*Official Benchmark Thread* - Post it here or it didn't happen :D

[jaybee83]

welp boiz i finished retesting Suprim Perf. / PNY Epic-X / Gigabyte Master / GIgabyte Master ICE vbios versions, this time with shorter, bursty benchmark runs.

 

Suprim came in last, with the PNY second to last. Master ICE often had the edge over the regular Master vbios, but its somehow more....volatile? like id have to rerun a bench several times to get that "good" score. with the regular Master, its just flash and forget, provides good scores with every run. so id say the Master is the overall most rounded vbios it seems, whereas if one is out for the best possible scores the Master ICE is a tad bit higher in scores (were mostly talking less than 1%, but sometimes up to 2-3%. but again, very volatlie and difficult to reproduce reliably).

 

and with those benchmark scores no wonder the guys over at overclock.net are swooning over it 😄

 

ill test run the Master vbios for a bit now. lately ive been having stuttering issues while gaming ever since switching over to the PNY vbios. hoping this will run more smoothly.

 

just thought id let you guys now, before anyone is disappointed with the PNY vbios 🙂 

 

as for the Suprim vbios...like, wth MSI?! with the 4090 it was basically the same: hardwarewise they know their stuff, cooling is awesome, same with fan noise/curves, power delivery but when it comes to vbios theyre like second fiddle....

Gigabyte Master ICE 10-Mar-2025 (274453).rom Gigabyte Master 22-Feb-2025 (273784).rom

[Mr. Fox]

5 hours ago, electrosoft said:

Kind of like a draw unfortunately. 😞

 

Was the Gene able to do 8400 too?

 

And yeah, the Dashboard and higher resolution BIOS is always nice. 🙂

No, not using the CPU and RAM I own. 8000-8200 was where they tapped out in the Gene. Those parts do better with the Apex but it is mostly a draw. The biggest win for the Apex is the fact that it is not mATX. The Gene is as great as it can be considering when it was made in spite of its form factor.

[tps3443]

5 hours ago, electrosoft said:

Linus building out the fastest PC on the planet  9800X3D + Astral 5090 with Splave.

 

Splave opted to go with the Taichi over the Asus....

 

Timestamped at the point they are "flashing a top secret unlocked BIOS" to the 5090:

 

 

Note the (sorry you can't do this part at home)....

 

Slider now goes to 120%

 

Pulling 890w during the last run....

 

Watching all the pins trip red.... 🙂

 

Crashing and after a few more crashes Splave to Linus "Is your card worse than mine?" and a subsequent bit on Silicon Lottery.

 

 

Kind of like a draw unfortunately. 😞

 

Was the Gene able to do 8400 too?

 

And yeah, the Dashboard and higher resolution BIOS is always nice. 🙂

 

 

 

Yeah, I thought they were ok with it till it went off the rails at the end. I just kind of went, "whoa, whoa, I thought you liked it?" 😕

 

4k non G-sync was an issue. I don't like 4k panels on laptops. QHD for 17.3" or greater or FHD for 16" or smaller is my preference.  I never had a problem with the trackpad but I'm a lefty. I don't know if that makes a difference. There is a new BIOS out for my unit I plan on trying along with tearing it down for a repaste and eval on extra copper under the hood soon as I get some time to start truly tinkering.

 

As for pricing, yeah they've come down a bit. On launch, they had a "special" price of $3999.99 for the Raider 285/5090 then it went back up to its "regular" price of $4899 but was then almost immediately discounted to $4499 and now it's at a regular price of $4099. 6400 speed ram and 2tb SSD isn't terrible but still....

 

Even the normal Titan dropped below $5k almost at launch.

 

9950HX3D models are all I would be interested in on the top end atm for gaming for 5090s though as the overall uplift versus the previous generation is mediocre.

 

 

I'm still waiting for Acer to send out their unit to me with a 275hx/5070ti OLED but they are currently switching distribution lines because of China tariffs for eval units because they will be hit regardless of import status. They had to push everything back to June according to their rep while they realign their supply lines.

 

 

Here's the cure to all your ills Brother.... 🤣

 

Or step into the 2 dimm land of the Apex

 

 

 

Yeah, this has rolled around in my mind for quite some time every since the launch of the 5090 series. The market continues to show Nvidia it can charge more and more for their halo products but I am a firm believer a product is only worth as much as consumers will pay.

 

For the lower tiers, Nvidia has routinely hit walls or competition and has had to price down accordingly, but the top end? Sky's the limit and even 5080 AIB models have skyrocketed into formerly 4090 land and brimming on 5090FE/PNY/Gigabyte launch pricing.

 

It really doesn't matter who won't pay but moreso who WILL pay.

 

Supply and Demand baby......supply and demand.....

 

----------------------------------------------------------------------------

 

On another note, my friend finally wised up and I snagged an Astral for him from Best Buy and he sold off his FE for $3500 locally cash. I just had a chance to test it earlier today. It hit +3000 no problem on memory but it's boost lock is only 2919 and 2970 vs 2977 and 3060 on mine. Voltage is 1.065 and 1.085 max vs 1.085 and 1.1105 on mine. It also as quite audible coil whine but nothing compared to his FE which sounded like R2D2 taking a shot while navigating. We just finished up with a two hour FT jam session and I told him I could return it and wait for another unit if he wants but he still wants it and plans on blocking it (well, having his son block it for him) so off it goes tomorrow to its new home.

 

 

 

 

 

@electrosoft The Astral in that video was just like yours! It too was 1115mv@2977. Anyways, you need to subtract your idle clocks to get a fair comparison between these Astral's VFcurve. 

So, my FE is idle 172-180Mhz. The Astral might idle at 226Mhz-262Mhz which would mean you need to subtract to get your correct VFcurve. Your cards might idle at a different frequency. This is all for fair comparison if you are comparing Astral's. So, whatever your idle is over +180, that is how much you will subtract from your VFcurve.. If I go just +75 on my core my cards 1065mv point is now 2,997 , my card would also idle at 240Mhz. Anyways, so with Astral typically its 225-262Mhz is idle but I heard it varies. So, if your 1085MV is 2,970Mhz. This number is inflated by your factory OC which is inflating the VFcurve. So, Check your idle clocks, subtract until you get 180 then subtract the difference from your Vfcurve point, and that is your true VFcurve. 

Idle Clock Speed= 232Mhz
 232-180= 52

Original VFcurve 2,977@1.115mv Minus 52= 2925Mhz. "True 1.125MV Vf Curve= 2925"

^This is just an example. 

But, I would definitely bench these two cards head to head. Not that the voltage matters all that much. I personally am after a high VFcurve regardless of what its max voltage is or may be. Also, make sure its on OC bios, that will affect its VFcurve directly. 

Also, do not discount that lower voltage card. It may do very well. My FE does very well it's only a 1050mv/1075mv. But so long as your VFcurve is good, it does not matter what the heck your max voltage is. 

You are going to have to put the pedal to the metal on both Astrals in Steel Nomad/Port Royal and see which one wins! Who knows, they could clock the same. I want to see some numbers!!! 😀
 

[electrosoft]

3 hours ago, tps3443 said:

@electrosoft The Astral in that video was just like yours! It too was 1115mv@2977. Anyways, you need to subtract your idle clocks to get a fair comparison between these Astral's VFcurve. 

So, my FE is idle 172-180Mhz. The Astral might idle at 226Mhz-262Mhz which would mean you need to subtract to get your correct VFcurve. Your cards might idle at a different frequency. This is all for fair comparison if you are comparing Astral's. So, whatever your idle is over +180, that is how much you will subtract from your VFcurve.. If I go just +75 on my core my cards 1065mv point is now 2,997 , my card would also idle at 240Mhz. Anyways, so with Astral typically its 225-262Mhz is idle but I heard it varies. So, if your 1085MV is 2,970Mhz. This number is inflated by your factory OC which is inflating the VFcurve. So, Check your idle clocks, subtract until you get 180 then subtract the difference from your Vfcurve point, and that is your true VFcurve. 

Idle Clock Speed= 232Mhz
 232-180= 52

Original VFcurve 2,977@1.115mv Minus 52= 2925Mhz. "True 1.125MV Vf Curve= 2925"

^This is just an example. 

But, I would definitely bench these two cards head to head. Not that the voltage matters all that much. I personally am after a high VFcurve regardless of what its max voltage is or may be. Also, make sure its on OC bios, that will affect its VFcurve directly. 

Also, do not discount that lower voltage card. It may do very well. My FE does very well it's only a 1050mv/1075mv. But so long as your VFcurve is good, it does not matter what the heck your max voltage is. 

You are going to have to put the pedal to the metal on both Astrals in Steel Nomad/Port Royal and see which one wins! Who knows, they could clock the same. I want to see some numbers!!! 😀
 

 

No, mine is 2977 @ 1.085. 🙂

 

 

 

 

[tps3443]

19 minutes ago, electrosoft said:

 

No, mine is 2977 @ 1.085. 🙂

 

 

 

 

Right, but that’s including a factory applied OC if so, it’s not the true VFcurve. 
 

To find out what your real core offset is on both Astrals, you have to subtract their stock idle speed from 180Mhz. 
 

Nvidia spec 5090 idle is 180Mhz. So, if your card idles any bit over 180Mhz, you have to subtract that from your 2977Mhz which correlates your real VFcurve. 
 

If your card idles 190Mhz it’s 2967@1085mv

 

If your card idles at 230Mhz it’s 2927@1085mv.

 

Anything on top of 180Mhz is a overclock on top of the default clock pre-applied through the bios from the AIB.

@electrosoft 5090 Astrals can have different idles. Crazy I know. But they can! So this can skew the default VF curve and what it actually may be. 
 

See below. 
 

[electrosoft]

1 hour ago, tps3443 said:

Right, but that’s including a factory applied OC if so, it’s not the true VFcurve. 
 

To find out what your real core offset is on both Astrals, you have to subtract their stock idle speed from 180Mhz. 
 

Nvidia spec 5090 idle is 180Mhz. So, if your card idles any bit over 180Mhz, you have to subtract that from your 2977Mhz which correlates your real VFcurve. 
 

If your card idles 190Mhz it’s 2967@1085mv

 

If your card idles at 230Mhz it’s 2927@1085mv.

 

Anything on top of 180Mhz is a overclock on top of the default clock pre-applied through the bios from the AIB.

@electrosoft 5090 Astrals can have different idles. Crazy I know. But they can! So this can skew the default VF curve and what it actually may be. 
 

See below. 
 

 

 

True story, from the jump,  I never, ever used the above approach (or facsimiles thereof from before) to calculate scores or ranking for 5090's before I even had a 5090. I want you to figure out why that is and why the above approach generalized is invalid.

 

--------------

 

I am noticing this new 1401 BIOS, my system temps and especially idling temps are much lower. I'm used to seeing the nifty diagnostic LED showing temps in the mid 40s during desktop/light work, but it is sitting solidly in the low to mid 30s now doing the same stuff.

 

This is with my exact same settings from before re-dialed in.

 

 

 

 

 

 

 

 

 

 

 

[tps3443]

12 minutes ago, electrosoft said:

 

 

True story, from the jump,  I never, ever used the above approach (or facsimiles thereof from before) to calculate scores or ranking for 5090's before I even had a 5090. I want you to figure out why that is and why the above approach generalized is invalid.

 

 

 

 

 

 

 

 

 

Simple explanation. 

My card idles at 180Mhz if I overclock it and add +50Mhz to my core offset it will idle at 230Mhz now. My VFcurve is also gone up +50Mhz, so instead of 2932@1065mv it is now 2982@1065mv. If I add a +475Mhz offset to my core, it’ll idle at 655Mhz. And my VFcurve would go up to 3407Mhz@1065mv
 

Your card has a factory overclock. You must subtract the idle speed from 180Mhz. If I could flash your bios to my FE, it too would have a factory overclock which would increase my VFcurve just like manually overclocking, which increases your VFcurve. You are not factoring what the silicon would run by its self based on its programmed VF curve, you are accounting for that+ the overclock.

 

[electrosoft]

10 minutes ago, tps3443 said:

Simple explanation. 

My card idles at 180Mhz if I overclock it and add +50Mhz to my core offset it will idle at 230Mhz now. My VFcurve is also gone up +50Mhz, so instead of 2932@1065mv it is now 2982@1065mv. If I add a +475Mhz offset to my core, it’ll idle at 655Mhz. And my VFcurve would go up to 3407Mhz@1065mv
 

Your card has a factory overclock. You must subtract the idle speed from 180Mhz. If I could flash your bios to my FE, it too would have a factory overclock which would increase my VFcurve just like manually overclocking, which increases your VFcurve. You are not factoring what the silicon would run by its self based on its programmed VF curve, you are accounting for that+ the overclock.

 

 

No, I get the methodology.

 

I also get why this approach is flawed and invalid.

 

Do you?

 

 

[jaybee83]

speaking of 5090 VCores,  i dont think ive seen ANY other card with as low as a peak as mine. absolute max i can see at idle and low temps is 1.025V. but under load i would never even see 1.02, 1.015 is as good as it gets.

 

so guys, out of curiosity: whats the highest D2D OC u can get on your cards with a given max VCore? were talking maxed out PL and voltage sliders here and max boost clocks shown at 30C core temp at idle.

 

mine would be 3240 Mhz at 1.025V (max. PL, max. voltage slider, gpu temp 30C).

[tps3443]

8 hours ago, electrosoft said:

 

No, I get the methodology.

 

I also get why this approach is flawed and invalid.

 

Do you?

 

 

I don’t know why it may be flawed. I just know that’s how everyone else gets/compares their baked in VFcurve on their GB202 die.

 

5 minutes ago, jaybee83 said:

speaking of 5090 VCores,  i dont think ive seen ANY other card with as low as a peak as mine. absolute max i can see at idle and low temps is 1.025V. but under load i would never even see 1.02, 1.015 is as good as it gets.

 

so guys, out of curiosity: whats the highest D2D OC u can get on your cards with a given max VCore? were talking maxed out PL and voltage sliders here and max boost clocks shown at 30C core temp at idle.

 

mine would be 3240 Mhz at 1.025V (max. PL, max. voltage slider, gpu temp 30C).

What application are you going by to show these max clocks?
 

Run steel nomad, and tell me your GPU voltage range when he’s driving through the gate on his hover bike. Reset your card to default clocks.

[Papusan]

 

Gigabyte said just clean it off. Sure, maybe do it weekly as a part of the maintenance/check for signs of melted trash connectors. In the end it will be no more putty on the components. This will hunt Gigabyte's reputation and RMA cost will increase within the cards support warranty period (Gigabyte offer 4 years if you register your card). Good they have larger margins due disgusting prices above MSRP. Yep, cost saving on the paste, gel and the job will in the end cost them more than what they saved/jacked up prices. Payback time for greed. And sad for the owners that need to send in their cards for RMA due badly executed job on the thermal concept (again due greed) on overpriced graphics cards.

Gigabyte leaky GPU problem continues, here's the thermal putty creep in action

Gigabyte's so-called "Server-Grade Thermal Conductive Gel" and a degrading thermal paste on the RX 9070 Gaming OC | Test Lab

Finally, there's the bitter aftertaste that Gigabyte apparently prefers to bask in the comforting haze of buzzy buzzwords like "server-grade thermal conductive gel" instead of simply developing a well-thought-out overall thermal concept. The fact that they opted for viscous paste with questionable long-term performance instead of proven PTM pads, precisely where thermomechanical stability is essential—namely, directly on the sensitive GPU die—seems not only unwise but almost cynical in retrospect.

[Mr. Fox]

16 hours ago, electrosoft said:

 

No, mine is 2977 @ 1.085. 🙂

Who would have ever thought that GPU core clock speeds (in MHz) would be smaller than the base price (pre-tax) price tags on NVIDIA GPUs? Historically, the exact opposite has been true.  Looks like the new price model is more than $1 per MHz.

[jaybee83]

8 hours ago, tps3443 said:

I don’t know why it may be flawed. I just know that’s how everyone else gets/compares their baked in VFcurve on their GB202 die.

 

What application are you going by to show these max clocks?
 

Run steel nomad, and tell me your GPU voltage range when he’s driving through the gate on his hover bike. Reset your card to default clocks.

 

any gpu tuning software u prefer, be it PX1, MSI AB, GPU Tweak, etc. and enable locked max boost, thats it 🙂 

[electrosoft]

4 hours ago, Papusan said:

 

Gigabyte said just clean it off. Sure, maybe do it weekly as a part of the maintenance/check for signs of melted trash connectors. In the end it will be no more putty on the components. This will hunt Gigabyte's reputation and RMA cost will increase within the cards support warranty period (Gigabyte offer 4 years if you register your card). Good they have larger margins due disgusting prices above MSRP. Yep, cost saving on the paste, gel and the job will in the end cost them more than what they saved/jacked up prices. Payback time for greed. And sad for the owners that need to send in their cards for RMA due badly executed job on the thermal concept (again due greed) on overpriced graphics cards.

Gigabyte leaky GPU problem continues, here's the thermal putty creep in action

Gigabyte's so-called "Server-Grade Thermal Conductive Gel" and a degrading thermal paste on the RX 9070 Gaming OC | Test Lab

Finally, there's the bitter aftertaste that Gigabyte apparently prefers to bask in the comforting haze of buzzy buzzwords like "server-grade thermal conductive gel" instead of simply developing a well-thought-out overall thermal concept. The fact that they opted for viscous paste with questionable long-term performance instead of proven PTM pads, precisely where thermomechanical stability is essential—namely, directly on the sensitive GPU die—seems not only unwise but almost cynical in retrospect.

 

Well there goes my vertical mount..... 🤣

 

Sad really, because the card is solidly built all around....

 

11 hours ago, jaybee83 said:

speaking of 5090 VCores,  i dont think ive seen ANY other card with as low as a peak as mine. absolute max i can see at idle and low temps is 1.025V. but under load i would never even see 1.02, 1.015 is as good as it gets.

 

so guys, out of curiosity: whats the highest D2D OC u can get on your cards with a given max VCore? were talking maxed out PL and voltage sliders here and max boost clocks shown at 30C core temp at idle.

 

mine would be 3240 Mhz at 1.025V (max. PL, max. voltage slider, gpu temp 30C).

 

With everything maxed, fans 100% to bring down the temps to sub 30c, I'm hitting 3348+ at +327 / +3000 1.105 which is my D2D.

 

How high can you push your boost clocks locked just for no load / idle shiggles? Does it allow you to push past 3300? 3400?

 

I did discover that running PX1 and GT3 will trip the power monitoring doing absolutely nothing and light them all up red randomly.

 

In the end my check list is:

 

No coil whine

3ghz+ gaming D2D profile at all times

fairly quiet under load

Gamerz branded colorz and lightz! (JK)

 

----------------------------------------------------------------------

 

@tps3443 GT3 and PX1 are reporting a default PL of 107% not 104% so even a touch higher than you thought. 🤣

 

5 hours ago, Mr. Fox said:

Who would have ever thought that GPU core clock speeds (in MHz) would be smaller than the base price (pre-tax) price tags on NVIDIA GPUs? Historically, the exact opposite has been true.  Looks like the new price model is more than $1 per MHz.

 

 

Agreed 100%. The pricing is just sick and gluttonous at this point but compared to other leisurely time pursuits, it is on the low end actually with an actual ability to recoup your investment and sometimes more.

 

Can't say the same for many other pursuits but it still doesn't excuse the pricing.

 

 

[Papusan]

Nice. Expect lower performance even up to 14th gen Intel Core processors. Me upgrading firmware? Nope. I'm still on the older 9905. And if the processor die or degrade I just ask for a new one. Or my money back.

 

 

23 minutes ago, electrosoft said:

Well there goes my vertical mount..... 🤣

 

Sad really, because the card is solidly built all around....

 

Maybe Gigabyte should deny repair/RMA if you confirm to them that you have used your GPU in vertical mount?😁 Remember Gigabyte is damn happy to use their stickers with the fat big red arrow stamped.

[johnksss]

Question, now that you guys have had a chance to play around with your cards.

 

What is your stock and top overclocked scores for Steel Nomad today?

@electrosoft @jaybee83

 

@tps3443 Were you able to run and keep your memory overclock speeds at 2125 average the whole run?

[tps3443]

5 hours ago, jaybee83 said:

 

any gpu tuning software u prefer, be it PX1, MSI AB, GPU Tweak, etc. and enable locked max boost, thats it 🙂 

Mine is was normally 3442@30c my voltage was not hitting 1075mv though, I think it's because the card was 31c this evening, and when I last checked I was on driver 572.83 which may be why it was 3,442Mhz before, right now I am on 576.28. Just had a hard gaming session though. Card was absolutely heat soaked to the max! 😄

[electrosoft]

1 hour ago, tps3443 said:

Mine is was normally 3442@30c my voltage was not hitting 1075mv though, I think it's because the card was 31c this evening, and when I last checked I was on driver 572.83 which may be why it was 3,442Mhz before, right now I am on 576.28. Just had a hard gaming session though. Card was absolutely heat soaked to the max! 😄

 

Yeah, I'm on 576.28 too. I did find 572.83 was better overall for clocks.

 

The wife and I finished up some serious WoW'ing and off to bed she went room is toasty. I rediscovered my love of my ceiling fan since the 5090 install and warmer days.

 

Seriously, when all is said and done, all roads lead to the same ballpark. Just slightly different seating. Only thing that will get you where you want to go on top of silicon is cooling and shunting along with all the tweaks the dedicated do to their environment/OS to extract every bit out of their cards. Me? Anything and everything I do is on top of my D2D bloaty OS. 

 

------

 

+3000 / +496 maxed nets me ~3465  at ~1065-1070 on 576.28. I have 576.40 drivers downloaded but still waiting a few days to see which direction they go.

 

The higher I pushed the offset, the lower the voltage dropped with no change to the cap out at 3465.

 

+465 / +3000  and +496 / +3000 = 3465.

 

Try lowering your offset to ~+465 and see if you maintain 3412 (don't forget to OC your memory too to max) and if your voltage goes up a bit.

 

+496 / +3000

 

 

+461 / +3000:

 

 

-----------

 

5090 mobile getting raked over the coals by Jarrod.....

 

 

 

 

 

 

[tps3443]

1 hour ago, electrosoft said:

 

Yeah, I'm on 576.28 too. I did find 572.83 was better overall for clocks.

 

The wife and I finished up some serious WoW'ing and off to bed she went room is toasty. I rediscovered my love of my ceiling fan since the 5090 install and warmer days.

 

Seriously, when all is said and done, all roads lead to the same ballpark. Just slightly different seating. Only thing that will get you where you want to go on top of silicon is cooling and shunting along with all the tweaks the dedicated do to their environment/OS to extract every bit out of their cards. Me? Anything and everything I do is on top of my D2D bloaty OS. 

 

------

 

+3000 / +496 maxed nets me ~3465  at ~1065-1070 on 576.28. I have 576.40 drivers downloaded but still waiting a few days to see which direction they go.

 

The higher I pushed the offset, the lower the voltage dropped with no change to the cap out at 3465.

 

+465 / +3000  and +496 / +3000 = 3465.

 

Try lowering your offset to ~+465 and see if you maintain 3412 (don't forget to OC your memory too to max) and if your voltage goes up a bit.

 

+496 / +3000

 

 

+461 / +3000:

 

 

-----------

 

5090 mobile getting raked over the coals by Jarrod.....

 

 

 

 

 

 

Can you try with MSI AB? GPU Tweak 3 is inflating mine for some reason. 

[electrosoft]

49 minutes ago, tps3443 said:

Can you try with MSI AB? GPU Tweak 3 is inflating mine for some reason. 

 

I double checked it with PX1 and AB. 🙂

 

I actually just used PX1 to boost it while having them running simultaneously. That is also when I found out PX1 was somehow tripping the power sensors detector on GT3. 🤨

 

But all three read the same. Also one reason why I decided on GT3 since it does everything I need and the others are kind of redundant but GT3 will freeze/lock up/crash randomly when setting it to launch with Windows so I manually start it.

 

 

I did snag a shot of PX1 running too with GT3 when I was cross referencing earlier: