Test of the Real Nvidia Gtx Titan, single-sli-tri Sli, 4 WAY SLI

[Teeth]

Hello Pepillo. Thanks for your help.

The graphic card holds up well up to 1150 OC, but I have it at 1084. Without touching voltage. The memos go well even at 6500, but I haven't touched them.

The ASIC is quite low, 66.9. That's what made me doubt about returning it.

Anyway, from 980 to 1006 it's not that there's an abyss, and, overclocking it, it goes up well to 1084 and stays there stuck.

What do you think about the ASIC?

[Pepillo]

1.084 Mhz without touching voltage, it seems more than correct to me, if it's without Throttling, many would already want it.

Look, I'll explain myself: These cards come limited by consumption and temperature. The maximum you can go up to, without changing the bios, is 106% of consumption, and the temperature to 95º, which doesn't normally affect because in a well-cooled case, neither with overclocking nor overvoltage does it go beyond 80º-82º. Due to a problem recognized by nVidia itself, the drivers read a lower consumption than the real one, but when you get close to 97%-98% you are in reality already at the limit of 106%, the 265W, and they start to lower the clock and the voltage, the famous throttling. This can only be fixed by flashing the bios to allow the 300W.

Therefore, touching the voltage immediately brings negative consequences, consumption skyrockets and throttling appears earlier. It also happens, for example, if you modify the fan profile to keep it cooler (something totally unnecessary, by the way, since the temperature limit is not reached due to its excellent series cooling) that it consumes more.

In short, in my experience, the best thing is to look for the maximum that you can do with the series voltage and without throttling appearing. Look, this is the maximum that I achieve in those conditions:

1.150 Mhz of core and the memories at 6.302 Mhz. The core can't hold much more without raising voltage, which would skyrocket consumption and lower my clock, and the memos, although it can go up more, I'm already at the consumption limit, peaks of 96%-97%, anything I touch already makes throttling appear. That is in my opinion the ideal overclock for a Titan. If you look closely at the graph on the left, you'll see how the GPU speed remained stable at those 1.150 Mhz throughout the benchmark. For example, with the core at 1.163 Mhz and the memos at 6.400 Mhz, it also holds without raising the voltage, but there are punctual drops in speed in the test for exceeding the maximum consumption. Although I can go up more if I touch the voltage, and even achieve some marginally higher score in the overall test, I would do it at the cost of losing that stability of the processor clock, with ups and downs, and that seems to me to be a bit of self-deception about the card's capacity. If you want more, you have to flash the bios to allow those 300W, otherwise, I assure you that I have read the entire thread of overclock.net of the titan, more than 6,000 posts, and most would be more than happy to have a Titan like mine, that without touching bios or voltages, reaches 1.150 Mhz of core with 0 Throttling. Don't believe everything you read about 1.200 Mhz or more, they do it with throttling or with modified bios that allow greater consumption, without knowing the effect it can have in the long term on the card. That said, check that the core clock is stable in your overclock.

As for the ASIC, in this series many claim that it has a lot to do with those differences in boost from one card to another. Go figure, in any case, I don't think there's a big difference between your card and mine.

Regards

[Teeth]

Thanks for your help.

Right now I'm testing the graphics with Crysis 3. At 1150 it goes to the desktop. It's more comfortable at 1110 or 1084. If I leave this frequency it usually stays there. It may drop to 1071.

I'm going to run the unigine and I'll tell you more.

The difference is that if I put 150 offset like you, due to the difference in the initial boost, it stays at 1110. It starts at 1124 to 1.16, but it goes as soon as it approaches 80 degrees (regardless of whether I mark the temperature at 94 degrees, which is curious).

Also, increasing by 13 mv to reach 1150 also fails.

What tool do you recommend for testing the correct operation of the overclock? I used to do it with OCCT, but it's no longer good for these series.

Thanks, really.

[Pepillo]

Run Unigine as I do, in extreme and in windowed mode, with the graphics open to verify that it doesn't throttle you. And I repeat, it's not the temperature (although higher temperature also means higher consumption), it's the consumption that causes the clock to drop, more than proven. Unigine is very demanding in that regard, you'll see that it's not easy to avoid throttling. Generally this appears before the graphics reach their maximum and you crash, especially if you test little by little. Look for what I have done, your stable maximum and tell me things. Then, a good session of Crysis 3 is a good confirmation of the stability of that overclock. Your lower boost when putting the same offset is probably that lower ASIC of your card. Saludos

[Teeth]

Thanks Pepillo, really, for going to so much trouble to help me out.

Right now I've passed a test with the unigine. It stays stuck at 1084, 1.15 and 250 of memory (6500).

It's one frame below your maximum, and two above your minimum, which you posted.

What I observe is that the clock oscillates when the fan gets close to 65-70%. There the clock goes down. If it stays below, it's fine.

I'm going for the next step, I'll let you know, and I'm going to bed, because tomorrow I have 10 hours of work...

[Pepillo]

It's what I told you, if the clock goes down, and the fan reaches 65%-70%, you're at a higher temperature than my tests (note that my fan didn't go over 61% on automatic), and exceeding the maximum consumption. Although you get a higher score, it's not ideal for a 100% stable overclock for gaming. Lower it a bit until the clock stays stuck at a frequency all the time (also play with the memory, everything influences consumption), and that will be the ideal maximum to leave your Titan in stock (another thing is for benchmarking). I also wake up early and have to go to sleep, we'll "talk" tomorrow.

Regards

[Teeth]

Ok. We will continue the discussion tomorrow.

I just passed the bench, 1110, 6500. Stable without crashes. The only thing that fluctuates is at the beginning, when it goes from 1.16 volts to 1.15, from 1124 to 1110. At that frequency it gets stuck.

It seems that I am beginning to see how the graph works.

Pepillo, we will continue with the work tomorrow.

Thanks again for your help. I will never get tired of telling you. Rest!!!

[bankuai]

¡Esta publicación está eliminada!

[fjavi]

@Teeth:

Thanks Pepillo, really, for taking so much trouble to help me out.

Right now I've passed a test with the unigine. It stays stuck at 1084, 1.15 and 250 of memories (6500).

There is one frame below your maximum, and two above your minimum, which you posted.

What I observe is that the clock oscillates when the fan gets close to 65-70%. There the clock goes down. If it stays below, it's fine.

I'm going for the next step, I'll let you know, and I'm going to bed, because tomorrow I have 10 hours of work…

Of course if the fan revs up more it consumes more and that's what limits these cards, the consumption is not as they said according to the temperature it's because of consumption that's why if you block the fan for example at 50% it's easy to hold the boost better, if you put it at 100% it will last less.

What you're interested in is a stable frequency for gaming, in the end with the Titans for gaming there's not much difference that it goes up a little more, in benchmark it would be noticed a little more but in games it should be very little.

What matters is a stable boost and that the graphics are at a good temperature and don't make much noise.

[Teeth]

@fjavi:

Sure, if the fan spins faster it consumes more and that's what limits these cards, the consumption isn't as they said, according to temperature, it's because of consumption, that's why if you block the fan for example at 50% it's easy for it to hold the boost better, if you put it at 100% it will hold less.

What you're interested in is a stable frequency for gaming, in the end with the Titans for gaming there isn't much difference that it goes up a little more, in benchmarks it would be noticed a little more but in games it should be little.

What matters is a stable boost and that the graphics are at a good temperature and don't make much noise.

Hi fjavi, well yes, when I get home I'll do some more tests. I don't mind that it doesn't go up to 1150 or anything like that, what bothers me (out of habit, more than anything else) is that, from what I see in forums or in reviews (I pay less attention to these), generally, without touching anything, they go up to 1006, 1058…, although then they fall to 980 or 1006 respectively.

The one I have, as I observed in the photos of the ELP3, goes up to 980. If it gets hot, it automatically goes down to 967 and 1.15.

Basically that's it, it's a matter of habit. It's true that, I've only had a day to tinker with it and that work doesn't allow me to spend more time with it, and that, I hope, that, at least, it stays at 1084 continuously. But, of course, this overclocking is a lottery and, generally, I don't usually have much luck with the components...

It is true what Pepillo says, I've taken a look around the EVGA forums and some seem to have been sent selected graphics cards, because you can see 1250 and other things, but, of course, going up to 1150 at full voltage and passing a Unigine doesn't seem to me to be a good methodology for seeking or fine-tuning an overclock...

Another thing I don't understand, and this question is for Pepillo, is that in the image you posted your graphics card goes to 1.15 from the beginning of the Unigine. However, mine starts at 1.16 and then goes down to 1.15. Is that normal?

And another thing I haven't mentioned is that the installation of the Titan was done without formatting from the SLI of the 480. It's true that I went through Driversweeper and other removal tools. Should I format? (Perezón).

Thanks!!!

[Pepillo]

The voltage thing doesn't work for me, as you can see in the graph it goes up to 1.15 and stays there, although I wouldn't give it much importance. If you've been to overclock.net, precisely in one of the last posts they mentioned that the usual ASIC for these cards is between 65 and 75, very few are 80.

As for the drivers, I had a clean install from a few days ago of Seven with the GTX 580 SLI, and I didn't format, nor driveswepeer, nor anything, just a clean install from the nVidia panel, I've always done well that way. I have it as I showed you, 1,150 Mhz of Core and 6.302 Mhz in memory without touching voltages, and it passes all the usual benchmark batteries 3dMark, Unigine, etc., but, above all, I've played the whole Dishonored and many hours of Crysis 3 and Dead Space 3, without the slightest problem. For me that's worth more than any stress program, I've been overclocking components that way for a long time and not running a few minutes of synthetic benchmarks. I've seen (and suffered) more than once that what seems stable in those benchmarks then isn't when playing. I use synthetic programs to get a quick idea of where I can go, then I lower it a bit, and I play for hours, that's what counts.

Regards

[Jotole]

Well, once again, and since I can't stay still, I have to swallow what I said about not flashing them….....:wall: Yesterday all the parts for the new RL arrived, including the graphics blocks. So before removing the reference cooler and putting in the blocks, I pulled the wool over my own eyes and flashed the Titans.........................:ugly: I put one of the most aggressive ones on them, since we were at it...........;D Just have to be careful when putting OC on them, in case they end up as a very pretty doorstop..............xD (and expensive).... I flashed them one by one, following the OCN guide, and a little help. and the truth is it's impressive. To see these cards at 1200 Mhz without any throttling and getting 9 fps in the Unigine test I have, with the graphics at 1200. I didn't push it any further, and it was the first OC I tried by eye, raising the vcore by 10, and the Power Limit to 120%. and 200 on the Clock. I don't have screenshots because right after I removed the graphics card, I put another one in to do the same process. Flash it, check that it boots up (very important...xD) and run a full Unigine to check that there were no problems. and I started to disassemble. But as soon as it's operational, you'll see how they pull without throttling, it's crazy... Regards....

[Pepillo]

Heh, I would have bet money that you would end up doing it, and when I saw that you had signed up for OCN, I thought that in less than 24h you would do it

I was about to the other day, but I stopped myself because I read that in idle, on the desktop, the voltage does not go down. Do you mind checking if that's the case? Which bios have you used? There are dozens of them....

I feel like trying it out, mine doesn't need a 200 offset, with 175 it already goes over 1.200 Mhz, I have the feeling that I got a good one and I feel a bit bad not squeezing it.

Regards

[Teeth]

;)Well, I've been running Unigine for an hour and it seems (and I say seems, touch wood) that it's stable at 1110. Not bad.

Quiet, cool and with pretty good performance (like my SLI, but without a quarter of the noise).

I don't know Pepillo, do I return it? ;D, what luck you have, damn...1200...surely without breaking a sweat...

[Pepillo]

No, the 1.200 are increasing voltage and with Throttling, I'll see if I get the courage to try a modified bios, like Jotele. Without touching voltage and without Throttling they are 1.150, 40 Mhz of difference does not seem to me to return a card.

Greetings

[Teeth]

It was a joke, man! I'm not going to return it for that. In fact, I was happy with the 1084.

I tried to set the voltage limit and see how high it goes. It stays at 1137-1150. It's not worth it for the gain. Moreover, as you mentioned, it's even worse, because when the TDP is exceeded, the frequency drops to 1097-1110, so in the end, you have the same thing but with more voltage.

It's a shame that NVIDIA doesn't allow a little more of the TDP, but of course, it would go beyond the standard for these cards.

Modifying the BIOS gives me a little scare... it's very risky, damn...

[fjavi]

@Teeth:

Hello fjavi, yes, when I get home I'll do some more tests. I don't mind that it doesn't go up to 1150 or anything like that, what bothers me (more out of habit than anything else) is that, from what I see in forums or reviews (I pay less attention to these), generally, without touching anything, they go up to 1006, 1058…, although then they drop to 980 or 1006 respectively.

The one I have, as I noticed in the photos of the ELP3, goes up to 980. If it heats up, it automatically drops to 967 and 1.15.

Basically that's it, it's just a habit. It's true that I've only been tinkering with it for a day and that work doesn't allow me to spend more time with it, and that, I hope, that, at least, it stays at 1084 continuously. But, of course, this overclocking thing is a lottery and, generally, I don't usually have much luck with the GPUs...

What Pepillo said is true, I've had a look around the EVGA forums and some seem to have been sent selected graphics cards, because you can see 1250 and other things, but, of course, going up to 1150 at full voltage and passing a Unigine doesn't seem like a good methodology for finding or fine-tuning an overclock...

Another thing I don't understand, and this question is for Pepillo, is that in the image you posted your graphics card goes to 1.15 from the start of the Unigine. However, mine starts at 1.16 and then drops to 1.15. Is that normal?

And another thing I haven't mentioned is that the Titan installation was done without formatting from the SLI of the 480. It's true that I ran Driversweeper and other removal tools. Should I format? (Perezón).

Thanks!!!

The thing with the 1006 doesn't happen with all of them, I think the ELP3 has two that are at 1006 but another one goes to 996 and the worst one to 980, so it depends on how good the chip is, even the memory, because worse memory will consume more and limit sooner, anyway, it's not bad that a single GPU already gives what an SLI of 580 does, plus with more than enough VRAM for the games that are coming out very demanding of memory.

Also, a single GPU works better than an SLI, less stuttering problems and without worrying about SLI profiles, it's about playing everything without changing drivers, I would change my two 480s for one of these if they were cheaper, because as a card I think they're pretty good, let's see if they go down in price and I'll go for one, at least.

regards

[Pepillo]

@Teeth:

Modifying the bios gives me a little scare…jotole is that it's very risky...

Well, I've also jumped into the pool:

Honestly, I think it's worth it. It's not just the 1.202 Mhz that it reaches without touching anything else but the power limit, in my case I've tried with 115%, and it's perfect on the first try. It's also that you can set a profile that's more suitable for the fan, and the temperature has stayed at 74° during the benchmark. That said, this bios sets the boost to 1.202 Mhz automatically, if you get one that doesn't reach it, you have to set the offset to a negative value, -10, -20, or whatever you need to make it stable. But since it raises the voltage to 1.20, it easily reaches the 1.202 Mhz, cooler, and without Throttling. And my doubt, I had read that in some cases the voltage didn't lower in idle, but it has worked for me without problems, it lowers the voltage to 0.862 at rest.

Tomorrow I'll run a session of Crysis 3 to be more at ease, and I'll test to see how far this Titan goes :mad:

By the way, the process of flashing the bios, extremely simple, I had never done it with a graphics card, but it couldn't be easier. Here's where they explain it best:

http://www.evga.com/forums/tm.aspx?m=1891166

Regards
P.D. Teeth, if you're not sure, don't do it, the truth is that it gives you a "yuyu" to do that to a card of this caliber that you can't see …...... but what would life be without the thrill of risk

[majo]

I don't have much idea about OC on GPU, but I've been intrigued by this one I have and without touching the voltage it gave me this last Sunday-Monday. Pepillo, tell me what you think. This weekend I'll build the second one and the RL blocks, when I get back from work. According to my coworker, the second one I'm going to build is somewhat better than this one, since he does stress tests on them, I don't, I do other types of tests, but I haven't really looked to compare, that's why I ask you Pepillo to tell me how you see it. This is what I got last Sunday from the one I have at home but without touching much, certainly the stock voltage, I mean I only touched the powertarget, I didn't raise any voltage:

[Teeth]

¬¬

Well, another one that goes up like foam... :wall:

I've been playing Crysis 3. Although in Unigine the graphics stay stuck at 1084, in the game it fluctuates between 1084, 1058, 1045 and usually doesn't go below that. So this whole overclocking thing isn't that simple if you don't use a modified BIOS.

Not Pepillo, my inhibitions are getting the better of me...