GeForce GTS 250 Roundup (ASUS, Gigabyte, Sparkle, EVGA)

[SKYMTL]

Tom Clancy’s Hawx (DX10)

Tom Clancy’s Hawx (DX10)

Even though HawX has a built-in benchmarking tool, we found that it did not provide accurate enough results needed to reflect what actual in-game performance would be. Thus, we played through the first 6 minutes of a mission and recorded the results with FRAPS.

1680 x 1050

1920 x 1200

2560 x 1600

 

[SKYMTL]

Heat & Acoustics

Heat & Acoustics

For all temperature testing, the cards were placed on an open test bench with a single 120mm 1200RPM fan placed ~8” away from the heatsink. The ambient temperature was kept at a constant 22°C (+/- 0.5°C). If the ambient temperatures rose above 23°C at any time throughout the test, all benchmarking was stopped.

For this test we use the 3DMark Batch Size test at it highest triangle count with 4xAA and 16xAF enabled and looped it for one hour to determine the peak load temperature as measured by GPU-Z. For Idle tests, we let the system idle at the Vista desktop for 15 minutes and recorded the peak temperature.

Gigabyte GTS 250 1GB OC

When it comes to both load and idle temperatures, the Gigabyte GTS 250 with its Zalman cooler was simply unmatched. Things were a little less rosy when it came to acoustics since the fan on the VF1050 can’t regulate its speed based on core temperatures. This leads to this card being (subjectively) the noisiest of the roundup but its acoustical footprint is small enough that it was easily drowned out by the two 120mm Yate Loon fans installed onto the tech station. Unfortunately, this is the only card which we couldn’t control its fan speed through software like Rivatuner.


Sparkle Calibre GTS 250 X250

We didn’t have very high hopes for the heatsink on the Calibre X250 and its performance bore this out to some extent. In its straight position with the fins as close as possible to the PCB, the fans had to operate at a constant 80% of their rated speed just to keep the core at a toasty 78°C. Meanwhile, when we elevated the two cooling assemblies, the peak temperatures dropped 4°C while the two little fans went to 74% output. High fan speed coupled with relatively near-80° temperatures tell us that the Dual Fly heatsink is very close to its thermal capacity and is struggling to keep the core within an acceptable temperature range.

Personally, I think there are two real stars of this show: the fact that elevating the heatsink assemblies really did decrease temperatures and that even when working their butts off, the fans stayed extremely quiet. The only real complaint we have about the fans is that at higher RPMs their slight vibrations are amplified by the copper fins into a very faint mechanical resonance.


ASUS GTS 250 1GB Dark Knight

At the beginning of this roundup, I expressed some doubts regarding the efficacy of the heatsink used on the Dark Knight. Well, ASUS proved me wrong since this heatsink literally competes blow-for-blow with the Zalman unit found on the Gigabyte card. In the end it lost by a few degrees but that didn’t stop it from impressing the hell out of me.

Noise-wise, the small fan usually keeps to its own company while happily gobbling away at temperatures but every now and then it will ramp up speed for a few seconds. This speed ramp-up doesn’t increase fan speed to the point where it will be noticeable fro within a closed-case environment.


EVGA GTS 250 1GB Superclocked

Considering this card uses what many would call the “reference” GTS 250 heatsink, the temperatures it showed where very good. Not once did its fan speed increase to levels where we could hear it and even though this is the largest heatsink of the bunch, the fact that it exhausts all of the hot air outside the case is a welcome addition.

 

[SKYMTL]

Power Consumption

Power Consumption

For this test we hooked up our power supply to a UPM power meter that will log the power consumption of the whole system twice every second. In order to stress the GPU as much as possible we once again use the Batch Render test in 3DMark06 and let it run for 30 minutes to determine the peak power consumption while letting the card sit at a stable Windows desktop for 30 minutes to determine the peak idle power consumption. We have also included several other tests as well. Please note that after extensive testing, we have found that simply plugging in a power meter to a wall outlet or UPS will NOT give you accurate power consumption numbers due to slight changes in the input voltage. Thus we use a Tripp-Lite 1800W line conditioner between the 120V outlet and the power meter.

Considering this roundup featured cards which were both overclocked and had stock speeds, most of the power consumption figures were well within expectations but there were some exceptions. You should also note that the GTS 512MB card used here is simply a reflashed 9800 GTX+ which explains why it is not showing the same idle power consumption levels as the other GTS 250 cards.

For some reason, the ASUS card had slightly higher power consumption than the reference GTS 250 1GB. Whether this was due to lower-quality component choices can be debate somewhere else but we reran the tests 4 times with the same results so this is more than just the usual margin of error.

As is quite apparent, the Sparkle card does not downclock when in 2D mode though in their SPAtune program, you can manually switch the clock speeds to “Green” mode to lower power consumption. This lack of downclocking led to higher than expected idle power consumption numbers.

Finally we have the Gigabyte card; it has slightly higher than reference idle numbers but its power consumption under load is excellent. It seems like the claimed efficiency with Gigabyte’s Ultra Durable VGA is more than just marketing hype.

 

[SKYMTL]

Overclocking Results

Overclocking Results

What do we consider a stable overclock? If the card can pass our test which involves a 20 minute loop of 3DMark06, 30 minutes of playing Far Cry 2 and an additional 30 minute gameplay sequence in Left 4 Dead, it is considered stable.

Since overclocking can vary from one card to the next within the same family of cards, you should be aware that these results will not be considered when voicing our final opinions about these four cards. As such, we will show you the final overclocking results but leave our opinions to ourselves and let you guys form your own conclusions.

Gigabyte GTS 250 1GB OC

Final Overclocks

Core: 805Mhz
Shader: 2002Mhz
Memory (DDR): 2432Mhz

Click image to zoom​

Sparkle Calibre GTS 250 X250

Final Overclocks

Core: 813Mhz
Shader: 2041Mhz
Memory (DDR): 2530Mhz

Click image to zoom​

ASUS GTS 250 1GB Dark Knight

Final Overclocks

Core: 808Mhz
Shader: 2004Mhz
Memory (DDR): 2404Mhz

Click image to zoom​

EVGA GTS 250 1GB Superclocked

Final Overclocks

Core: 829Mhz
Shader: 2032Mhz
Memory (DDR): 2644Mhz

Click image to zoom​

 

[SKYMTL]

Conclusions (Gigabyte & Sparkle)

Conclusions

Considering that before this roundup came to fruition, we hadn’t reviewed a single GTS 250 card I didn’t really know what to expect. Granted, specs-wise the GTS 250 is nothing more than a rebadged 9800 GTX+ but because of this, Nvidia’s board partners can be adventurous in the way they go about designing their cards. They already knew what was and wasn’t possible with the G92b architecture and have used this opportunity to put forward some damn interesting designs. Power consumption also seems to have been lowered over the 9800 GTX+ even though in this article we flashed the older card to a GTS 250. This also goes to prove that a mere BIOS flash will not magically lower the power consumption of older cards.

Naturally, the GTS 250 is in tough against the HD 4850 with both cards retailing for approximately the same price. However, we feel that the Nvidia 1GB card seems to have a slight edge over its equally-equipped competitor.

As much as possible, we have tried to make sure that this roundup has focused less on pitting one GTS 250 against another and looking at each card separately. Naturally there are some which excel in certain areas but when push comes to shove we wouldn’t hesitate in recommending most of the four cards featured in this roundup. They all offer a good amount of performance for your hard earned buck and there isn’t really much more to ask for other than that. With that in mind, let’s get on to the individual evaluations.


GIGABYTE GTS 250 1GB OC

GIGABYTE really hit the nail on the head with this card by flexing their engineering know-how to good effect. We see this with additional efficiency through the use of their Ultra Durable VGA initiative. Its Zalman heatsink performed above and beyond the call of duty by cooling the core to some pretty impressive levels. We have seen some complaints about fan noise but we didn’t hear it above the noise of our other fans. That coupled with the fact that it was also the shortest card of the bunch and carries all of the necessary outputs and connectors for HDMI video and audio makes it a perfect card for an HTPC. Let me tell you, a GTS 250 1GB like this should definitely be considered by those of you are looking to put some serious gaming muscle in your home theatre PC system.

Ummm….no. ​

If we would stop right here and now, GIGABYTE would probably win our Dam Good Award but there is one thing that stops us from giving it out: the fact that this card isn’t overclocked as claimed. To us, honesty to consumers comes before everything else and that includes performance. Thus, when you claim to have overclocked memory on the front of your box, in your marketing materials and throughout retailer listings, you better be damn sure of what the reference clocks are. Unfortunately for GIGABYTE, that wasn’t meant to be. While this may have been a simple misunderstanding on GIGABYTE’s part, a claimed overclock can and will change a customer’s perception of a product and that is unacceptable in this case. This is all compounded by the fact that GIGABYTE’s site is completely lacking in any reference to clock speeds so the consumer really does go into this purchase blind.

Overall, we highly recommend the GIGABYTE GTS 250 1GB OC to those of you looking for an all-round great performer that has amazing cooling potential. Even though it features a high-end heatsink, extremely good efficiency and a perfect selection of connectors, it actually carries the lowest average price of all the cards featured here. For that reason, it gets our Dam Good Value Award.

Sparkle Calibre GTS 250 512MB X250

Being the only 512MB card in this roundup meant that the Calibre GTS 250 X250 had an uphill battle from the very first benchmark. Having the highest clock speeds of all the competitors meant incredibly impressive results through many of the tests where it routinely bested the EVGA 1GB Superclocked card. Unfortunately, that 512MB came back a few times to bite the Calibre card in its butt. There were many situations at higher resolutions where its performance dropped to the earth like a meteor due to the memory bottlenecking the GPU.

Its Dual Fly cooling system did prove to be quite effective at cooling off the card at its pre-overclocked speeds but once we began longer gameplay sessions, its limitations came into focus. Stability wasn’t an issue at all but judging from the 80% speed the fans were running at and the 78°C core temperature after an hour of testing, we believe that the heatink was very close to its threshold. If you are using a case without much airflow, you may want to keep a close eye on temperatures.

I think what surprised me most was how quiet this card was even with its fans running near their maximum speeds, it was nothing more than a background whisper. Another pleasant surprise came after realizing that Sparkle’s claims of an elevated heatsink reducing temperatures was more than just window dressing; the heat really was lowered.

If you are looking for gaming at or under 1920 resolution, then you should seriously consider looking at this card. Due to its unique (and quiet) cooling solution, its amazing performance for a 512MB product and all the outputs anyone could want, we give the Sparkle Calibre GTS 250 X250 our Dam Good Award. The only thing I find myself wishing for is better availability so the price can come down.

 

[SKYMTL]

Conclusions (ASUS & EVGA)

Conclusions

ASUS GTS 250 1GB Dark Knight

Let’s put our cards on the table right away: this ASUS card’s performance disappointed us. We are used to ASUS coming out with all guns blazing and when I saw the heatsink on this thing, my hopes seemed justified…until I looked at the specs. In my push to get the roundup done, I surged ahead with testing before even looking at the specs for many of these cards so when the Dark Knight started posting scores which were lower than a stock GTS 250 1GB, I was puzzled. A quick GPU-Z click later I found out that the memory was operating at 2Ghz which is 200Mhz below Nvidia’s reference specs for 1GB card. This translates into sub-standard performance for the ASUS card when compared to the other 1GB models.

To make matters even worse, the stock GTS 250 512MB uses memory operating at 2.2Ghz. Guess where that leaves the ASUS Dark Knight: trailing the stock 512MB card in several tests. Did we mention that this card is priced higher on average than the Gigabyte model we tested? It’s embarrassing really.

Luckily for ASUS, it isn’t all doom and gloom since I am happy to report that I was completely wrong in my initial assessment of the Dark Knight’s heatsink. It was among the best of the best when it came to cooling off the G92b core and it went about its business at sufficiently low noise levels. We also have to give ASUS credit for including the 10% off game voucher as the discounts apply to games that are actually worth the money.

Unfortunately, a kick-ass heatsink and 10% off select games does nothing to diminish the reality of what this card is: overpriced and underperforming. Just remember that if ASUS actually decides to cut this card’s price, it could be the perfect solution if you are looking for a 1GB GTS 250. If anything, the Dark Knight proves that you really, really have to do your homework before making a purchase.


EVGA GTS 250 1GB Superclocked

We are used to seeing EVGA right in the thick of things performance-wise and the GTS 250 1GB Superclocked did not disappoint. In every game we tested it posted results which were well ahead of the reference GTS 250 1GB and in some rare cases its advantage in framebuffer size allowed it to close with the GTS 260 216. Its dual slot cooler is definitely an advantage as well since (unlike other cards in this roundup) it is able to exhaust hot air outside of your case and it does so in near-silence. This would all be for nothing if the Superclocked’s temperatures were high but the heatsink proves itself to be more than adequate for cooling off the overclocked core.

Indeed, there is very little not to like about EVGA’s entry into this roundup. The shadow of cost-cutting did rear its ugly head with the anemic accessory package but that was the most glaring fault the Superclocked Edition showed…until we saw its price. We understand paying a premium for higher clock speeds is inevitable but EVGA priced this card high enough that it is now in the same bracket as the GTX 260 216. Considering the performance difference between these two cards in the vast majority of benchmarks we feel the mere $30 difference between the two cards is unjustified.

EVGA’s entry into the GTS 250 1GB arena is definitely something for them to be proud of as it provides performance in spades at every resolution and EVGA’s customer service is literally second to none. Those two facts alone should take it to the top of your list…as long as you can find the GTS 250 Superclocked for more than $40 less than a GTX 260 216.

Our thanks to all the manufacturers who participated in the roundup.

http://www.hardwarecanucks.com/forum/video-cards/17697-gts-250-roundup-comment-thread.html​