Reference 7950 faster than His 7950 IceQ at same clock.. why??

D

Drangueos

Limp Gawd
Joined
Feb 4, 2012
Messages
231
I have 2 reference 7950 800/1250 running on my system....
I got my hands on 2 His IceQ Boost Edition 850(925)/1250

so I did some tests... I found out that even running with higher clocks the Boost edition was getting same performance as the reference.

so I underclocked the Boost edition to 800/1250 and runned all the tests again, more than once... and running at same clocks the boost edition was 3 to 5% slower than the reference.

So to make the His IceQ Boost edition match the Reference one.. I have to run the GPU around 880mhz OR the memory at 1350mhz...

My question is: WHY????

the only explanation I can think is that the reference uses hynix memories and the iceq boost uses elpidia ones... and MAYBE the elpidia ones have higher latency.

any other ideas?
 
That's a pretty interesting read. I think you have a good idea about the memory timings. I'm wondering if there is any software out there that would actually tell us what we are running at, I know GPU-Z doesn't. Looking forward to others thoughts on this too. Perhaps something in the GPU's BIOS can tell us?
 
May be the latency of the memory.

Another theory that I have is the ASIC score may effect the performance.

And one other theory of mine is that IceQ has higher voltage requirement to run stable at desired speeds. More voltage = more electrical heat and noise which interferes with performance.

I have noticed many "BOOST" cards use a higher default voltage of 1.25v under load even at slower speeds.
 
About the voltage / temps.
I could undervolt the boost card to 1.1V without problems, they were running 15C lower than the reference ones.
Even with default voltages, the IceQ ones run cooler than the reference.

About the Asic
the reference ones are 77 and 73 asic, the boost ones are 61 and 60.
 
Last edited:
Drangueos said:
About the voltage / temps.


About the Asic
the reference ones are 77 and 73 asic, the boost ones are 61 and 60.
Click to expand...

That right there may be the problem. Higher ASIC score is associated with better elelctrical conductivity and performance.
 
SonDa5 said:
That right there may be the problem. Higher ASIC score is associated with better elelctrical conductivity and performance.
Click to expand...

If this was true a higher voltage on the same card would increase the performance until a limit and a lower voltage would decrease.
I think the higher voltage already compensate the lower electrical conductivity...
 
Drangueos said:
If this was true a higher voltage on the same card would increase the performance until a limit and a lower voltage would decrease.
I think the higher voltage already compensate the lower electrical conductivity...
Click to expand...


Higher voltages will negatively effect heat, resistance, and put more strain on vrm and other all electrical parts. Less voltage is always better. Bettet ASIC score is better.
 
SonDa5 said:
Higher voltages will negatively effect heat, resistance, and put more strain on vrm and other all electrical parts. Less voltage is always better. Bettet ASIC score is better.
Click to expand...

to run the cards at 1100mhz..
I need 1.150v on both... reference and IceQ
And the IceQ runs at least 10C cooler, but the reference still perform 3 to 4% better.

I think it's something with the memory and not the actual GPU.
 
I wonder what the performance delta is on cards that are truly the exact same thing.
 
Drangueos said:
to run the cards at 1100mhz..
I need 1.150v on both... reference and IceQ
And the IceQ runs at least 10C cooler, but the reference still perform 3 to 4% better.

I think it's something with the memory and not the actual GPU.
Click to expand...


The lower ASIC score card doesn't conduct as well. The cooler temps are probably result of better heat sink but the gpu itself will still not conduct electricity as well and it probably has stability problem.


Run HWinfor64 in background and compare speeds, voltages, temps, power out put and you can get a good idea of what all is going on between the two cards. Keep in mind the lower ASIC score GPU is not going to conduct electricity as efficiently as the higher ASIC score GPU.
 
SonDa5 said:
Run HWinfor64 in background and compare speeds, voltages, temps, power out put and you can get a good idea of what all is going on between the two cards. Keep in mind the lower ASIC score GPU is not going to conduct electricity as efficiently as the higher ASIC score GPU.
Click to expand...

but as far as I know a higher voltage would compensate this lower efficiency of conducting electricity, making a lower asic card using a higher voltage perform like a higher asic with lower voltage.

I tested the 77% at 0.993V and still it was 3 to 4% faster than the 61% at 1.256V 1.2V, 1.15V or 1.1V .

I tested one card that is 61% and another one that is 53%, both same model.. and the difference between them is 0.

That is why I think the problem is the memories...
 
I got an answer from HIS support about this: "This is due to their bios timing and settings are different."
 
Drangueos said:
I got an answer from HIS support about this: "This is due to their bios timing and settings are different."
Click to expand...

I guess this also explains why I see in some forums for people who changed their BIOS and then state "it flashed properly but I had instability in games after"
 
You must log in or register to reply here.
Back
Top