| VisionTek Killer HD5770 Combo Card |  |
| Reviews - Featured Reviews: Video Cards | |
| Written by Steven Iglesias-Hearst | |
| Tuesday, 15 February 2011 | |
|
Page 2 of 19
Closer Look: VisionTek Killer ComboThe Killer HD5770 combo card is much larger than the regular Radeon HD5770 coming in with a 10" long PCB, this might be something one would need to consider when looking to purchase such a card. Due to its spec's the Killer HD5770 combo card would be ideal for a HTPC/Gaming PC but due to its size it may not fit in some Micro-ATX / HTPC cases.
The Killer HD5770 combo card has a visually appealing design, this is somewhat wasted when the card is installed into your system even if you have a windowed side panel as this side will be facing down, coolers should be more about function than looks but if a good balance can be found without sacrificing one for the other then all is well.
Looking from the top we can see two CrossFire fingers which means the Killer HD5770 is good for 2- or 3-way CrossFireX, the HD5770 is a relatively low end card but when doubled up it can compete with is bigger brothers. The PCI bracket is vented but I wouldn't think a lot of the hot air will be leaving this way.
The Killer HD5770 requires a 450w (or greater) PSU with one 75w 6-pin PCI-E power connector, or a 600w (or greater) PSU with two 6-pin PCI-e power connectors for CrossFireX mode. You may also notice that there is no active cooling on the VRM section, this isn't good for extreme overclocking as this suggests that there is no overvoltage support, but there should definitely be some headroom for a nice moderate overclock. The design of the shroud directs the intake of air over the heatsink and then towards the front and back end of the PCB thus cooling the other components.
Removing the cooler and shroud reveals a very busy PCB, we will have a look at this in more detail in the next section. It's good to see that VisionTek haven't used too much TIM on the GPU, it is still more TIM than I would personally use but it is nowhere near as extreme as I have seen in the past. The cooler uses a copper plate to make contact with the GPU transferring the heat to the aluminium heatsink which is cooled by the airflow from the fan.
The back of the PCB is well populated, the soldering quality is excellent and the PCB is nice and clean as you would expect, there are also a few chips on this side we will be taking a more detailed look at in the next section, so let's move on and get a more in depth look.
|
|
Comments
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.
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.
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.
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.
"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...
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.
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.
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.