- 3GB 384-Bit GDDR5
- Core Clock 800 MHz
- 1 x DVI 1 x HDMI 2 x Mini DisplayPort
- 1792 Stream Processors
- PCI Express 3.0 x16
FINEST BUILD QUALITY I HAVE SEEN – part 2
10/03/2012
This review is from: HIS IceQ Radeon HD 7950 3GB GDDR5 PCI Express 3.0 x16 CrossFireX Support Video Card H795Q3G2M
Pros:
Overclocked to 975 this card can handle 2560x1600 resolution Sleeping Dogs at high res texture pack and DX11 and max settings, with a fluid average 28 fps.
This is a followup two months later to my 8/5/12 review below. Both cards have performed flawlessly alone, or in crossfire.
If you are put off by some of the bad reviews, remember that you are protected by Newegg if you have to RMA, and if the card is defective, you will only be out the cheap shipping. So when you get your card, thoroughly test it - more on that below. And keep all of your packaging for at least the 30 day return period.
If you look down at my 8/5/12 review, you will find that I was having problems fitting two of these cards in a beautiful mid-tower Antec Sonata case. In that review I speculated that I might move all my components to a full-tower Thermaltake Spedo case that I had bought about 4 years ago, which was still in its box brand new in the garage. That case is a bottom psu design. Since the part of the HIS IceQ card that is 3 slots wide, is the end where the fan is, that part extends way past the bottom PSU, and so there are no issues any more of the second card fitting in the bottom PSU case. I could have made it work in the mid-tower Antec in slot 6, only by cutting a hole in the case bottom and adding feet to the case - I was seriously considering doing that.
Back to single card performance - I pulled out Crysis, Crysis 2, and Warhead, which I bought but never played, waiting for this upgrade. I ran them with one card, 2560x1600, max AA, max AF, everything max, and averaged about 32fps, and it was completely fluid. On Crysis 2 I opted for DX11 - I took every max or ultra setting available.
A super tech guru warned about a major fps drop in Crysis once you hit the ice. Guess what - it never happened. The card sailed through and I never got below the low 30s which for me is fluid - never any feeling of lag. Same for BF3 ultra.
Finally along came Sleeping Dogs, at 30" gaming, 2560x1600, high resolution texture pack, DX11. The game looks awesome. It is the only game so far that has seriously tested this card.
Initially I got about 24fps average in the game, a little laggy, but at the rock solid overclock of 975, 1350 memory, and vddc 1087, it gets a fluid 28 fps, and never drops below that. I ran the game in crossfire, where I get fps in the 40s and up - pretty good scaling on the second card - but the higher frame rates seem to put more stress on my slow 9450 cpu, and a few times I noticed my cores max out at 100% load every core!!
I finished the game, but at the end I dropped to stock clocks on both cards in crossfire, and fps at about 40. I didn't really notice a major change from one card overclocked at 28 fps. It's a gorgeous game. If the karate is trouble (I think pc reacts slower than console) find all the health shrines, and before major fights, like Stickup, do 3 things: A. Eat, and drink: b. energ
Cons:
I have discovered what I think is a driver incompatibility problem on BF:BC2. I am running the latest Catalyst, 12.8. I have tested both cards singly and in crossfire. Singly, each card stops reporting gpu temp, and load, within about 10 minutes.
I am monitoring temps and load in-game with MSI Afterburner. But I am setting my overclock with Sapphire Trixx. I have tested gpu core clock of 975, memory at 1350, and vddc of 1087, and had that be completely stable. Sleeping Dogs at stock clock runs only about 24fps at 2560x1600, but at the 975, as I mentioned, I get 28 or more, which is fluid. With both cards, I get near 40fps.
There is nothing that tests a card like a game, and the 975 is stable, whereas 1000 appeared to be stable in furmark, and in Heaven, but in-game, that overclock caused a hang in Sleeping Dogs after 10 minutes. So 975, 1350, and vddc of 1087 seems a solid combination. Vddc you can set with Trixx. (Close CCC and Mom.exe.) MSI Afterburner gives me the OSD (on-screen display) and because I formerly had an old nvidia card, and I was setting my OSD with Riva Tuner, when I now start Afterburner, and then riva, the cpu core information from Riva shows up in my Afterburner display - that is wayy cool.
You could do the same by picking up a dirt cheap nvidia card for $25, preferably an older one (but maybe a new one would work) getting everything set up in Riva, then putting the HIS IceQ back. Riva won't work with the 7950, and in fact initially Afterburner wouldn't either, but I won't let Afterburner update to the new revision that WILL work, because I don't want it adjusting my Trixx settings - I only want it for its OSD. I decided I didn't like how the core temp and load info was laid out, so I took out my HIS, put the old nvidia back in, set everything how I wanted it, then reversed. That was a lot of driver changing, and I ran Driver Clean each time, but my OSD now is awesome, with gpu temp and load, fan rpm, clock and memory clock, and memory vram usage, for both cards, then a full set of core info on all 4 cores, temps, load, and speed, which is the same on all cores of course, but that lines up the load so I can glance from top to bottom and see my cpu usage - in Sleeping Dogs it has occasionally hit 100% all 4 cores - and THEN I might suffer severe fps drops to the low teens. This is because my processor is a slow 9450, overclocked to 3.343ghz, with 8 gigs RAM-which is sufficient. This only happened in crossfire, overclocking the HIS at 975.
So, in crossfire, I dropped to stock 800 clocks on the HIS cards so my cpu didn't have to work so hard, my core loads dropped to around 95%, and I still got a smooth 40fps or more in Dogs.
The BF:BC2 incompatibility affects one card more than the other - why I have absolutely no idea. The least affected card gives me several hours before a hang - so that one is now in main slot 3. I run that game with crossfire disabled, all max = 45
Overall Review:
When I said make sure you test your card during the 30 day return period - what is a good test?
Well, Furmark is a basic test, but not really that good to detect flaws. People who have written reviews on overclocking, choose some massive overclock, and say "It runs great!" but they are testing with Furmark. Also, be careful of Furmark - it may overheat your card.
These HIS IceQ cards can run full 2560x1600 Furmark, and at fixed max fan speed, can stay under 80, even the main card which does more work than the second card, and therefore runs hotter. Furmark will graph each card in crossfire, provided you run it full screen - crossfire only works in full screen mode, and that is also true for games I believe. If you allow auto fan, you may get temps to 85. I suggest you not let it get hotter than that.
Furmark will graph temps, and let you see the effect of fan speed on temps. Using Trixx, I either use 100% fan speed, or I use a custom fan setting, a line from 0 to 80% and 80 degrees. Since my HIS hits near max rpm at about 75% tachometer, that keeps my cards cool.
But on Sleeping Dogs, in crossfire, I was so close to maxing out my cpu core load, I dropped the custom fan setting, which put my cpus at 100%, and I went back to max fan on Sleeping Dogs. Now that I have finished that game, I am back to the Trixx custom fan setting. The custom setting seems to only affect the main card - the other runs slower fans at auto fan setting.
That is okay - that card, while it may show 100% load like the other, is not charged with assembling the whole frame. The main card, as shown by gpu-z logging, carries more core amperage than the second card, therefore that shows it does more work, and it gets hotter.
Ok, so Furmark is just a basic test - it will show you also if your OSD is working - don't let temps get past 85. Better to keep them below 80, and my temps often run below 70 in-game, depending on how hot my room is - we have had some hot 100 degree weather and I had to game only at night. Other than free Furmark, I recommend you pay $19 for the new 3dmark11, as I did. And I am glad I did. That test picked up artifacts on a Power Color 7950 that I RMA'd, in the first test segment with the submersibles. For testing there is also Unigine Heaven - which is free - but which however did not exhibit any artifacts with the defective Power Color card. That gives you 3 tests, Furmark, Heaven, and 3dMark11.
At 975 core, my 3dmark11 was about 2560, my Heaven at 1920x1080, normal tesselation, 4x AA, and 16x AF, was about 1630. With Heaven in crossfire, I broke 2900, but at 1000 core, which later hung in-game. So I consider 975 my max stable. (My overall 3dmark11 is low due to my slower cpu.)
Once you get through with your testing, do some serious gaming, on max settings. You can start up gpu-z, which is free, and do some logging - to the desktop. Open the log with notepad, and you will clearly see down the var
Share this review:
3 people found this review helpful.
Did you?
Yes
/
No
- Richard E.
-
Verified Owner
-
Owned For:
1 month to 1 year
LOADING...
