for 1.
1a. X-bit is measuring DC voltage/amps to video card.
(neglects power saved/used by other components)
1b. SilentPCReview measures total AC power usage.
(sums power saved/used by other components.)
Method 1a should be accurate for the specific video card in any system.
Method 1b is more accurate to actual power use, but the numbers are specific to the particular motherboard and power supply used during the testing. A negative power change from IGP to a discrete card is possible under method 1b.
@QuietOC: So, to make better idea how much a real system uses, SPCR should modify method 1b. Modern IGP can use 5W or more
and that should be taken into account. SPCR should find an old discrete 2D only card and measure the actual power use.
Then they could estimate IGP consumption and then correct VGA consumption.
My 'reference' discrete card is old Matrox Millenium. I don't know it's power consumption, but it can't be very high. The card does not
even have a heatsink! And as this was 'high-end' card at that time it's possible that other models consume even less.
@Matija: Post results for HR-03 and pictures also.
and that should be taken into account. SPCR should find an old discrete 2D only card and measure the actual power use.
Then they could estimate IGP consumption and then correct VGA consumption.
My 'reference' discrete card is old Matrox Millenium. I don't know it's power consumption, but it can't be very high. The card does not
even have a heatsink! And as this was 'high-end' card at that time it's possible that other models consume even less.
@Matija: Post results for HR-03 and pictures also.
Anything bigger/better than the Accelero S2 (not S1) is overkill IMO.Matija wrote:Installed the card.
Damn, the thing is loud. Under load it's actually okay, but it's idle that's killing me. Manual fan speed control = fail.
I'll test the HR-03 on it eventually. I thought about doing it over the weekend, but the noise is killing me, so I might try tomorrow.
When you say fail, do you mean that manual fan speed control doesn't work or that it doesn't get it quiet enough for your system? I hope it's the latter as the Sapphire card is on my "list" of cards for which manual fan speed control works. . .Matija wrote:Damn, the thing is loud. Under load it's actually okay, but it's idle that's killing me. Manual fan speed control = fail.
@Aris: You know that's not true! Do a little research before you post.
Your 8800GT sucks with the power consumtion. It has idle at 35W DC and 3D load at 80W DC.
We are talking here about a 4670 card that has idle at 9W DC and 3D load 47W DC.
(from xbitlabs)
So buy two 4670 cards and you'll have lower power consumtion as 8800GT while not gaming
and better fps while gaming.
Your 8800GT sucks with the power consumtion. It has idle at 35W DC and 3D load at 80W DC.
We are talking here about a 4670 card that has idle at 9W DC and 3D load 47W DC.
(from xbitlabs)
So buy two 4670 cards and you'll have lower power consumtion as 8800GT while not gaming
and better fps while gaming.
Last edited by lechuck on Wed Dec 03, 2008 5:50 am, edited 1 time in total.
-
- Posts: 10
- Joined: Sat Nov 29, 2008 11:31 pm
- Location: Finland
I have a DC clamp meter, so I can measure the DC idle power, but I'm not doing it before I get a waterblock for this card.
Does anyone know if the Zalman waterblock for 3850/3870 card is ok for 4670? http://www.zalman.co.kr/ENG/product/Pro ... sp?Idx=286
Vicotnik's idle power dropped -3W when changing from Sapphire to HIS. What was the idle power clocks and voltages in that Sapphire? HIS card can drop it's memory clock to 250Mz and Sapphire's minimum is 500Mz so is the -3W drop form the memory clocks? If so, 3W difference is no reason for me to change the card, but for new buyers it's good to know.
Does anyone know if the Zalman waterblock for 3850/3870 card is ok for 4670? http://www.zalman.co.kr/ENG/product/Pro ... sp?Idx=286
Vicotnik's idle power dropped -3W when changing from Sapphire to HIS. What was the idle power clocks and voltages in that Sapphire? HIS card can drop it's memory clock to 250Mz and Sapphire's minimum is 500Mz so is the -3W drop form the memory clocks? If so, 3W difference is no reason for me to change the card, but for new buyers it's good to know.
I did my research, how about you do yours.lechuck wrote:@Aris: You know that's not true! Do a little research before you post.
http://archive.atomicmpc.com.au/forums. ... t=9354&p=0
HD4670 10W-70W
8800 GT 35W-78W (512MB)
8800 GT 31W-50W (256MB)
I did specify the 256mb version, not the 80w load 512mb version.
-
- Patron of SPCR
- Posts: 857
- Joined: Fri Dec 27, 2002 1:49 pm
- Location: Somerset, WI - USA
- Contact:
Not to get in the middle, but these quoted values come from different sources. You can't really make any exact conclusions from this list if they weren't measured the same. SPCR's own review of the HD4670 gave a much lower load wattage of 40W than the 70W quoted in this article.Aris wrote:I did my research, how about you do yours.
http://archive.atomicmpc.com.au/forums. ... t=9354&p=0
HD4670 10W-70W
8800 GT 35W-78W (512MB)
8800 GT 31W-50W (256MB)
I did specify the 256mb version, not the 80w load 512mb version.
Code: Select all
HD4670 ______________#¦)######}| 10W-70W %
8800 GT _____________###¦)##| 31W-50W ^ (256MB)
% = The official "peak power consumption" from the manufacturer
^ = Found at www.anandtech.com
@Aris: I guess you did some research, but your source is not reliable.
Why? Even in theory 4670 can't have 70W consumption. Remember, the card has no external power
and max. it can get from MB is 75W (on paper). No company would go so close to the limit, so about
50W is more realistic. I didn't test the card myself yet, but I read every review out there...
Xbitlabs proved to be accurate over time. You should not thrust any review 100%, there are variations
between brands too.
And as we are talking here about silent card, the idle power is very important. I overlooked 256MB, but
anyway, any card with more than 20-25W in idle is not suitable for passive.
So 8800GT -> more than 30W idle + no or very little airflow = cooking
Why? Even in theory 4670 can't have 70W consumption. Remember, the card has no external power
and max. it can get from MB is 75W (on paper). No company would go so close to the limit, so about
50W is more realistic. I didn't test the card myself yet, but I read every review out there...
Xbitlabs proved to be accurate over time. You should not thrust any review 100%, there are variations
between brands too.
And as we are talking here about silent card, the idle power is very important. I overlooked 256MB, but
anyway, any card with more than 20-25W in idle is not suitable for passive.
So 8800GT -> more than 30W idle + no or very little airflow = cooking
-
- Posts: 353
- Joined: Sat Oct 18, 2008 6:30 am
- Location: Moldova, exUSSR
At last had some time to tinker about the 4670.
Well, my Force3D doesn't go to 0.9V in idle. Instead of 0.9V I see 1.28V, and 1.34V under 3D load
And how much volts do you have folks on GPU?
Maybe to have full PowerPlay support through firmware we must have as much similar to reference design board as possible -- I tried two different firmwares and I can say that my card doesn't support VID shifting, and I think this is the only reason not getting 4W idle as SPCR reviewer has got.
1. IXBT's 4670 reference:
2. My Force3D:
Does anybody know what those technological micro switches on reference card are actually for? I also found three resistors that are missing on mine (see pics).
Well, my Force3D doesn't go to 0.9V in idle. Instead of 0.9V I see 1.28V, and 1.34V under 3D load
And how much volts do you have folks on GPU?
Maybe to have full PowerPlay support through firmware we must have as much similar to reference design board as possible -- I tried two different firmwares and I can say that my card doesn't support VID shifting, and I think this is the only reason not getting 4W idle as SPCR reviewer has got.
1. IXBT's 4670 reference:
2. My Force3D:
Does anybody know what those technological micro switches on reference card are actually for? I also found three resistors that are missing on mine (see pics).
I just put a Palit 4670 in my new build. I'd read it was "quieter" than other 4670's... so I thought together with the catalyst fan control I could have it near silent when not under load.
Wishful thinking that was. Even at the lowest speed setting it sounds like a vacuum.
I've read the accelero s2 is plenty... but can someone please confirm it will fit no problem?
Wishful thinking that was. Even at the lowest speed setting it sounds like a vacuum.
I've read the accelero s2 is plenty... but can someone please confirm it will fit no problem?
It should fit, based on this image: http://www.arctic-cooling.com/pics/accs2_01l.jpg
There are three sets of screw holes on its base. I suppose the second one is for PCBs with mounting holes spaced 43mm apart.
I still can't figure out what to do with my HR-03... Damn it, all I want is a passively-cooled graphics card
There are three sets of screw holes on its base. I suppose the second one is for PCBs with mounting holes spaced 43mm apart.
I still can't figure out what to do with my HR-03... Damn it, all I want is a passively-cooled graphics card