CrossFire and Test Setup
CrossFire with the R7870 HAWK
It is not secret that AMD is a bit behind NVIDIA when it comes to multi-GPU functionality. That being said, AMD is still leaps and bounds from where they once were. We saw a pretty significant jump in scaling from the HD 5000 series to the HD 6900 cards. One issue that does crop up in discussion is the amount of “micro-stutter” that the AMD cards have as compared to the NVIDIA SLI solutions. Something that is not widely known among users is that NVIDIA has dedicated units in each GPU which help to even out frame rates and cuts back on the stutter. So while the NV and AMD multi-GPU solutions may benchmark fairly evenly with each other, most users will agree that the NV solution does feel smoother.
The cover of the GPU Reactor dominates the back of this board. Note as well the nice cover for the entire back, which protects the PCB and provides rigidity.
That being said, I decided to give CrossFire another shot with the latest HD 7870 cards. I used the Catalyst 12.6 beta release drivers which not only fixed a bunch of issues, it also improved performance and compatibility with many games in CrossFire. The cards were instantly recognized by Windows 7 and CrossFire was enabled once the drivers were installed. The HD 7870 cards only have one CrossFire over-the-top connection, so tri and quad card solutions are not available for this level of product from AMD.
Test Setup
The base system is an overclocked Phenom II X6 1100T. It is set at a constant 3.7 GHz instead of the standard 3.3 GHz with six cores and the turbo mode of 3.7 GHz with three cores or less. I have been using this CPU rather than the latest FX-8150 because it honestly has better overall performance and is a far less choppy CPU in these applications. This will likely change when Vishera hits the market, but for the time being I am still very interested in gaming performance with high end cards on the AMD platform, and this is the best CPU that AMD has to date for gaming.
Removing the plastic cover reveals the secondary PCB that powers the GPU Reactor. Tantalum capacitors as far as the eye can see…
I tested the MSI R7870 Hawk against the XFX R7850 DD Black Edition and the MSI R7950 TFIII/OC. The R7850 comes in at approximately $274 and has 1024 stream units clocked at 975 MHz combined with 1250 MHz (5000 MHz effective) memory. The $329 MSI R7950 has 1792 stream units clocked at 880 MHz and a 384 bit memory bus at 1250 MHz (5000 effective). I tested the R7870 in CrossFire, but I utilized a 1250 MHz memory speed rather than the stock 1200 MHz. We will see that MSI was a bit conservative when it comes to memory speed, and at the core 1100 MHz GPU clock, it is somewhat memory starved.
I like to test these setups in actual cases, and I used the new Enermax Fulmo GT case as the basis for this review. Previously I had used the Silverstone Raven RV02, but that case was getting a bit long in the tooth when it comes to design. The Fulmo is much more spacious, has far superior cable routing, and is packed full of fans to keep even the hottest video card cool. I also started using the Corsair AX1200 power supply, which is probably one of the absolute best units that can be found on the market. A bit pricey, but very efficient and can push any video card on the market (in single and multi-GPU configurations).
The top card is the upcoming GTX 670 Power Edition, followed by the R7870 HAWK, with the last gen HD 6950 Twin Frozr II. Oh, how the HAWK has grown…
Phenom II X6 1100T @ 3.7 GHz
GSkill 2 x 4GB DDR-3 1600 @ 9.9.9.27 timings
Asus SABERTOOTH 990FX Motherboard
Corsair AX1200 Power Supply
Western Digital Caviar Black 1TB HD
Lite-On BDR-Drive
Enermax Fulmo GT Full-size ATX Case
Windows 7 Ultimate 64 Bit
Catalyst 12.6 Beta Drivers
Great review Josh!
Great review Josh!
Haha@gaming with AMD CPU’s.
Haha@gaming with AMD CPU’s.
…Fool.
Great review, by the
…Fool.
Great review, by the way
I agree, what a waste. I
I agree, what a waste. I find it easier to scale the results down, than trying to imagine what the difference would be if it were an Intel based system.
You sure about that
You sure about that statement…? Have you looked at th next gen games coming out? They all use 6-8 core CPUs… You better buy one before they reach $500…
Thanks for the great review.
Thanks for the great review. Maybe my next Corssfire System 🙂
BTW >>@anonymous “Haha@gaming with AMD CPU’s.”<< ...stupid people writing stupid things!...
I see that the Lightning
I see that the Lightning Edition GTX680 is out now. Is it possible to get a review of that? If so, could some of the benchmarks it’s compared to be the MSI 7970 Lightning Edition that’s Stock O.C’d to 1070MHZ AND a 7970 GHZ Edition (any manufacturer will do). That way we could see how a stock overclocked 680 fairs against this 7970 GHZ card and a normal 7970 o.c’d over a GHZ? What’s your thoughts?
But not to deter, great review, glad to see that AMD is really bringing the heat against nvidia.
I’ll see what Ryan has
I'll see what Ryan has planned for that card!
I would be great to game with
I would be great to game with this esspecially Crossfire with an AMD APU
A current APU really wouldn’t
A current APU really wouldn't push this in CrossFire all that well. While the current A8-3870K overclocked is decent at its max 3.5 GHz speed, it still will not outmuscle the latest Intel Ivy Bridge quad core products. Kaveri might be a very different beast though, but it still only looks to be 2 module/4 core with GPU attached. Perhaps next summer we will finally see an APU with some serious muscle?
APU and serious muscle is an
APU and serious muscle is an oxymoron. AMD makes discrete graphics too, remember?
6600 series MAX for the apu
6600 series MAX for the apu dual graphics ….
6600 series MAX for the apu
6600 series MAX for the apu dual graphics ….
hopefully a six core
hopefully a six core piledriver apu with no gpu, or use the internal gpu for fpu calculations? that would DESTROY intel.
Yea because just last ye…
Yea because just last ye… er, back in 2005, AMD was the better CPU.
It is a great review….but
It is a great review….but why were the cards not benched while OCed. Also my were the cards tested on a AMD platform and not at least a OCed 4.5ghz 2500k/2600k/3570k/3770k ?????? It just does not make sense to do a crossfire review with a CPU that clearly bottlenecks the video cards at lower resolutions. You can clearly see this in every benchmark where the scaling stinks at resolutions below 2560/1600 since at that resolution it really starts to load the GPUs more then the cpu since the frame rates are finally low enough that the CPU is not bottlenecking the Graphics cards.
You clearly need to upgrade your video card benchmarking Platform to a intel z77 overclocked 3570k or 3770k.
My Oced 4.7ghz 2600k with SLI Evga superclocked GTX560TI’s @ 1015mhz cores and 2400 memory score better then these totally AWESOME HAWK 7870 cards. Its just a shame you are using such a lame cpu that its slowing them down on benchmarks and not showing the full potential of the video cards. I am pretty sure At least with a INTEL benchmarking platform you would see that the 7870 hawks are faster then a plain jane 7950.
Ah yes. Another Intel
Ah yes. Another Intel “NOTHING ELSE COMPARES” fan boy. First and utmost, I am not an AMD fan boy. I would say I’m more of an Intel Hate boy! Due to the way they treat their customers, the way they carry out business which has been more than morally questionable over the past decade, and the fact with a 10nm process decrease, “new/improved” architecture, they only yield a 5% performance increase over sandy bridge (power improvements of course, but that’s a given of a smaller process)? Clearly capable of a lot more; but they release it, to just milk their loyal customers. Argument could be the same if not worse than AMD’s Bulldozer. BUT, AMD doesn’t have anywhere near the R&D budget of Intel, nor was that their focus. They knew they couldn’t beat Intel with IPC so they leaned towards the higher core count and Multi-Tasking/Threading capability… Oh, not to mention the anti-trust suits with Intel.
Anyway, Getting off track here.
My question to people such as yourself, is that so what if the CPU is “bottle-necking” a xFire GPU setup at low resolutions? Guess what? You are getting well in excess of 60fps, even 100fps in some cases (60Hz is usually the highest NATIVE refresh rate of modern PC monitors) regardless.. so what does any higher performance achieve? Absolutely nothing, other than a lighter wallet. Also… do you play at low resolutions? Any gamer worth their weight in water would game at no less than 1080p. Why would you Crossfire for low resolutions? That’s like boosting your grandma’s daily commute so she can drive to the shops once a week in a 500HP Corolla…
Even the argument of future proofing has it’s flaws. As by the time the performance difference will make a noticeable/significant difference, your PC will still be out of date by then and you could buy something with 50-100% better performance at half the price. I have purchased absolute high end PCs before (in the early and mid 00’s when there was an argument it was ‘needed’). Waste… Of… Money.
I get absolutely fantastic video converting performance (comparable to sandy i7’s/Ivy i5’s), more than sufficient gaming performance and great multi-tasking/everyday performance. All this for $180 (FX-8320). A motherboard with 16/16x xFire, all the trimmings for $30 less than the same branded but Intel equivalent.
Not to mention I have never had an AMD CPU die on me. My Athlon 64 Clawhammer lasted 9 years (still working today) and 6 of those years was pretty much running 24/7 overclocked. I have had 3 Intels die on me (one even within the warranty period). Excluding stories from friends. I haven’t even built 70 PC’s yet. Worse I’ve heard first hand of AMD is I had a friends FX-4300 run at the wrong voltage (too high) so he had to set it manually until they fix it with a BIOS update.
Don’t get me wrong. If I absolutely needed the performance/IPC capabilities of an i7 (for work etc) I would buy one. However, I doubt many people would truly be in that position. Buying an overpowered processor is often caused by ego or pretentiousness. I understand that, just don’t label it as something else.
I can understand your point
I can understand your point of view, and even some of your frustration. But this man is correct. It isnt fair to Amd OR the readers of his review to skew the results of these benchmarks by not using a setup that completely takes cpu overhead out of the picture. Say whT you want about intel as a business, or as a producer of technology… they have better performing products. In the interest of letting users know exactly what levels of performance these cards are capable of, it is irresponsible to present a picture of performance based on an outdated setup. This is not to take anything away from this excellent and thorough review. But as the poster noted, these results are CLEARLY bottlenecked at resokutions lower than 2560, as evidenced by the general similarities of the frane rates at differing levels of resolution on different video cards. Are you trying to say that all cards tested, even though from different generations and product families ALL have similar performance?
Good Intel CPU’s eliminate
Good Intel CPU’s eliminate bottlenecks when testing GPU’s.
AMD CPU’s are a bottleneck. FACT.
7870 hawk CF bottlenecked by
7870 hawk CF bottlenecked by a X6 1100T… so sad!
AMD fanboys, please, accept the fact that an i7 or even an i5 could push those FPS a lot higher than this, at any resolution.
best review i’ve read on 7870
best review i’ve read on 7870 xfire! Thanks for the informative review, helped me make a decision!
7870 cf beat 7970/680
7870 cf beat 7970/680