Table of Contents
The war has begun and battle is being waged. AMD started this particular battle with their HD 7790. NVIDIA is here today hoping to win it with their newest addition to the Kepler line – the NVIDIA GTX 650 Ti BOOST.
As evident in AMD’s slides, the HD 7790 was aimed at the GTX 650 Ti. Now it gets more interesting. NVIDIA has introduced GPU BOOST (v.1.0) to its GTX 650 Ti in the hopes of enticing those all-important mainstream 1080p gamers. Can NVIDIA take the day? Let’s find out!
Specifications & Features
The GTX 650 Ti BOOST uses the same GK106 GPU contained in the GTX 660. The GPU in the GTX 650 Ti BOOST is reduced to 768 CUDA cores and 64 texture units (versus 960 & 80, respectively on the GTX 660). Both GPUs have 24 ROPs. While the GPU is cut down a bit, it operates at the same 980 MHz base / 1033 MHz boost as the GTX 660. The new GPU’s TDP is 134 W. Memory is also identical, with 2 GB of GDDR5 operating at 1502 Mhz (6008 MHz quad-pumped) with a 192-bit interface. There will also be 1 GB versions available.

The actual stock boost frequency on this particular card ends up coming in at 1071.3 MHz, as shown via monitoring in 3DMark 11 and HWBot Heaven Xtreme.

As you can see, the GTX 650 Ti BOOST is more closely related to the GTX 660 than the GTX 650 Ti. As such, it performs considerably better than the GTX 650 Ti.

This slide is a bit over stated, so let me break it down. If you wait five years between upgrades, you can get a large increase in performance. that applies to GPUs from both sides of the fence, making this a little silly. NVIDIA says people usually wait two years or greater to replace their GPUs in this product segment, which is why they like to mention this.

This graph is a bit more relevant, comparing the GTX 650 Ti BOOST with the HD 7850 in NVIDIA’s internal testing. You also see the target for this GPU is not, in fact, the HD 7790 that came out last week, but the HD 7850, which hints at its pricing. The GTX 650 Ti BOOST with 2 GB of memory (as tested in this review) has an MSRP of $169. That may increase depending on board partner choices like overclock and cooler, but that’s the base pricing.
That said, they do expect to release a 1 GB version in April that has an MSRP of $149, just like the HD 7790. Did I mention a war was being waged? I love it!

In our conference call, NVIDIA also says their open beta of the GeForce Experience is getting positive feedback. They say they’ve even tweaked the experience based on user feedback. If you’re not familiar with it, GeForce Experience is a program that will monitor your PC’s GeForce driver version and notify you when upgrades are available.
More importantly, it can automatically set games’ video settings to match your system’s performance level. Most overclockers prefer to set their own settings, but if you want to give it a try, check out the GeForce Experience page.

AMD has a strong hold on the biggest titles for their gaming bundles, but NVIDIA isn’t letting that stop them from sending something with their cards. Rather than giving free games, they’re using Free-to-Play (F2P) games and giving you $75 in in-game credit.

Now you’ve seen the company line, let’s look at the GPU itself.
Meet the NVIDIA GTX 650 Ti BOOST
Unlike the reference GTX 650 Ti, the GTX 650 Ti BOOST comes with a bit stronger squirrel-cage cooler. As you can see it also has a single 6-pin PCIe power plug.

Photo shoot time.
Not a bad looking card at all so far, but you probably won’t see too many identical models; most will have custom coolers placed on them by board partners.
As far as video outputs, there will be dual DVI and single HDMI & DisplayPort outputs.

Let’s take the cooler off and see how it looks underneath.
Under the Hood
The good thing about NVIDIA is that they use actual thermal paste unlike that chunky mess ASUS used on the HD 7790. Both are effective I suppose, but from a reviewer’s standpoint, if nothing else it’s easier to remove.
Like the reference GTX 660 Ti cooler we saw, the shroud is much larger than the heatsink itself, which is really a small slab of copper attached to aluminum fins. It’s definitely not much of a heatsink as GPU heatsinks go, but this is a budget GPU. Regardless, you’ll see better solutions than this on most, if not all partner cards.
Ahh, here’s the GTX 650 Ti BOOST in the flesh, sans cooler. It’s a short PCB like the rest of the low- and mid-range GTX series (as well as the HD 7790 we reviewed Friday).

The power section is also very similar to the HD 7790, with four phases for the GPU and one for the memory (you can see the memory phase in the photo above left, to the left of the far left memory chip).

Speaking of memory, NVIDIA continues to go with Samsung GDDR5 memory clocked at 1502 MHz (6008 MHz quad-pumped).

Lastly we have the GK106 GPU.
Overall, it looks the part of a solid budget card. Time to install it in a system and see what we can get out of it.

Test Setup
Our test setup includes a 4.0 GHz i7 3770K with memory clocked at DDR3-1866 / 9-9-9-24. Competition comes courtesy AMD and NVIDIA above and below the GTX 650 Ti BOOST’s price range.
CPU | i7 3770K @ 4.0 GHz |
MB | ASUS Maximus V Extreme |
RAM | G.Skill TridentX DDR3-2600 @ 1866MHz 9-9-9-24 |
GPUs | ASUS HD 7790 DirectCU II OC EVGA GTX 650 Ti SSC Galaxy GTX 660 GC PowerColor PCS+ HD 7850 |
OS | Windows 7 Professional x64 |
**Important note about drivers** The HD 78xx cards graphed in this review have not been re-benched with the 12.x (or later) Catalyst drivers, which showed large performance gains over previous Catalyst versions. We are in the process of reviewing a newer HD 7870 and a HD 7850 is on the way for review as well, so you’ll get those numbers in the future.

Installed on the Maximus IV Extreme, it’s not a bad look for a reference card at all.
Cooler Performance & Power Consumption
The GTX 650 Ti BOOST comes in right where it should, with cooler performance very similar to its slightly more powerful GTX 660 brother. Temperature wise, it just doesn’t compare to the HD 7790, but neither does the cooler itself, so that’s to be expected.
Being a blower-style cooler, it definitely doesn’t compare to the acoustics of the partner-designed cooling on the HD 7790. While the latter couldn’t ever be heard over my under-volted radiator fans, it was pretty obvious when this cooler kicked in. That said, it’s not really a fair comparison, because that was a partner-designed cooler and this one isn’t designed to compete with it.

Power-consumption wise, it is and acts like a cut down GTX 660, trimming a few watts off the top. The card is driving power consumption here too. While the HD 7790’s peak came when the CPU was loaded (in the 3DMark 11 Combined test), the GTX 650 Ti BOOST’s peak was during the GPU tests.

It fits right where it should in the power consumption chart. While it’s not as efficient as AMD’s latest offering, if it out-performs it, that wattage might be forgiven. We’ll just have to see.
Overclocking
Overclocking comes to us courtesy of EVGA’s PrecisionX. The total overclock, with voltage control, was +110 MHz on the GPU and +300 MHz (in PrecisionX, which translates to +150 MHz in reality) on the memory. It’s not as strong MHz wise as the HD 7790 and this overclock had voltage control to help it out. However, it’s not a bad overclock by any stretch.
These are the settings used to obtain the overclock. Disregard the Temp Target; this was the version of PrecisionX already installed on my system from the Titan review. Rest assured, this card has GPU Boost 1.0, not 2.0.

Interestingly, this overclock was the max 24/7 stable, but it was not hammering up against the card’s power target. It picked a frequency and stuck to it. In this case, the actual boost frequency ended up being 104.5 MHz over stock, and it kept it throughout the benchmark.

Stability testing comes to us courtesy 3DMark 11 and HWBot Heaven Xtreme.
Not a bad clocking card at all and the overclock you set keeps on giving, without running into the pesky power limit and throttling back all the time.
Performance Results
All of our benchmarks are run per our GPU testing procedure guidelines, which you can check out right here. Long story short: benchmarks are run at their default settings and games are run with everything maxed. Because of the previously mentioned driver notes re: HD 78xx GPUs, I’ll keep my comments between the GTX competition and the HD 7790.
Synthetic Tests
3DMark03 is a known variable when it comes to Kepler cards. They just don’t have what they used to for this now-ten-year-old, which is pretty irrelevant to anyone but benchmarkers. However, we still keep it for old times’ sake (and because benchers are a solid chunk of our audience).

Moving to more relevant benches, it looks like the GTX 650 Ti BOOST steps up to the plate and takes out the HD 7790 in DirectX 10.

Well, look at this. Looks like we might have a battle on our hands for benchmarking. The HD 7790 wins the day in 3DMark 11.

The HD 7790 is quickly unseated in Heaven Xtreme, with the GTX 650 Ti BOOST taking a healthy lead.

The newest 3DMark is a back and forth even within the same benchmark. The GTX 650 Ti BOOST wins at stock, but the further-overclocking HD 7790 takes the overclocked prize.

In case you’re wondering, 3DMark Fire Strike wasn’t included in the last review because there weren’t enough results in the HD 7790’s price range.
Game Tests
Now we reach the point where the rubber meets the road – game tests. Getting right to it, the GTX 650 Ti BOOST gets a solid “W” in Aliens vs. Predator.

More of the same, with results over ten FPS better in Batman: Arkham City

Even Battlefield 3 falls in favor of the GTX 650 Ti BOOST.

Civilization V is closer and the overclocked result takes a step back (not uncommon for this game in cards from both sides), but still doesn’t lose to the HD 7790.

Dirt 3 continues the slightly-over-10-FPS lead.

GPU killer Metro 2033 is much closer, but the GTX 650 Ti BOOST still comes out on top.

In every game test, the GTX 650 Ti BOOST beats the HD 7790. That’s to be expected for $20 more and it does precisely what NVIDIA set out to do – stave off AMD’s newly increased power in this market segment.
Frame Times – a Short Introduction
There has been a lot of talk on the interweb recently about frame times, and this is the first step in considering including a frame time component in our reviews. It will start with with just me and just one game. It is a point of discussion with our editors and when Haswell comes, if we upgrade our test platform, we may include Frame Time testing in the future. If you have an opinion on frame time measurements, please feel free to share it in the comments.
If you’re unfamiliar with frame times, it is how long the GPU takes to render a frame. FPS is a good measurement to tell how many frames are being produced per second on average, but as we all know, FPS can vary as you’re gaming. That’s because your GPU often takes longer than that average to produce the frame. The simplest explanation I’ve seen came from SKYMTL, who writes for Hardware Canucks from his research:
- Below 20ms: unnoticeable
- Rapid, cyclical fluctuations above 25ms: you’ll notice it
- Sudden “spikes” (non-cyclical) above 40ms: you’ll notice it
- Grey area between 21ms and 25ms: debatable. Some will notice it while others won’t.
What I went with in our first frame time graph is the first 3000 frames of our Battlefield 3 run. As you can see, both of these GPUs have their moments above 30 s, some as much as 45-50 ms. The big spikes came from the GTX 650 Ti BOOST and it spent more time north of 30 ms than the HD 7790 did.

That said, don’t make a mountain out of a mole hill here. This is a lot of data thrown into a small graph and it looks intimidating. Both of these GPUs felt fine coming out of my monitor, even on Ultra settings, which is what you see graphed, and that’s what’s important.
Final Thoughts & Conclusion
The war wages on, but this battle between the newest GPUs from NVIDIA and AMD has come to a close, with the NVIDIA offering coming out on top.
In this market segment, money talks. The GTX 650 Ti BOOST is listed as having a (reference) MSRP of $169. That’s $20 more than the HD 7790’s $149, or 13%. However, that 13% price premium translates to 17-31% FPS increase, depending on the game. AvP, Batman, CivV & Dirt 3 come in at 27-31% higher FPS. That’s nothing to sneeze at. In fact, those games increase double FPS per dollar over the HD 7790.
Now, that said, the HD 7790 we reviewed was quieter, cooler, overclocked better and sipped less power. If it had been able to access voltage control it would likely have overclocked even further. However, we’ll have to see what NVIDIA’s board partners come out with before making any snap judgements on temperatures and acoustics, as reference coolers on the GTX 650 Ti BOOST won’t really be coming to market, just like with the HD 7790.
Also important to consider is that the HD 7790 comes with a copy of Bioshock: Infinite. Sorry NVIDIA, but to me, a brand new premium title is worth more than some credit to spend in some F2P games. The latter may have a large fan base, but a full game is better than some weapon upgrades. Of course, if you have no interest in Bioshock: Infinite and just want more horsepower for the games you have, there goes that advantage.
A significant problem for AMD is that the GTX 650 Ti BOOST (both reference and overclocked models) will be on sale at e-tailers today. AMD’s HD 7790 “launch” was a paper launch, with availability expected in early April. NVIDIA expects to get their 1 GB, $149 version of the GTX 650 Ti BOOST out in early April as well. The pricing and positioning wars continue on and that’s great news for everybody in the market for a GPU.
Reference vs. non-reference acoustics and temperatures aside, the GTX 650 Ti BOOST does one thing well – it beats its competition in this market segment with pure horsepower. When you’re looking at saving every dime you can but still game happily at 1080p, that’s the most important measurement.
26 replies
Loading new replies...
New Member
Member
Water Cooled Moderator
Member
Water Cooled Moderator
Member
Water Cooled Moderator
Member
Member
5up3r m0d3r4t0r
Join the full discussion at the Overclockers Forums →