Kingpin breaks 3DMark world record by four Titan X Pascal's

[EarthDog]

Oh, I accept your opinion... no doubt. With respect, I accept it as the opinion shaped by someone that doesn't know much about what was done there to reach those scores... and that is OK! TPU isn't exactly known for hardcore enthusiast level overclocking like other sites. We are here to share and learn from each other!

That said, he did not reach those clocks without modifying the BIOS at minimum. I'd also venture a guess that is actually on LN2 on the cards. You also cannot clock 2 cards as high as one, 3 cards as high as 2 or 1, 4 cards as high as 3, 2, or 1. Those are the highest clocked set of 4 cards in this list and top 5 against a SINGLE card clockspeed wise. As I said in an earlier post, more was done than more people think.

The point here is never about gaming, never about having some performance gain linked to applications except the ones at Hwbot where people like us (though I am NOT in his league, I have used LN2 many times over the past few years) go to 'compete'. This is no different than wondering why someone's track car isn't a dialy driver. Its OK for you to have the opinion its not relevant... but if that is your opinion, go poo poo in a thread that isn't about getting these high clocks.

I digress...

 

[Wimpzilla]

Well, for what i know, there is no way to edit the pascal gtx bios, maybe i'm wrong but on the net there is no tool that showed up accessible to large amount of people, if it exist.
The only thing u can do it is try the different bios that come out of the box between the different brand, some will allow u more OC than the others, but that's all, if i remember well the VGPU is linked to the TDP and there is no way to really unlock it without touching the TDP.
So i rather suppose that EVGA gifted him some really nice samples, with really good asic, that allow u to get max clock stable with the default VGPU allowed by the max TDP with simple air/water cooling.
After that i think u did not grasp the technology behind these new chip, they really do not need LN2 to reach these clock, even a good aircooling heatsink will allow u to get this frequency, under LN2 these chip break every dream u could have, reaching frequencies above 2.4Ghz, because u lower the TPD and trick the card to not boost down the GPU frequency.
So for me the only thing i can recognize to him it is the tweak/OC of the CPU/MOBO and OS to get this great score.
I really doubt he get them under LN2, even if i'm not an expert there, i would be really tricky to fit 4 LN2 heatsink and destroy more than 5k$ caused by the isolation treatment.
He just take the card out of the box, slide the shit out, profit!

 

[cdawall]

Guarantee he didn't just toss 4 cards in and click the next key regardless to binning.

 

[Wimpzilla]

No for sure i do not disagree, but neither, IMO, did a big tweak/modification to the cards, these one are pretty terrific out of the box.
But if he is sponsored by EVGA, i could assume they helped him a bit, with good asics cards.

 

[Onglar]

As far as i can tell his titans arent clocked unusually high (might be wrong, not very familiar with titan xps), maybe he was given a couple of gold samples, i dont know. But fact is that he doesnt really have a lot of competition, there's only a few people on the HOF with quad titan XPs. And considering he is one of the best overclockers there are im sure he has a couple of tricks up his sleeves to give himself an edge in terms of total score. He also happens to have the highest clocked 6950x (5.2GHz, which needs good understanding of the chip itself and overclocking in general.) if i am not mistaken, thus has the highest score. Go figure.

 

[EarthDog]

Well, for what i know, there is no way to edit the pascal gtx bios, maybe i'm wrong but on the net there is no tool that showed up accessible to large amount of people, if it exist.
The only thing u can do it is try the different bios that come out of the box between the different brand, some will allow u more OC than the others, but that's all, if i remember well the VGPU is linked to the TDP and there is no way to really unlock it without touching the TDP.
So i rather suppose that EVGA gifted him some really nice samples, with really good asic, that allow u to get max clock stable with the default VGPU allowed by the max TDP with simple air/water cooling.
After that i think u did not grasp the technology behind these new chip, they really do not need LN2 to reach these clock, even a good aircooling heatsink will allow u to get this frequency, under LN2 these chip break every dream u could have, reaching frequencies above 2.4Ghz, because u lower the TPD and trick the card to not boost down the GPU frequency.
So for me the only thing i can recognize to him it is the tweak/OC of the CPU/MOBO and OS to get this great score.
I really doubt he get them under LN2, even if i'm not an expert there, i would be really tricky to fit 4 LN2 heatsink and destroy more than 5k$ caused by the isolation treatment.
He just take the card out of the box, slide the shit out, profit!

No for sure i do not disagree, but neither, IMO, did a big tweak/modification to the cards, these one are pretty terrific out of the box.
But if he is sponsored by EVGA, i could assume they helped him a bit, with good asics cards.

LOL, they are terrific, but these are not out of the box cards man... how many times do I need to say it? These cards don't hit 2200MHz+ (one card) on water bud.... none the less FOUR cards at 2200MHz.

As far as being tricky to cool four of these and a CPU under LN2... look at my avatar, that was his rig. Granted it was for marketing purposes, but its been done before, quad SLI under LN2 while a CPU under LN2.

Look at the clocks for all cooling....
http://hwbot.org/benchmark/3dmark_-...#start=0#interval=20#coolingType=1#undefined=

water.....
http://hwbot.org/benchmark/3dmark_-...740&cores=1#start=0#interval=20#coolingType=3

Notice how 2200Mhz+ isn't found until LN2? Now tell me these werent on sub ambient cooling of some sort...it may have been chilled water... but I would suspect LN2.

What you don't seem to grasp is that 2200Mhz+ is subambient territory according to Hwbot. In no way did he take these out of the box, 'slide the shit out' and profit... none.

As far as i can tell his titans arent clocked unusually high (might be wrong, not very familiar with titan xps), maybe he was given a couple of gold samples, i dont know. But fact is that he doesnt really have a lot of competition, there's only a few people on the HOF with quad titan XPs. And considering he is one of the best overclockers there are im sure he has a couple of tricks up his sleeves to give himself an edge in terms of total score. He also happens to have the highest clocked 6950x (5.2GHz, which needs good understanding of the chip itself and overclocking in general.) if i am not mistaken, thus has the highest score. Go figure.

Its what I said many posts earlier, but........... well, you see, LOL!

 

[Hood]

The hobbyt of benchmarking is useful to him and those that participate. Think of it as a nerdy way to modify a car and race it. That's all its for

Good analogy - and like the car, it's pretty much useless for everything but racing, undriveable for trips to the grocery store, and uses way too much gas! But props to the man, somebody has to be the fastest.
I guess if I had unlimited funds, and was bored enough, I would spend $50,000 or so and try to beat these scores, but I'm not that bored (or rich).

 

[cdawall]

LOL, they are terrific, but these are not out of the box cards man... how many times do I need to say it? These cards don't hit 2200MHz+ (one card) on water bud.... none the less FOUR cards at 2200MHz.

As far as being tricky to cool four of these and a CPU under LN2... look at my avatar, that was his rig. Granted it was for marketing purposes, but its been done before, quad SLI under LN2 while a CPU under LN2.

Look at the clocks for all cooling....
http://hwbot.org/benchmark/3dmark_-...#start=0#interval=20#coolingType=1#undefined=

water.....
http://hwbot.org/benchmark/3dmark_-...740&cores=1#start=0#interval=20#coolingType=3

Notice how 2200Mhz+ isn't found until LN2? Now tell me these werent on sub ambient cooling of some sort...it may have been chilled water... but I would suspect LN2.

What you don't seem to grasp is that 2200Mhz+ is subambient territory according to Hwbot. In no way did he take these out of the box, 'slide the shit out' and profit... none.

It was all air cooling how many times do I have to tell you?

 

[Recon-UK]

Waiting for KingPin to put them under LN2, i thought he would have done that straight away so i understand the under 2ghz clocks.

 

[cdawall]

Waiting for KingPin to put them under LN2, i thought he would have done that straight away so i understand the under 2ghz clocks.

This was probably a baseline run for proof of concept that he didn't expect to take the record with.

 

[EarthDog]

Perhaps when he typed the clocks in at Hwbot, he did it wrong? it's there where he says 2200MHz...

Also, how would being so far under the clocks of next best person enable him to beat him by 2k points? you notice second place person is 100mhz less on the core?

Are the cpus THAT different?

Edit: nope.. 200mhz on same cpu...

 

[cdawall]

Perhaps when he typed the clocks in at Hwbot, he did it wrong? it's there where he says 2200MHz...

Also, how would being so far under the clocks of next best person enable him to beat him by 2k points? you notice second place person is 100mhz less on the core?

Are the cpus THAT different?

It is actually really easy to figure that part out. Physx score isn't off by a crazy amount. So I would say that probably does account for 500-750pts in the score, the rest is pure GPU and like you have said boost clocks are into the 2200mhz range. Note that second place run was done with the cards on water, I didn't bother to read if it was chilled water. Kingpin does score some 4000pts higher in the GPU tests that points out to me this is a purely better GPU situation. They are using different motherboards which could have different efficiency without a doubt, but I don't have the EVGA board to do any personal tests with to confirm that.

Here is the second place run

Here is the new record.

 

[EarthDog]

I never even looked at the regular FS, just extreme is what I am talking about.

I dont understand what you are talking baseline run? Nothing was under 2 GHz... I understand GPUz is showing that, but the boost clocks (and perhaps boost tables on the bios?) go higher as you said. All I know is he says its 2200 MHz and that makes sense with his scores and CPU used compared to the others in FS Extreme.

Its also not an air cooled item. At best, its chilled water...

 

[cdawall]

I never even looked at the regular FS, just extreme is what I am talking about.

I dont understand what you are talking baseline run? Nothing was under 2 GHz... I understand GPUz is showing that, but the boost clocks (and perhaps boost tables on the bios?) go higher as you said. All I know is he says its 2200 MHz and that makes sense with his scores and CPU used compared to the others in FS Extreme.

Its also not an air cooled item. At best, its chilled water...

All air baseline run he got a higher score because it says kingpin on it

 

[EarthDog]

Sorry about how dense I am right now... what are you talking about? What baseline run? I don't see an air based basline run here? Anywhere? LOL!!

 

[Onglar]

lol

 

[cdawall]

All air baseline runs.

 

[EarthDog]

Hate you cdawall!!!


(Finally caught on...........because of your PM... now, back to your regularly scheduled cluelessness in this thread!)

 

[cdawall]

Hate you cdawall!!!


(Finally caught on...........because of your PM... now, back to your regularly scheduled cluelessness in this thread!)

Since apparently I have confused people let's clear some things up.

The CPU for this was on LN2, the GPU's were either chilled water or LN2 to reach those clocks, the clocks were not 17xx mhz they were in the 2200mhz range with boost. Anyone who would like to disagree with that is just wrong. Kingpin isn't magical he can't just make numbers appear on screen. This took a substantial amount of tweaking to accomplish. I am sure his parts were all heavily binned probably by himself from a stack of near perfect ASIC's.

 

[CAPSLOCKSTUCK]

Good on him i reckon its brilliant and not as easy as some might think. i would love to have the opportunity to play with kit like that

Take it for what it is........... an advert for EVGA.

 

[cdawall]

Good on him i reckon its brilliant and not as easy as some might think. i would love to have the opportunity to play with kit like that

Take it for what it is........... an advert for EVGA.

I still remember back when he was pushing AMD CPU's he had quite a few records with the FX57, 60 and 62.

 

[Vellinious]

Couple things.

The X99 EVGA motherboards are shit on a stick. They lack options in the bios that other boards have, and they take more voltage to achieve the same clocks as the ASUS and MSI boards...I owned 2 of them. Never again.

For those saying EVGA gave him the GPUs. The Titans aren't sold by board partners for pascal. NVIDIA isn't allowing it. If you want a TitanXP, you order from NVIDIA. They may have gave him the money for them, but there's no EVGA branding on a TitanXP, so.....no "advert for EVGA" happening there, other than the individual connecting his name with their brand. /shrug

2200 on a TitanXP is awesome....I would guess it would take a minimum of chilled water to get to those clocks and hold them long enough for even a short benchmark run, like the FS benchies.

Anyway...carry on. Just thought I'd make a couple of points.

 

[buildzoid]

Well, for what i know, there is no way to edit the pascal gtx bios, maybe i'm wrong but on the net there is no tool that showed up accessible to large amount of people, if it exist.

All Nvidia GPU vendors have access to non public BIOS editing tools. There's also HEXeditiors if you have the patience to reverse engineer the atrocity that is an Nvidia BIOS.

Other than that 2.2Ghz on a TITAN XP is hard. In the first few weeks after launch people were having trouble getting single cards into the 2.3GHz range because Nvidia has crammed them full of safety features and the Boost 3.0 algorithm is completely retarded when on LN2(if the card goes over 0C you get an extra +130mhz on the core clock which leads to a very quick crash).

Also I wish luck to anyone who thinks they can handle a 6950X at 5.2GHz without any past experience. I have a hard time with a 4790K and that pulls around half the power of the 6950X and that's ignoring all the issues that the X99 platform has with memory stability that get amplified with every GPU you add.

 

[cdawall]

Also I wish luck to anyone who thinks they can handle a 6950X at 5.2GHz without any past experience. I have a hard time with a 4790K and that pulls around half the power of the 6950X and that's ignoring all the issues that the X99 platform has with memory stability that get amplified with every GPU you add.

I am at 4.9 with my 5960x and 3 980ti's. but it is at the expense of memory speed I couldn't keep stability over 2400mhz. That chip also clocks a lot better than a 6950x will. People really don't realize how much tweaking was involved with all of that.

 

[cadaveca]

I am at 4.9 with my 5960x and 3 980ti's. but it is at the expense of memory speed I couldn't keep stability over 2400mhz. That chip also clocks a lot better than a 6950x will. People really don't realize how much tweaking was involved with all of that.

Meh. ES CPU = don't care if it pops (speaking about the topic at hand, not your chip). I know what Intel values these ES chips at, dollar wise, and I think most might find that figure very surprising.


I have one retail and one ES still in my possession. Guess which one clocks best?

Anyway, no matter how "skilled" you think Vince is, he's not designing these parts.. he literally is just changing BIOS options and sliders as someone suggested earlier. Yes, he has a lot of experience, but nearly ANYONE can change BIOS options and have the same results provided they have the same time to build experience and the same access to hardware.

The real heroes to this story are those that designed the parts that he used to get that score, IMHO, the EE's.