Archive Home arrow Reviews: arrow Video Cards arrow VisionTek Killer HD5770 Combo Card
VisionTek Killer HD5770 Combo Card E-mail
Reviews - Featured Reviews: Video Cards
Written by Steven Iglesias-Hearst   
Tuesday, 15 February 2011
Table of Contents: Page Index
VisionTek Killer HD5770 Combo Card
Closer Look: VisionTek Killer Combo
VisionTek Killer HD5770 Detailed Features
Features and Specifications
Killer NIC Software and Testing
Video Card Testing Methodology
DX10: 3DMark Vantage
DX10: Street Fighter IV
DX11: Aliens vs Predator
DX11: Battlefield Bad Company 2
DX11: BattleForge
DX11: Lost Planet 2
DX11: Tom Clancy's HAWX 2
DX11: Metro 2033
DX11: Unigine Heaven 2.1
VisionTek Killer HD5770 Temperatures
VGA Power Consumption
VisionTek Killer HD5770 Overclocking
Final Thoughts and Conclusion

VisionTek Killer HD5770 Overclocking

The clocks on the Killer HD5770 are already quite high so I wasn't expecting to have much headroom to play with clock speeds, also working against me was the lack of over voltage support. Armed with my lack of confidence I went ahead to try and break any barrier that was put in front of me, first I fired up AMD Overdrive in the Catalyst Control Center and noticed that the GPU clock had a cap of 960 MHz but within not too much time I had hit that wall and AMD Overdrive still called it stable, it wasn't until I fired up FurMark that I found out otherwise.

VisionTek_Killer_HD5770_Combo_Video_Card_GPU-Z_OC.gif

Several restarts and runs of FurMark later and a switch from AMD overdrive to MSI Afterburner found a stable GPU overclock of 910MHz (+60MHz). After tweaking the GPU I moved on to the memory, a little research showed that the Elpida RAM chips used on the Killer HD5770 were rated at 5GHz (1250MHz) so I knew there was a little headroom to play with. Armed with this I proceeded to crank up the memory speed till I hit a wall at 1315MHz (+115MHz).

Test Item Standard GPU Overclocked GPU/RAM Improvement
AMD Radeon HD5770 850/1200 MHz 910/1315 MHz 60/115 MHz
DX10: Street Fighter IV 27.53 36.37 8.8 FPS (31.9%)
DX10: 3dMark Jane Nash 16.46 18.09 1.2 FPS (7.3%)
DX10: 3dMark Calico 12.10 13.29 1.2 FPS (9.9%)
DX11: HAWX 2 40 43 3 FPS (7.5%)
DX11: Aliens vs Predator 17.5 19

1.5 FPS (8.5%)

DX11: Battlefield BC2 34.23 36.44 2.2 FPS (6.4%)
DX11: Metro 2033 14.51 15.33 0.8 FPS (5.5%)
DX11: Heaven 2.1 16.60 17.70 1.1 FPS (6.6%)
DX11: Battle Forge 20.30 21.30 1 FPS (4.92%)

Armed with a moderate overclock we went back to the bench and ran through the entire test suite. Overall there is an average 10% increase in scores (at 1920x1080 resolution) which amounts to an average 1.5% to 2% increase in frames per second. We also re ran temperature tests at the overclocked speeds at the same ambient temperature of 26.6°C. The Killer HD5770 idled at 50°C (fan speed 43%) and at full load it was 84°C (fan speed 64%), at 100% fan speed we saw the temperature drop to 82°C but once again the fan noise was too much to bear for the sake of 2°C.



 

Comments 

 
# I like the ideaktew 2011-02-16 07:19
I've had very good history with 5770s and the network speed improvements are very interesting (if you need that type of thing). I think my core disappointment is they did not use a quality cooling solution. Better heatsink or double fan or something to get to or below reference standards.

If I put a mid range machine together again, I'm looking for cool and quiet just as much as performance. This one seems to miss on each of those details and the price has moved it beyond a range I'd be considering in a mid-range build.
Report Comment
 
 
# HmmIM0001 2011-02-16 10:55
The 5770 is a nice midrange card and paired with the Killer Nic, it would be the perfect card for IMO, a WoW player who wants a little graphics upgrade (or a lot) and a nice nic boost to boot. (It is one of the big games that does show a benefit using the Killer Nic vs Onboard.

A lot of people continue to cry foul and say it is just a placebo effect, but a lot of the high end gamers still pick one up since it does do a bit of cleaning and prioritizing of gaming traffic over other traffic while you are gaming. I do have a new Killer 2100 at home here to try but I have yet to have the time to open my rig and install it. Just too busy to deal with it as the comp also needs a bit of tweaking while I have her out.

Either way this card would be a prime candidate for a good crossfire setup to make it a potent little combo.

Also the cooling solution is similar to what VisionTek has been using on a lot of their card and it may make the system itself run a bit hotter since it vents into the case, it also cools the card a reasonable bit better than the stock reference design while also being quieter. What is the bummer is the 5700 series chip runs hotter than the 5800 series for whatever reason. It is the same chip used on the 5870M on laptops and they just get hot under load. The VT 5870 I own on my desktop has a very similar cooler (but more copper) but it idles at around 30-40C and under load never goes above the 70C's. A nice improvement over the older awesome yet super hot running 4800 series. (I had a VT 4870X2) It idled at 58C if the idle clocks were enabled, and under load it often hit 80+ with the VRM's themselves hitting 100C quite often on a lot of owners cards. They ran hot but ran great while they lasted.
Report Comment
 
 
# Not that cheap, IMO.Olle P 2011-02-17 04:11
While the VisionTek Killer 2100 is a pretty expensive NIC, you can get a similar card at
Report Comment
 
 
# most of the reply disappeared...Olle P 2011-02-17 04:14
A similar NIC can be had at
Report Comment
 
 
# Darn code!Olle P 2011-02-17 04:18
It's the "less than" sign that cut my messages short... Here we go again:

A similar NIC can be had at less than $10.
A basic HD5770 less than $130.
Combined less than $140, which is way less than the $200 combo card.

The combo card is just about only worthwhile when you have a micro-ITX motherboard (only one expansion slot) *and* need the extra network speed *and* need fair graphics performance.
Report Comment
 
 
# RE: VisionTek Killer HD5770 Combo CardSteven Iglesias-Hearst 2011-02-18 00:32
Similar NIC? or are you talking about any old network card?
Value rating was based on buying a VisionTek HD5770 and a Killer NIC since there are no other VGA+NIC comparison products.

AS I pointed out in my article, this card is aimed at MMORPG / RTS players.
Report Comment
 
 
# RE: RE: VisionTek Killer HD5770 Combo CardOlle P 2011-02-18 05:20
The only reasons for using a NIC when there is a network controller on the motherboard are to a) offload the CPU and/or b) to have more network connectors.
"Any old network card", as you put it, *will* do this. And when it comes to MMORPG gaming the LAN speed is fairly irrelevant, since the server typically doesn't reside within the player's LAN. The vast majority of players have 100 Mbps internet speed or less, which will bottleneck any NIC.
So for relevancy the comparison should also include a cheap, separate, network card, and it should be made towards some server on the internet, or at least on a 100Mbps LAN incapable of higher speeds. I have a hunch that the Killer card isn't decidedly faster in that situation...
Report Comment
 
 
# RE: RE: RE: VisionTek Killer HD5770 Combo CardSteven Iglesias-Hearst 2011-02-18 14:49
I do not own a game server that I can test on, also I can't just choose 'some' game server to conduct tests on because there are too many variables involved for it to be a controlled test. If the test isn't controlled then the results might as well just be plucked from a hat.

All I wanted to find out was whether UDP transmission was faster on the Killer NIC than on my on-board NIC since game traffic relies on UDP protocol. Testing this on a LAN resricted to 100Mb/s won't prove asnything as it is still a LAN.
Report Comment
 
 
# RE: RE: RE: RE: VisionTek Killer HD5770 Combo CardOlle P 2011-02-21 04:03
"Testing this on a LAN resricted to 100Mb/s won't prove asnything as it is still a LAN."
It will prove what happens when the top speed is governed by an external bottleneck. Restricting the speed to 10Mbps should be even more realistic, since that's closer to the real world limits.
Your own test showed that the motherboard's NIC ran both TCP and UDP at about 95 Mbps, which is what you get with a 100Mb restriction.
Report Comment
 
 
# RE: RE: RE: RE: RE: VisionTek Killer HD5770 Combo CardSteven Iglesias-Hearst 2011-02-21 04:40
"It will prove what happens when the top speed is governed by an external bottleneck. Restricting the speed to 10Mbps should be even more realistic, since that's closer to the real world limits."

Since both NIC's are capable of near 100Mbps on TCP and UDP then the results would be equal and inconclusive.

The only real way to test would be to set up a game server and control the load on that server, connect to it with two identical machines at the same time (one with killer NIC one without) and record average frame rates and ping and also run the same test as detailed within this article.

My budget won't allow for that, so I test within my means.
Report Comment
 
 
# usefulldie 2011-03-26 03:14
6870 I like it for price and power consumption I think for best buy
Report Comment
 

Comments have been disabled by the administrator.

Search Benchmark Reviews
QNAP Network Storage Servers

Follow Benchmark Reviews on FacebookReceive Tweets from Benchmark Reviews on Twitter