Comparative: Dell 2GB 680M OC vs Clevo 4GB 680M OC vs Dell AMD HD7970M 2GB OC

Hi guys!

Me and my mate Riri-Fifi got some spare time today to run some proper tests and benchmarks as we are lucky enough to own very nice gaming machines with new gen GPUs that we can clock properly!

The purpose of this thread is to show you the capability of each new gen card, with some proper overclocking that we use for 24/7 gaming use. Riri is very luck as both his 7970M's can get cranked even higher but we have tried to keep reasonnable overclocks that can be achieved by anyone of you with a bit of effort Also this can be used as a database for those who are keen on getting the best of their system!

We chose the very usual benchmarks that didn't require too much of downloading (sorry we don't own Crysis 2 nor Metro2033) and I hope you will like them.

The main reason also we did this is because last night, Riri and I were doing some Unigine Heaven benchmark 3.0 runs and noticed that with his much lower clocks with the Dell 680M (950/1000), he was getting similiar/higher score than my 4GB Clevo 680M @ 1000/1250! That was pretty insane to me even though I knew the Dell 680M 2GB is quite a monster @stock and with slightly bumped clocks (and no throttles)!

We wanted to show that the 2GB 680M is very capable and runs very very cool! Although and that's a shame, we would have liked to test it and compare it with its 4GB version in-game but we were lacking of time!

The 7970M isn't far behind at all - to me it still has the best bang for buck by far!

If we get our hands on Crysis 2, Metro2033 and Dirt3, that could be the next tests!

I'll try to keep the table updated but any suggestions, please let me know!

Here are the complete results:



And the screenshots (currently uploading):
**3dmark11
Performance Preset

Spoiler

Spoiler

Spoiler

Extreme Preset

Spoiler

Spoiler

Spoiler

**3dmarkVantage
1280x1024

Spoiler

Spoiler

Spoiler

1920x1080

Spoiler

Spoiler

Spoiler

**AVP Benchmark
Settings: 1920x1080, default

Results:

Spoiler

Spoiler

Spoiler

**LostPlanet 2
Settings:


Results:

Spoiler

Spoiler

Spoiler

**Resident Evil 5 Benchmark
Settings


Results:

Spoiler

Spoiler

Spoiler

**Stalker Call of Pripyat Benchmark
Settings:


Results:

**The Last Remnant
Settings 1920x1080, default

Results:

Spoiler

Spoiler

Spoiler

**Unigine Heaven Benchmark 3.0
Settings


Results:

Spoiler

Spoiler

Spoiler

 

Hey, you guys did a wonderful job on this. Thank you for the time and effort, and for sharing the results. That Dell 2GB GTX 680M is a real beast.

 

Really curious about how you get the gpu boost enabled in AW since someone told me that Dell doesn't support that feature on 680m. I'm waiting for my 680 now and I want to set a turbo for it instead of overclock the defult clock straightly, is that a good idea?
Thanks for your response in advance

 

To be honest I don't know how valid your results are for a few reasons:

1) Different voltage and clocks
-----> The chances of throttling due to high temperatures can skew results further, making a higher overclocked card look worse than one with lower clocks.
2) Different video drivers
-----> This is a big one. The Dell card you tested was using more updated drivers than the Clevo version, which could also skew results.

If you don't control for as many variables as possible, your results become much less valid. Regardless, thanks for taking the time to run these tests and put that chart together; it will likely get referenced quite frequently in the future!

 

With the new M17xR4 BIOS, I think it enables the gpu boost clocks to crank up properly so no worries! We had to get SVL7's hand to get the base clocks set to 950/1000 since in the R3 GPU boost is disabled.

Interesting - we own both cards and have tested different drivers - we used the best drivers for the card we think.
The new 306.02 gives a pretty nice boost with the Dell 2GB 680M while it gives horrible crashes and issues in-game with the 4GB 680M.
The 304.79 are the best drivers to me for gaming and perhaps not the best for benchmarking but it's not that important.
What we wanted to highlight is the potent of the Dell 680M as I think many of us got disappointed since the gpu boost and the throttles have been disabled for the former and present for the latter.

Now regarding the throttles, at least I can tell you that there are none with the 4GB 680M because I use the QS/ES MSI vBIOS - the card is exploited at 100% and the core clock stays at 99.8% of the time at 1002mhz.

The results are quite accurate, except for TLR where my gpu utilization was barely going over 75% which results in a much lower fps than expected.

I think that clarifies what you tried to say.

 

Nice report.

Interesting food for thought.

Thanks for the time invested to show us! +rep

 

I think the results are valid on face value for the systems they're installed in. It might not be a lab-grade controlled science project, but it still offers a great point of reference for people trying to cut through all of the noise and see how these beasts run.

The thing about drivers is, using the same one could actually skew the results. The point of the exercise was not testing drivers, but testing hardware that has been tuned to run its best. You should always use the driver that works best. Sometimes that means using a different driver on one machine than you do on another. So, using the driver that works best gives the better test results. Think about it... are you going to use a driver that hinders performance just so it can be the same version as someone else's?

Edit: I read this ( below) after posting, but it says exactly what I was trying to convey.

 

Is the boost automatically be set to 850MHz or I have to do it myself? I haven't used the Kepler card ever so I'm not sure will the VBIOS PATCHER (which I'm using it now for making 7970m custom VBIOSs) work on it both in editing the base and boost clock?

 

@Mr.Fox
Don't pretend you haven't read my edited post before posting yours Yeah this table shows you the best of our systems (and possibly your systems!), with our 24/7 settings!

edit: my temps could look a bit/quite high and that's because I've got a 3820QM ES that runs a bit hot - and because MSI parts share a single powerful fan, the heat starts to build up between both GPU and CPU causing the GPU temp to increase a lot.
With the 3610QM, I got like 5-7 degrees lesser under long heavy load!

Yeah the boost automatically increases when you increase the core clocks! As easy as before! You can't use the SVL7's vbios patcher unfortunately, this is only meant to be used with the 7970M.
Just use NVInspector or MSI Afterburner to bump your clocks

 

Hello

Thx @ Mr.Fox

 

Consider me re-educated! I see where you're coming from about drivers, and it looks like you used a BIOS which prevents throttling to circumvent that issue.

Your results are also quite telling. The 7970M is solid competition for the 680M, and Dell's 680M doesn't seem gimped at all compared to its Clevo sibling. They seem pretty even as far as I can tell, though if you can find an overvolted bios for the Dell card you might be able to push it further, judging by the temperatures you recorded.

 

My 2 Gb. 680M .. This OC is 893 / 2106 Mhz. on stock voltage of .9870 V. Highest GPU temp. I.ve seen is 72*C on TW 2 (2 hr. gaming) on this same OC. I'm pleased with the temps. I could finally say that it "owned" TW2 non-enhanced, very pleased with the graphics and smooth gameplay now, in any map.









One of the favs..



I have Metro 2033 (I forgot). I will install and do the in-game benchmark on diff. clocks.

My Metro 2033 in-game benchmark results.



On stock clocks:




At 893 / 2407:




At 893/2407 with DOF:



DOF, MSAA 4X and Very High Quality Texture are very taxing. My NVidia Control Panel is also set to high quality.

 

I thought the Dell 2GB cards doesn't OC and crashes at every voltage?

 

That Dell GTX 680M is a good card however 1920X1080 should be 800X600. I mean its not a big res. You wanna see that Clevo show its teeth its needs more res.

1920X1200 you will see a little but 2560X1440 is where its at. I know most people dont have that but that's when the 2gb vs 4gb fight gets fun. And the Clevo will handle 3D a little better on games with big textures.


Cant say the cards are even because 2gb vs 4gb. 4gb cards were made for games with huge textures but most people wont even see a diff because most wont use over 1gb -1.5 even on their demanding games. Its when you play in 3D with high res and huge textures the extra vram comes in handy and really shines.

We should do scaling with resolution as well as with 3D enabled.

 

No they work fine in games and single card situations even better than sli when it comes to achieving max potential out of the dell cards.

 

Widezu said that if you OC the core you get throttle and if you OC the mem you get a display driver crash. And if both means a crash at any voltage.

Strange how dell also had throttling problems with the 580M

 

4GB is only really useful for 3D however none of us Clevo users have been able to get the 3D working so far.

 

Thanks to Pau1ow I'm now able to compare my dell 680M's Unigine 3 results with their cards. My 3DMark 11 benchmark results has been screwy..




This is at 893 / 2106 Mhz.

This was the first benchmark at version 3, my 2nd try yielded 55.8 FPS. This atleast gave me some peace of mind due to my confusing Mark 11 results. My Intel HD card is adding issues to these benchmark softwares.

 

My clevo card runs 3d just fine and I use the latest driver release without any crashes like the OP experienced.

Sent from my GT-N7000

 

Note that "Clevo users" not "Clevo card users" . For Clevo's P150EM/P170EM the display ports are connected to the integrated GPU.

 

Thats because you dont have optimus.

Optimus rules out having a 3d screen.

This is why Sager are going to be adding it to the 370EM which does not have optimus since it too supports dual graphics cards.

 

And a 5 minute battery life? xD

 

It's not THAT bad, but I would expect 90-120 minutes.

That's the price you pay I guess.

 

I'm still wondering if the Dell 680M can push the same levels of OC/OV as the Clevo/MSi 680M, Widezu's post a couple of weeks back stated that the Dell 680M cannot be OC'd

._.

 

The Samsung DRAM modules on the Clevo GTX680M is capable of delivering 1250MHz (NVIDIA standard 2500MHz) as factory default, but that's with 1.5V. However, at 1.35V, 1250MHz (NVIDIA standard 2500MHz) can still be achieved

 

So

Clevo DDR5= 1.5v
Dell DDR5= 1.35v

?

I'm also concerned about the core OC+OV throttling.

 

Clevo GDDR5 = 1.35V

But can still push it up to 2500MHz.

Not sure about Dell, if you can physically find the DRAM module code printed on the DRAM banks then you can search on Hynix's official site and find out what voltage it's running and what is the maximum frequency supported with the maximum voltage.

 

Aw man /:

I really do hope there's a second batch dell 680M or a bios mod/fix for the current dell 680M's

Considering a 680M end year

 

First test Alienware M17X R4

Spoiler

NVIDIA GeForce GTX 680M video card benchmark result - Intel Core i7-3720QM Processor,Alienware M17xR4 score: P6971 3DMarks

 

^ your GTX 680M is downclocked by 10Mhz base/boost?

 

This is Clevo with i7-3610QM and GTX680M @ 854/2000.
NVIDIA GeForce GTX 680M video card benchmark result - Intel Core i7-3610QM Processor,CLEVO P15xEMx score: P7040 3DMarks

GTX680M @854/2200
NVIDIA GeForce GTX 680M video card benchmark result - Intel Core i7-3610QM Processor,CLEVO P15xEMx score: P7077 3DMarks

Alienware's GTX680M is seriously slower even with high clock...

 

No i did downclock it to 940mhz so it is perfectly stable.

 

aka throttle .

 

3dmark Vantage

Spoiler

 

Rifi, if you don't mind. Can you overclock to near 1000MHz core/2400MHz vRAM and still bench with it? (no crash/artifacts)

 

Spoiler

 

^ Those are some nice scores, sir!

 

Just found out what the Dell 680M Hynix RAM was:
SK Hynix

It doesn't mention anything regarding the bandwidth/frequency

Also just benched with the 306.02 BETA and indeed they offer a great boost benchmarks wise as my GPU score in 3dmarkVantage went up from 26649 to 28513 and Unigine Heaven from 1405 to 1515.

 

Good job guys. Although I found 304.79 to be very problematic with my Clevo 680m causing higher scores in some cases, and lower in others, and even crashes with 3DMark11 in some instances. 302.77, latest Clevo drivers may not be the fastest, but are more consistent and stable. Also, running at higher clocks / voltage can result in a throttle despite temps listed, especially with 304.79.

Curious on the temps though. If you run full fans on the Sager you can drop temps ~ 10C, but it's not an option on all Clevo laptops, only Sager I believe actually. Edit: D'oh! I realized your chart says "turbo fans on". My results are without turbo fan on.

I noticed that the Dell 680m Last Remnant was likely run at full screen and MSI was run at windowed because there shouldn't be such a huge discrepancy. I just ran it at windowed and got 98FPS, but 173FPS full screen.

Compare your results with mine at 1000/1200:

Benchmark: Dell 680m / MSI 680m / Clevo 680m
3DMark 11 P: 6805 (58C) / 7865 (77C) / 7942 (82)
3DMark 11 X: 2190 (55C) / 2495 (78C) / 2695 (82)
3DMark Vantage 1280x1024: 26022 (65C) / 26649 (87) / 24838 (86)
Unigine Heaven 3.0: 1468 - FPS 58.3 (62C) / 1405 - FPS 55.8 (93C) / (need to run your settings)
Stalker Pripyat: 62-63-63-50 (64C) / 77-74-76-54 (93C) / Ultra 89-84-87-59 (87C)
Alien vs Predator: 72.6 (64C) / 78.3 (80C) / 76.0* (76C)
Resident Evil 5: 139 (67) / 146 (75) / 123 (86)
Last Remnant: 164 (58C) / 105 (75C) / 98* (72C - 67C with Turbo fan)
Lost Planet 2: 73 (62) / 75 (90) / 91 (86)

*Run at 900/1200 stock voltage

 

I also get crashes in 3DMark11 with some of the later drivers.

The 302 series seems much more stable.

 

Within that batch you have 3 ratings, R0C, T2C or T0C at either 1.35V or 1.5V operation, so if you have the extra codes on the chip you can read them.

 

That's no fun, I only get P6200 score on my rig with the new BETA 306.02 drivers, was getting 5900 on the Dell driver.

 

That's normal for stock clock, what he showed is OC'd clock. However his OC'd dell GTX680M seems to be under performing at such clock. Clevo cards can achieve 7000+ on GPU score with 854/2000. But his dell GTX680M couldn't even break 7000 on the GPU score with base clock 940/1100.

 

Doh, just realized the overclock. GPU-Z on his pic tricked me by showing default clock as almost identical.

 

Excuse my naivite, but how does one OC the 680m SLI in an m18x? I've found a smattering of posts mentioning a modded vBIOS is necessary, but no consolidated guide.

 

M17x R4's running BIOS A05 supposively removed that bug without needing a modded vBIOS.

Idk about the M18x.

 

+rep work is work and you two did a lot of it.

I can only comment on the 7970m (as it's all I have), but I think your 1000/1500 OC is working against you. In 3DMark11 I was consistently cracking X2000 barrier with 950/1450 clocks whether it was the original Dell driver, 12.6whql or 12.7beta and with my highest stable core of 974 along with bumping to 1550/60 mem was consistently cracking X2100 (best was X2123 on 12.7beta 974/1560). With AMD drivers, P scores are also a hair better than yours with lower core clocks (e.g. 970/74 vs 1000).

I'll run some of the other tests at the exact same settings and see how they compare, but 3DMark11 alone tells a story and I only have a measly 3.1GHz 3610 to boot.

EDIT: This may have something to do with the different rigs though, idk.

 

The Hynix chips are T2C rated 1250Mhz (2500/5000) @ 1.5v & 900Mhz (1800/3600) @ 1.35v

 

So they are rated like the Samsung vram then! Thanks for the clarification Premabooss

 

Still no info on core throttling?