What does "troll" have to do with anything if those are the Hynix design specs for the Titan X memory and the GDDR temps THG recorded are accurate? If you have proof that contradicts either one, by all means, please post it. Because it doesn't really make sense to me, yet that's the way it appears.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
NVIDIA Unveils The Titan X At GDC 2015
- Thread starter HardOCP News
- Start date
What does "troll" have to do with anything if those are the Hynix design specs for the Titan X memory and the GDDR temps THG recorded are accurate? If you have proof that contradicts either one, by all means, please post it. Because it doesn't really make sense to me, yet that's the way it appears.
5150Joker
Supreme [H]ardness
- Joined
- Aug 1, 2005
- Messages
- 4,568
Do you have any proof the memory will fail? No you don't, nor do you have the qualifications to know (neither do I), all you are doing is spreading anti-NV FUD as usual. However, the engineers at NVIDIA should be smart enough to know that and anticipated MTBF for these. Otherwise there wouldn't be 3 years warranty on the cards.
Last edited:
Again, do you dispute either the THG's thermal image or the Hynix datasheet?
5150Joker
Supreme [H]ardness
- Joined
- Aug 1, 2005
- Messages
- 4,568
See above post, I don't pretend to know what the upper tolerances are for these memory chips. The spec sheet could be a very conservative number and these are still well within tolerance. You just don't know and unless you have spoken with Hynix and have insider knowledge, you're just making things up. You know..being a troll and spreading FUD as usual. I would assume a company like NVIDIA has validated every component and stress tested the design before release and that's why they are backing it up with 3 years warranty.
I'm not disputing the fact that it's possible the datasheet doesn't reflect the limits of the memory on the Titan X. But you can't just dismiss the possibility that it's true, either. Not without some proof. You have to admit that the thermal image and the datasheet do open up the door to the possibility that the RAM is being run out of spec.
Dayaks
[H]F Junkie
- Joined
- Feb 22, 2012
- Messages
- 9,773
If I wasn't putting my card in a cooled chamber I would be slightly concerned. 
If you're really worried buy some heatsinks for $10 and nonconductive epoxy for $8. Bam problem solved.
If you're really worried buy some heatsinks for $10 and nonconductive epoxy for $8. Bam problem solved.
Last edited:
5150Joker
Supreme [H]ardness
- Joined
- Aug 1, 2005
- Messages
- 4,568
AMD runs their GPUs at 95C, were you up in arms about that? Did you go look up data sheets to make sure everything was in tolerance? Doubt you did. If AMD can run their hardware at 95C then NVIDIA can run their memory chips at 95C as well. They wouldn't risk losing millions of dollars on warranty claims by running something outside of specifications that would be prone to failure early on. So again, this is just you spreading FUD. I'm done talking with you.
Last edited:
misterbobby
2[H]4U
- Joined
- Mar 18, 2014
- Messages
- 3,814
Yet they were willing to risk lying about the specs of the 970 and playing dumb for months.
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,688
Exactly. We have no idea what kind of slick tactics Nvidia is capable of these days.Yet they were willing to risk lying about the specs of the 970 and playing dumb for months.
It's hard to trust any new products they release. And frankly, I have trouble convincing myself to buy products from a company that has intentionally done things to violate my trust.
Dayaks
[H]F Junkie
- Joined
- Feb 22, 2012
- Messages
- 9,773
Yeah, I don't trust em. If I were you guys I'd never buy their shoddy products again.
So this was a stock, open air test? What happens when it's in a mildly warm case!?!?? And OC'd!?!!?!!
Last edited:
So GPU's and memory typically run at the same temperature now?
If you'd like to disprove either the thermal image or the datasheet then do so. Otherwise, just spouting off doesn't prove anything. If you will recall, Nvidia also recently said the 970 came with 64 ROPs and 2 MB of L2 cache. We all know how that changed four months later.
Attack the post, not the poster.
Sheesh.
Yet they were willing to risk lying about the specs of the 970 and playing dumb for months.
Flawless Victory! LOL
5150Joker
Supreme [H]ardness
- Joined
- Aug 1, 2005
- Messages
- 4,568
Well obviously it catches on fire and burns down your home! Damn, NVIDIA is out to make people homeless! I knew they were evil!
Yet they were willing to risk lying about the specs of the 970 and playing dumb for months.
you're talking about the marketing department here, and i agree.. not cool. that being said, the engineers at nvidia are not that stupid. they get paid good money to make good products. they know the capabilities and limitations of the components that go on their boards.
you guys are making accusations based on a tomshardware image and a random pdf you found on the internet. do you know how many man hours went into design and testing of these cards? hah
Many others were, and there were a lot of brouhaha about the 95C thing on pretty much every forum.
It was simply pointed out the memory ICs are being run out of Hynix's specs. You're right we don't know what the upper tolerance is, so one could only speculate. However, I still see nothing wrong with pointing out the facts.
In regard to the "If AMD can run their hardware at 95C then NVIDIA can run their memory chips at 95C as well" comment, very true. But I would also add "If AMD can run their hardware at 95C and get blasted for it on every forum, then NVIDIA can run their memory chips at 95C as well and receive the same amount of lambasting.
To be absolutely fair, Guru3D's recorded GDDR5 temps were a bit lower, but still in excess of 90C. While not as bad as Tom's, we at least have a second result showing the GDDR5 chips on the Titan X run hot.
Last edited:
5150Joker
Supreme [H]ardness
- Joined
- Aug 1, 2005
- Messages
- 4,568
Yet those memory chips running at 95C is a non-issue because they've been validated by NVIDIA to run that hot and as I said, nobody knows the real tolerance of these because the spec sheet could be very conservative w/their figures. NVIDIA isn't stupid, they won't produce a card that will have memory failure a few months or even years after ownership and risk millions in warranty claims. This is even more true now than ever considering the headaches they had with bumpgate etc. So I'm fairly certain the engineers that designed the Titan X know exactly what the tolerances are and stress tested it in the lab to make sure it meets their standards.
Therefore pointing it out is fine but trying to play it up as some sort of issue when there is none puts into question certain individuals ulterior motives. Now to discuss something more interesting, people that are running the Titan X on a modded vbios are apparently hitting 1500+ MHz stable which is pretty amazing considering this thing has 8.1 billion transistors (more than Titan) and still manages to stay under 90C on a 28nm node. I don't know what kind of voodoo these NVIDIA engineers are doing to accomplish that but I won't complain I can't wait until Amazon finally gets around to shipping mine.
Therefore pointing it out is fine but trying to play it up as some sort of issue when there is none puts into question certain individuals ulterior motives. Now to discuss something more interesting, people that are running the Titan X on a modded vbios are apparently hitting 1500+ MHz stable which is pretty amazing considering this thing has 8.1 billion transistors (more than Titan) and still manages to stay under 90C on a 28nm node. I don't know what kind of voodoo these NVIDIA engineers are doing to accomplish that but I won't complain
I can't wait until Amazon finally gets around to shipping mine.
Last edited:
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,688
Take this post, change "Nvidia" to "AMD" and "Titan X" to "290X" and it sounds like 2013 all over again, but this time from the opposite side of aisle.
I wonder how many people who shit on the 290X for high temperatures are now flipping and defending Nvidia on the exact same issue. It has truly been an enlightening week thus far.
Maybe we're not so different after all...
5150Joker
Supreme [H]ardness
- Joined
- Aug 1, 2005
- Messages
- 4,568
People were crapping on AMD for the cooler from what I recall. In this case those of us buying Titan X's are still not too thrilled w/the cooler so it's hardly the same thing. Personally I'd have appreciated a more robust stock cooler and higher power than 250W. But they obviously had certain limitations in mind when they designed the card. Fortunately, it can exceed those pretty easily.
Yet those memory chips running at 95C is a non-issue because they've been validated by NVIDIA to run that hot and as I said, nobody knows the real tolerance of these because the spec sheet could be very conservative w/their figures. NVIDIA isn't stupid, they won't produce a card that will have memory failure a few months or even years after ownership and risk millions in warranty claims. This is even more true now than ever considering the headaches they had with bumpgate etc. So I'm fairly certain the engineers that designed the Titan X know exactly what the tolerances are and stress tested it in the lab to make sure it meets their standards.
Therefore pointing it out is fine but trying to play it up as some sort of issue when there is none puts into question certain individuals ulterior motives. Now to discuss something more interesting, people that are running the Titan X on a modded vbios are apparently hitting 1500+ MHz stable which is pretty amazing considering this thing has 8.1 billion transistors (more than Titan) and still manages to stay under 90C on a 28nm node. I don't know what kind of voodoo these NVIDIA engineers are doing to accomplish that but I won't complain
Fair enough but, I would counter that many people on many different forums played up the 290X hitting 95C issue, which is pretty analogous to what's happening here, but there were few if any defenders (and most of them were fanboying defenses so did more harm than good).
I would even argue the 290X case is more egregious because people didn't have access to AMD's engineering specs, but jumped to conclusions anyway. At least here there's a tenable "concern". Just seems a bit like double standards to me.
Anyway I'm just saying it like I see it, I have no horse in this race either way.
Memory shown running hot in two thermal images: cause for concern and worth bringing up in an enthusiast forum to discuss.
I don't know why people insist on assuming people are hating or insist upon defending NVIDIA right away. Let's have a constructive discussion about it and see what we can find out.
5150Joker - you said NVIDIA validated the memory chips to run that hot? Do you have a link or proof?
I've personally had insane OC results with my TITAN X SLI and the memory OCs to the moon, as well. So I'm curious because I'm wondering if we should be looking for an aftermarket solution. I personally will not be putting my cards under water.
I don't know why people insist on assuming people are hating or insist upon defending NVIDIA right away. Let's have a constructive discussion about it and see what we can find out.
5150Joker - you said NVIDIA validated the memory chips to run that hot? Do you have a link or proof?
I've personally had insane OC results with my TITAN X SLI and the memory OCs to the moon, as well. So I'm curious because I'm wondering if we should be looking for an aftermarket solution. I personally will not be putting my cards under water.
It's not hard to stick some tiny heatsinks made for memory chips on the backside of the card unless there's no room for some reason.
Personally if they were my cards they would be under water. It's not all about cooling but quiet running cards.
In that case - you're good. These cards are not loud at all. This is not a 290/X situation. I had those cards at launch. They weren't even that bad - unless you had them kicked on for benching or a huge OC.
AFD
2[H]4U
- Joined
- Aug 7, 2013
- Messages
- 2,154
Both listings at EVGA also state "24 GB system memory (48 GB or higher recommended)".
The EVGA tech had this to say about it..
Stuck the Titan on my POS Gigabyte Z68X-UD3H-B3 that received such shit support they never bothered to finish the EFI bios upgrade they promised, just released a beta and never released a final. The computer would power on with the titan and show the splash screen, but would take 5 minutes to get past that to a blank screen then apparently froze. That system only had 16GB RAM in it, but I'm pretty sure the fault lies squarely with that pile of shit Gigabyte board. Upgraded to a X99 with 64GB RAM, works just fine.
As mentioned before, nVidia could've done a better job with retooling the blower instead of being lazy and slapping on the one from vanilla Titan.
But you also have to consider GM200 is pretty much 28nm being pushed to its limits. 390X appears to be 28nm's last hurrah so I'm kind of curious to see what AMD will be able to knock out. And also maybe why they're going with a CLC for reference cooling.
But you also have to consider GM200 is pretty much 28nm being pushed to its limits. 390X appears to be 28nm's last hurrah so I'm kind of curious to see what AMD will be able to knock out. And also maybe why they're going with a CLC for reference cooling.
TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,688
The cooler used on the 980 was a gimped Titan cooler. No vapor chamber, iirc. Yeah, it looked identical to the original Titan. Most people think it is... but it's not. The same way people thought their 970 actually had 4 GB.
Are we certain the Titan X uses the original Titan cooler and not the cheap 980 one? Food for thought.
Are we certain the Titan X uses the original Titan cooler and not the cheap 980 one? Food for thought.
Looks like you were wrong after all
The EVGA tech had this to say about it..
Quote:
I am very sorry for any confusion as the memory requirements listed in the manual are incorrect. Please however keep in mind that the Titan X is able to run with less than 24GB of system memory, but that is the minimum recommended by Nvidia at this time.
..
If you do not use over the internal 12GB of VRam then you should not necessarily see a problem. However if you do use over 12GB of VRam, which will not be easy at this time, then your system will be looking for 12GB of internal memory for a swap file.
http://forums.evga.com/System-Requir...-m2310062.aspx
Yeah it seems to run 2-3C warmer than og Titan by default
5150Joker
Supreme [H]ardness
- Joined
- Aug 1, 2005
- Messages
- 4,568
The proof is in the fact that you own two Titan X's right now which are doing fine w/out failing despite you "OC'ing to the moon". These things are very highly likely validated in house before production. I don't think they just stick them on a PCB and take a gamble, do you? Not everything needs a "link", sometimes common sense can prevail. That said, if you are OC'iing, then it doesn't hurt to take extra measures to cool them down if you feel you want to. Personally I'll be putting mine underwater after Amazon gets around to delivering it and I build my loop.
Last edited:
Factum
2[H]4U
- Joined
- Dec 24, 2014
- Messages
- 2,455
They should of had it shipping with an AIO LN2 cooler, and clocked at 2ghz+.
Whach
[H]ard|Gawd
- Joined
- Dec 22, 2011
- Messages
- 1,244
So. Worth going from SLI'd 980's to a single Titan X? Price wise, not much in it.......
Do not be so sure in ironizing guys, there are serious reasons why the 24GB card should appear:
1. NVidia has K80 professional card with 24 GB VRAM so this amount is not new
2. K80 is dual GPU card based on max Kepler chip
3. Historically there was dual GPU card with max chip of any generation
4. Titan Z is the right example, it has 12 GB VRAM which is double of the singlet Titan
5. It is thus logical that the dual GPU Titan XXX based on max Maxwell chip and with 24 GB VRAM is in the making
There are only two problems:
1. The dual GPU/dual VRAM card appears quite some time after the launch of the singlet card
2. NVidia will charge outrageous premium for the dual card like it was with the Titan Z
Anyway, be certain that Titan XXX will be available before the April 1st 2016 which will make fool of you on this day.
1. NVidia has K80 professional card with 24 GB VRAM so this amount is not new
2. K80 is dual GPU card based on max Kepler chip
3. Historically there was dual GPU card with max chip of any generation
4. Titan Z is the right example, it has 12 GB VRAM which is double of the singlet Titan
5. It is thus logical that the dual GPU Titan XXX based on max Maxwell chip and with 24 GB VRAM is in the making
There are only two problems:
1. The dual GPU/dual VRAM card appears quite some time after the launch of the singlet card
2. NVidia will charge outrageous premium for the dual card like it was with the Titan Z
Anyway, be certain that Titan XXX will be available before the April 1st 2016 which will make fool of you on this day
.
Whach
[H]ard|Gawd
- Joined
- Dec 22, 2011
- Messages
- 1,244
And with Half-life 3 !!