Irinikus wrote on 2023-09-23, 03:30:

If Pascal is as far as I'll be able to push it, then the Quadro P6000 is the best card that can be slotted here: (It's slightly more powerful than the TITAN XP)

Absolutely not, because power limits on Quadro cards are very tight. Also it has slower memory.

The Serpent Rider wrote on 2023-09-23, 04:20:

Irinikus wrote on 2023-09-23, 03:30:

If Pascal is as far as I'll be able to push it, then the Quadro P6000 is the best card that can be slotted here: (It's slightly more powerful than the TITAN XP)

Absolutely not, because power limits on Quadro cards are very tight. Also it has slower memory.

When you state power limits, what exactly do you mean? (CPU performance or Power Supply?) I don't intend do do any overclocking here!

So would you suggest the TITAN XP as the most powerful card. (Or card that would be most effective in this system?)

The TITAN XP has higher memory clocks, but the Quadro P6000 has higher GPU clocks

Just remember that the goal here it so get the absolute maximum performance out of this dual NetBurst system, regardless of the bottleneck on the GPU, meaning that you obviously have to dial back your expectations regarding the GPU's performance in this system (More than significantly), VS its actual capability in the type of system it's meant to be slotted into!

I want absolutely no compromise, otherwise this build is senseless!

Titan Xp has one 8-pin and one 6-pin PCIe connector. P6000 has one 8-pin connector which already works over official spec (175w).

The TITAN XP has higher memory clocks, but the Quadro P6000 has higher GPU clocks

Are you aware how modern GPU boosting algorithm works on Nvidia cards?

The Serpent Rider wrote on 2023-09-23, 04:48:

Titan Xp has one 8-pin and one 6-pin PCIe connector. P6000 has one 8-pin connector which already works over official spec (175w).

Ok I understand!

That would probably also require it to draw more power off the PCIe slot, which I certainly don't want! (I have seen a motherboard damaged in this way in the past! It was a cheap shitty one, but it happened non the less!)

This Tyan's pretty darn expensive so I absolutely don't want to run the risk of damaging it!

The Serpent Rider wrote on 2023-09-23, 04:48:

Are you aware how modern GPU boosting algorithm works on Nvidia cards?

No, I'm not

I do see that there's only 101MHz difference on the base clock between the cards, but the TITAN XP's memory's clocked 299MHz faster. (I assume that the cards won't make use of their boost clocks in this system.)

Making the TITAN XP the obvious better choice in this case. (And significantly cheaper!!!)

Thanks for your input! 😀 It's much appreciated!!! It's always good to thrash these things out! (That's ultimately what this hobby's all about!)

Making the TITAN XP the obvious better choice in this case.

Well, also no. Because some custom 1080Ti have much larger power limit, with 2x/3x 8-pin connectors, and better cooler.

The Serpent Rider wrote on 2023-09-23, 09:04:

Making the TITAN XP the obvious better choice in this case.

Well, also no. Because some custom 1080Ti have much larger power limit, with 2x/3x 8-pin connectors, and better cooler.

Please give me the model number of the card you have in mind?

I suppose this is one of the best of them: (ASUS ROG STRIX GTX 1080Ti OC)

Any top-tier card with custom PCB from brandname.

https://www.techpowerup.com/gpu-specs/galax-g … d-edition.b4588
https://www.techpowerup.com/gpu-specs/asus-ro … gaming-oc.b4306
https://www.techpowerup.com/gpu-specs/gigabyt … e-edition.b4329
https://www.techpowerup.com/gpu-specs/msi-gtx … htning-z.b10604

When comparing the TITAN XP to the ASUS STRIX, the TITAN's better in every way:

When comparing the TITAN XP to the GALAX, The GALAX has better overall theoretical performance, but its memory throughput is quite a bit worse than that of the TITAN (And to be really honest, the conservative look of the TITAN vs that of this card would be far more appropriate in this kind of build)

The really big question is what effect the internal (Internal to the card) memory throughput difference would make, in a situation such as this, where the card is so bottlenecked?

Also bear in mind that the guy who made the video showing the 1080TI in a P4 system mentioned that it was only 3DMark Fire Strike, which made the fans of that card spin up at all, indicating that the card won't be near to its maximum power dissipation in a system such as this one?

If I do indeed end up going for a 1080 Ti OC Edition, this will more than likely be the one that I'd go for: (The ZOTAC GTX 1080 Ti AMP! Extreme)

Here's how it ranks among the 1080 Ti's (It's the second most powerful card produced. There were only 50 of the Colorful iGame GTX 1080 Ti KUDAN's produced, so they're pretty much unobtainum at this point!!! 🙁 )

Here's how it compares to the TITAN XP: (The memory throughput is still quite a bit lower!!!)

Just to summarise, I'm in the process of doing something crazy, as I'm wanting to pair a modern GPU with a dual NetBurst Xeon from 2005.

The purpose of this exercise is to achieve the absolute maximum performance that I can from this system.

This is the Motherboard that I'll be using for this exercise: (It's a Tyan Thunder i7525 (S2676) fitted with two 604-pin Xeons @ 3.8GHz)

After discussions with "The Serpent Rider", I've reached a point where I'm considering these two GPU's: (Right click and select to open the image in a new tab if you can't see it clearly)

The ZOTAC 1080 Ti clearly has better theoretical performance, but the TITAN XP has significantly better memory throughput and more CUDA Cores.

So my question is the following: In the case where these GPU's are CPU bottlenecked to the extreme, will the difference in memory throughput (as well as the higher CUDA Core count on the TITAN) be an overriding factor when it comes to real world performance?

Remember that the goal here is to get maximum performance out of this dual NetBurst system, so please don't think as to how these cards would compare in the more modern type of system (i7) that was intended to run them.

Seeing that GPU's are SIMD-type systems, I'm personally of the opinion that the higher CUDA Core count would be a factor in this case. (Rather than the speed at which each individual core runs.) As you would accomplish more parallel calculations per instruction issued to the GPU by the CPU.

Seeing that the rate at which these CPU's can issue instructions the card are much lower than the desired rate, would greater parallelism not be a more desirable feature in this abnormal case?

Any inputs would be much appreciated!

Irinikus wrote on 2023-09-24, 10:48:

Just to summarise, I'm in the process of doing something crazy, as I'm wanting to pair a modern GPU with a dual NetBurst Xeon fr […]
Show full quote

Just to summarise, I'm in the process of doing something crazy, as I'm wanting to pair a modern GPU with a dual NetBurst Xeon from 2005.

The purpose of this exercise is to achieve the absolute maximum performance that I can from this system.

This is the Motherboard that I'll be using for this exercise: (It's a Tyan Thunder i7525 (S2676) fitted with two 604-pin Xeons @ 3.8GHz)

After discussions with "The Serpent Rider", I've reached a point where I'm considering these two GPU's: (Right click and select to open the image in a new tab if you can't see it clearly)

The ZOTAC 1080 Ti clearly has better theoretical performance, but the TITAN XP has significantly better memory throughput and more CUDA Cores.

So my question is the following: In the case where these GPU's are CPU bottlenecked to the extreme, will the difference in memory throughput (as well as the higher CUDA Core count on the TITAN) be an overriding factor when it comes to real world performance?

Remember that the goal here is to get maximum performance out of this dual NetBurst system, so please don't think as to how these cards would compare in the more modern type of system (i7) that was intended to run them.

Seeing that GPU's are SIMD-type systems, I'm personally of the opinion that the higher CUDA Core count would be a factor in this case. (Rather than the speed at which each individual core runs.) As you would accomplish more parallel calculations per instruction issued to the GPU by the CPU.

Seeing that the rate at which these CPU's can issue instructions the card are much lower than the desired rate, would greater parallelism not be a more desirable feature in this abnormal case?

Any inputs would be much appreciated!

Hi! Irinikus, I would like to get in touch with you. Not with regards to this but regarding a few Silicon Graphics questions. Apologies for dropping in here but not sure how else to reach you. Not sure if you have the ability to private message on here ?

You can PM me here or make contact with me on IRIX NETWORK or the Silicon Graphics User Group forums.

Irinikus wrote on 2023-09-27, 12:48:

You can PM me here or make contact with me on IRIX NETWORK or the Silicon Graphics User Group forums.

I am unable to PM (I dont think I have posted enough) but I will try the other options you suggested. Thank you!

After some thought I've placed an order for eight Thermaltake TOUGHFAN 14 Pro High Static Pressure 140mm Fans and one Thermaltake TOUGHFAN 12 Pro High Static Pressure 120mm Fan, together with a Thermaltake Commander FP 10-Port Hub for PWM Fans: (These ended up being a bit more expensive than the Noctua's but they should run more silently, while still offering good static pressure performance!)

Here are the specs on the TOUGHFAN 14 Pro High Static Pressure 140mm Fans:

Here are the specs on the TOUGHFAN 12 Pro High Static Pressure 120mm Fan:

Here are the specs on the Thermaltake Commander FP 10-Port Hub for PWM Fans:

Not sure if Thermaltake make their own exclusive fans, most likely you can buy these fans as OEM much cheaper. Also it won't be any better than typical Noctua offerings. If you want something better, look for fans with magnetic bearing.

The Serpent Rider wrote on 2023-09-29, 09:10:

Not sure if Thermaltake make their own exclusive fans, most likely you can buy these fans as OEM much cheaper. Also it won't be any better than typical Noctua offerings. If you want something better, look for fans with magnetic bearing.

They’ve been ordered, so that’s what it’s going to be, I’m afraid!

According to the reviews I've read, these fans compare with their Noctua equivalents, however they do have one feature which I really value, and that's an extremely long cable, which will prove useful when wiring this case up, as I won't need to use any extenders to fit the cables into the fan hub! (A feature which will give me greater flexibility when it comes to the fan hub placement and neat routing of the cables.)

This is what the system currently looks like , as it awaits its fans and RAM: (It will only start to look like something once the fans are installed!)

As for the wiring, I want to keep things as neat as possible, so the long fan cables will help allot!

I'll only start to experiment with GPU's once the fans and 16GB of RAM are fitted.

Planning for the fitment of the fans, I'm planning to have the bottom fan blowing outward (into the front cavity) and the top two blowing inward towards the CPU's

This is an attempt to overcome the restricted airflow into the front cavity due to the small air intake at the bottom, and the top two fans are the important ones, as they're the ones blowing directly onto the motherboard and CPU's)

The planned airflow pattern for the case: