Overclock.net › Forums › Graphics Cards › NVIDIA › NVIDIA Drivers and Overclocking Software › GTX 1070/1080/TITAN X(2ND GEN) BIOS - Who has it?
New Posts  All Forums:Forum Nav:

GTX 1070/1080/TITAN X(2ND GEN) BIOS - Who has it? - Page 18  

post #171 of 600
Thread Starter 
Quote:
Originally Posted by plant View Post

Hey Laithan, any guesses at how much power is drawn on an FE card at 2275mhz and 1.2v? That is what I was able to achieve with XOC bios however, performance was lower than FE bios at 2100mhz. But the card itself held up fine in all of my games and benchmarks. I would guess ~280w maybe? I know there was significantly more heat coming off my radiator.

Also if you remember the 980ti reference with 6 power phase running pretty good at 1.3v. Maybe this is just my bias as a Fanboy Edition owner but I think perhaps single 8 pin and 5 power phases will not be too limiting. If you are willing to run way out of spec...


Hi bud. Just to clarify, the power draw has no relation to the voltage applied. The voltage is what allows you to maintain a certain clock speed (boosting or not). The power draw is based on GPU workload and complexity regardless of voltage and clock speeds. Insufficient power can restrict boosting but this is minimum power required to boost to a given speed but not to be confused with drawing maximum power possible.

So to answer your question, the Frowner's edition with the single 8-pin is 240/250W max right now basically. 175Wrail + 75Wmb which pretty much everyone is saying isn't enough to satisfy the boost clocks that the GPUs are sometimes capable of. If there was more power available could it boost higher, YES.. If there was more power available could it maintain a high boost clock longer (without throttling), YES. I think the 10x0 cards sold with a single 8-pin are going to find that when we can (more easily) mod the BIOS, there's probably no hope of removing the power perfcaps therefore removing throttling will probably not be possible..
post #172 of 600
Quote:
Originally Posted by Laithan View Post

Hi bud. Just to clarify, the power draw has no relation to the voltage applied. The voltage is what allows you to maintain a certain clock speed (boosting or not). The power draw is based on GPU workload and complexity regardless of voltage and clock speeds. Insufficient power can restrict boosting but this is minimum power required to boost to a given speed but not to be confused with drawing maximum power possible.

So to answer your question, the Frowner's edition with the single 8-pin is 240/250W max right now basically. 175Wrail + 75Wmb which pretty much everyone is saying isn't enough to satisfy the boost clocks that the GPUs are sometimes capable of. If there was more power available could it boost higher, YES.. If there was more power available could it maintain a high boost clock longer (without throttling), YES. I think the 10x0 cards sold with a single 8-pin are going to find that when we can (more easily) mod the BIOS, there's probably no hope of removing the power perfcaps therefore removing throttling will probably not be possible..

Thanks. Yes I understand the first point and should have clarified. With the FE bios I did hit the power limit under high loads. WIth the XOC bios, not only did I never hit the power limit (because it does not exist in that bios), I also ran an extra 175mhz on the core, with an extra .15v. So I would assume that I exceeded 217w/240w (whatever the FE limit is), but by how much, I don't know.
post #173 of 600
Quote:
Originally Posted by Neopsy View Post

Here is a list of power limits and max RPM of all BIOS from https://www.techpowerup.com/vgabios/.
Every single one of them is useless for 1080FE, max RPM is too low.
...
*power is in mW, first number is for 100% power limit, second is max

Note, that some cards offer up to 99W power from motherboard, which could lead to reasonable performance.
In 5k resolution, 1080FE is power limited at 120%, default clock and voltage (voltage drops to ~ 0.981v), in almost any game, all the time. Huge disappointment .
Source code with offsets (Click to show)
def d (n) print $b[n,4].pack("C*").unpack("V") end
def d2(n) print $b[n,2].pack("C*").unpack("v") end
for n in Dir["*.rom"]
s=File.binread n
$b=s.unpack "C*"
d 0x2E12A
d 0x2E12E
print ' total '
d 0x2E158
d 0x2E15C
print ' MotherBoard '
d 0x2E186
d 0x2E18A
print ' PCIe1 '
d 0x2E1B4
d 0x2E1B8
print ' PCIe2 '
d2 0x2E7C0
d2 0x2E7C2
d2 0x2E7C4
d2 0x2E7C6
d2 0x2E7C8
d2 0x2E7CA
print ' fan '
d2 0x2E79C
print 'maxRPM '
p n
end
I compared the evga and gigabyte to the evga sc using hexdump/ediff and then I borrowed your ruby script and extended it to the regions that differed that did not look like vendor Ids and date strings. Not sure it provided any clarity other than:
a. the SC does not change the power limits from the FE card(s)
b. Interestingly my unedited gigabyte and evga 1080s run consistently under gaming load at 1843 so long as I crank the fans manually.

Maybe someone already knows what these offsets mean?
Here is the output - see below for the ruby script:
NOTE: I don't know what the alignment/width of the fields that mismatch are so I decoded first as 32 then after the "-" are the 2 16bit values in that same 32bit word:
Code:
EVGA GTX1080 SC (86.04.17.00.80).rom:
        27504:  [0]-[0][0]
        27508:  [0]-[0][0]
        27510:  [4274913792]-[512][65230]
        27514:  [0]-[0][0]
        2b9bc:  [3892613194]-[36938][59396]
        2b9c0:  [201327245]-[653][3072]
        2b9c4:  [1327373914]-[7770][20254]
        total:  [180000][217000]
        MB:     [66000][75000]
        PCIe1:  [162000][175000]
        PCIe2:  [0][0]
        2e78c:  [1677721602]-[2][25600]
        2e798:  [105]-[105][0]
        2e7bc:  [4004608]-[6912][61]
        fan:    [1920][0][2560][1600][2880][3000]
        maxRPM: [3000]
        2e7ec:  [16909570]-[1282][258]
        2e7f0:  [203000]-[6392][3]
        2e7f8:  [598212864]-[256][9128]
        2e7fc:  [2550202420]-[52][38913]
        2e800:  [536885345]-[14433][8192]
Gigabyte1080FE.rom:
        27504:  [13107]-[13107][0]
        27508:  [13107]-[13107][0]
        27510:  [4292673797]-[261][65501]
        27514:  [26214]-[26214][0]
        2b9bc:  [805889359]-[58703][12296]
        2b9c0:  [201327209]-[617][3072]
        2b9c4:  [1411259994]-[7770][21534]
        total:  [180000][217000]
        MB:     [66000][75000]
        PCIe1:  [162000][175000]
        PCIe2:  [0][0]
        2e78c:  [1679032322]-[2][25620]
        2e798:  [72089705]-[105][1100]
        2e7bc:  [5057820]-[11548][77]
        fan:    [1344][1100][2640][2200][2880][3700]
        maxRPM: [4000]
        2e7ec:  [132354]-[1282][2]
        2e7f0:  [0]-[0][0]
        2e7f8:  [0]-[0][0]
        2e7fc:  [0]-[0][0]
        2e800:  [536870912]-[0][8192]
EVGA_FE.rom:
        27504:  [13107]-[13107][0]
        27508:  [13107]-[13107][0]
        27510:  [4292673797]-[261][65501]
        27514:  [26214]-[26214][0]
        2b9bc:  [805889359]-[58703][12296]
        2b9c0:  [201327209]-[617][3072]
        2b9c4:  [1411259994]-[7770][21534]
        total:  [180000][217000]
        MB:     [66000][75000]
        PCIe1:  [162000][175000]
        PCIe2:  [0][0]
        2e78c:  [1679032322]-[2][25620]
        2e798:  [72089705]-[105][1100]
        2e7bc:  [5057820]-[11548][77]
        fan:    [1344][1100][2640][2200][2880][3700]
        maxRPM: [4000]
        2e7ec:  [132354]-[1282][2]
        2e7f0:  [0]-[0][0]
        2e7f8:  [0]-[0][0]
        2e7fc:  [0]-[0][0]
        2e800:  [536870912]-[0][8192]
Source Code with differing offsets (Click to show)
#!/usr/bin/ruby
def d (n) print $b[n,4].pack("C*").unpack("V") end
def d2(n) print $b[n,2].pack("C*").unpack("v") end
for n in Dir["*.rom"]
print n,":"
s=File.binread n
$b=s.unpack "C*"

print "\n\t27504:\t"
d 0x27504
print "-"
d2 0x27504
d2 0x27506

print "\n\t27508:\t"
d 0x27508
print "-"
d2 0x27508
d2 0x2750a

print "\n\t27510:\t"
d 0x27510
print "-"
d2 0x27510
d2 0x27512

print "\n\t27514:\t"
d 0x27514
print "-"
d2 0x27514
d2 0x27516

print "\n\t2b9bc:\t"
d 0x2b9bc
print "-"
d2 0x2b9bc
d2 0x2b9be

print "\n\t2b9c0:\t"
d 0x2b9c0
print "-"
d2 0x2b9c0
d2 0x2b9c2

print "\n\t2b9c4:\t"
d 0x2b9c4
print "-"
d2 0x2b9c4
d2 0x2b9c6

print "\n\ttotal:\t"
d 0x2E12A
d 0x2E12E
print "\n\tMB:\t"
d 0x2E158
d 0x2E15C
print "\n\tPCIe1:\t"
d 0x2E186
d 0x2E18A
print "\n\tPCIe2:\t"
d 0x2E1B4
d 0x2E1B8

print "\n\t2e78c:\t"
d 0x2e78c
print "-"
d2 0x2e78c
d2 0x2e78e

print "\n\t2e798:\t"
d 0x2e798
print "-"
d2 0x2e798
d2 0x2e79a

print "\n\t2e7bc:\t"
d 0x2e7bc
print "-"
d2 0x2e7bc
d2 0x2e7be

print "\n\tfan:\t"
d2 0x2E7C0
d2 0x2E7C2
d2 0x2E7C4
d2 0x2E7C6
d2 0x2E7C8
d2 0x2E7CA
print "\n\tmaxRPM:\t"
d2 0x2E79C

print "\n\t2e7ec:\t"
d 0x2e7ec
print "-"
d2 0x2e7ec
d2 0x2e7ee

print "\n\t2e7f0:\t"
d 0x2e7f0
print "-"
d2 0x2e7f0
d2 0x2e7f2

print "\n\t2e7f8:\t"
d 0x2e7f8
print "-"
d2 0x2e7f8
d2 0x2e7fa

print "\n\t2e7fc:\t"
d 0x2e7fc
print "-"
d2 0x2e7fc
d2 0x2e7fe

print "\n\t2e800:\t"
d 0x2e800
print "-"
d2 0x2e800
d2 0x2e802

print "\n"
end

Edited by cekim - 7/8/16 at 10:02pm
post #174 of 600
.So after spending the whole day building a custom loop, I have some results to report with the XOC bios and my MSI gaming X

Firstly, too much voltage causes instability. I was not able to make it fully stable at 1.2v at 2202mhz, but at 1.15v it is rock stable at 2202mhz. Below I am attaching a 3d mark score at this frequency.

http://www.3dmark.com/3dm/13061604?
post #175 of 600
Quote:
Originally Posted by Xeq54 View Post

.So after spending the whole day building a custom loop, I have some results to report with the XOC bios and my MSI gaming X

Firstly, too much voltage causes instability. I was not able to make it fully stable at 1.2v at 2202mhz, but at 1.15v it is rock stable at 2202mhz. Below I am attaching a 3d mark score at this frequency.

http://www.3dmark.com/3dm/13061604?
That's a really low score. I get 24 777 at 2126MHz (http://www.3dmark.com/fs/9009210). Try to see if you get better scores at lower frequencies or with the stock BIOS.
Edited by versions - 7/9/16 at 12:53am
post #176 of 600
Quote:
Originally Posted by Xeq54 View Post

.So after spending the whole day building a custom loop, I have some results to report with the XOC bios and my MSI gaming X

Firstly, too much voltage causes instability. I was not able to make it fully stable at 1.2v at 2202mhz, but at 1.15v it is rock stable at 2202mhz. Below I am attaching a 3d mark score at this frequency.

http://www.3dmark.com/3dm/13061604?

Hi xeq, nice score...

Are you getting better scores in graphics score at 2200???

The thing is trying with the same settings in your system except for gpu clock. Your cpu clock should be higher to get the best of your 1080 (4.7 or above).
post #177 of 600
I was always at 23800 to 24300 with my 1080 so did not really see big gain, but it does have more performance in games.

My 5820k is at 4.4 so i will try going to 4.8 for benches ( requires 1.4v so cant use it 24/7) and compare it.
post #178 of 600
Quote:
Originally Posted by Xeq54 View Post

I was always at 23800 to 24300 with my 1080 so did not really see big gain, but it does have more performance in games.

My 5820k is at 4.4 so i will try going to 4.8 for benches ( requires 1.4v so cant use it 24/7) and compare it.
Do you have a game with a canned benchmark to test that?
Edit: Also, how did you limit the voltage to 1.15?
Edited by Noshuru - 7/9/16 at 8:12am
Nosh's Rig
(21 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 6700K Z170-A NVIDIA GeForce GTX 1080 G.Skill RipJaws 4  
RAMRAMRAMHard Drive
G.Skill RipJaws 4  G.Skill RipJaws 4  G.Skill RipJaws 4  Seagate Barracuda 2TB 
Hard DriveHard DriveHard DriveCooling
Seagate Barracuda 3TB Samsung 850 EVO 1TB Crucial MX100 512GB Corsair H110i GT 
OSMonitorMonitorKeyboard
Windows 10 Pro 64Bit Benq XL2411T 144Hz Asus VS247H 60Hz  Ducky Shine 3 Tenkeyless  
PowerCaseMouseMouse Pad
be quiet! PowerZone 850w Corsair 450D  Steelseries Rival Steelseries QCK+ 
Audio
Asus Xonar DX 
  hide details  
Nosh's Rig
(21 items)
 
  
CPUMotherboardGraphicsRAM
Intel Core i7 6700K Z170-A NVIDIA GeForce GTX 1080 G.Skill RipJaws 4  
RAMRAMRAMHard Drive
G.Skill RipJaws 4  G.Skill RipJaws 4  G.Skill RipJaws 4  Seagate Barracuda 2TB 
Hard DriveHard DriveHard DriveCooling
Seagate Barracuda 3TB Samsung 850 EVO 1TB Crucial MX100 512GB Corsair H110i GT 
OSMonitorMonitorKeyboard
Windows 10 Pro 64Bit Benq XL2411T 144Hz Asus VS247H 60Hz  Ducky Shine 3 Tenkeyless  
PowerCaseMouseMouse Pad
be quiet! PowerZone 850w Corsair 450D  Steelseries Rival Steelseries QCK+ 
Audio
Asus Xonar DX 
  hide details  
post #179 of 600
so this bios, does it force the voltage to 1.24 at all times? or can I limit it? I am wondering if I could test it..but don't have watercooling on my card, its the MSI GTX 1080 Gaming (non-x). I can't even get to 2000 stable, so I am hoping that some additional voltage can help, and these cards have more power available than FE.
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel i5 6600k @ 4GHz ASUS Maximus VIII Hero MSI Gtx 1080 G.Skill Ripjaw 16GB Dual Channel 
Hard DriveOSMonitorKeyboard
AMD R7 2x256GB RAID 0 Windows 10 64Bit 1xSamsung U28E590D 4K Monitor, 2x ASUS VW266H Corsair strafe rgb silent 
PowerCase
EVGA 1300W G2 NZXT Noctis 450 
  hide details  
My System
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel i5 6600k @ 4GHz ASUS Maximus VIII Hero MSI Gtx 1080 G.Skill Ripjaw 16GB Dual Channel 
Hard DriveOSMonitorKeyboard
AMD R7 2x256GB RAID 0 Windows 10 64Bit 1xSamsung U28E590D 4K Monitor, 2x ASUS VW266H Corsair strafe rgb silent 
PowerCase
EVGA 1300W G2 NZXT Noctis 450 
  hide details  
post #180 of 600
Yes you can limit it with the curve. Just make it flat after your desired voltage.

I.e my curve goes up to 2202 at 1150mv and then remains at 2202 for all higher voltage points. This makes the card run at 1.15v
New Posts  All Forums:Forum Nav:
  Return Home
This thread is locked  
Overclock.net › Forums › Graphics Cards › NVIDIA › NVIDIA Drivers and Overclocking Software › GTX 1070/1080/TITAN X(2ND GEN) BIOS - Who has it?