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

[Papusan]

31 minutes ago, electrosoft said:

 

Did Galax officially cancel the HOF or is it just a rumor?

 

 

 

Rumors but most of Galax cards will be D models for the Asia market. And I mean we would see HOF cards by now if they where made Is this the end of GALAX HOF?

 

"According to our information, GALAX will be using 90% 5090D chips and 10% 5090 chips"

 

And overclocking.com editors have always had love for the HOF models. I'm sure they would find one in the 5000 series if they was real.

 

Edit. I think also Colorful will prioritise the Asia market. Aka most 5090 chips will be the D variant.

COLORFUL GeForce RTX 4090 iGame LAB GPU Claims.....

 

Everyone have China as number one market...

 

MSI optimizes affordable Intel 800 mobos for China's first homegrown DDR5 memory chips

[Tenoroon]

As I typically do, I have been lurking here, and many of you don't seem too happy with the state of technology right now, which I agree with very much. While it will be a LONG ways out, mark my words when I say I will eventually cobble together some motherboard(s) that gives complete control over the pre-BGA mobile chips, so everyone here can do their magic and dominate the HWbot leaderboards.

 

I'm already a third of the way done getting my undergrad as a EE with the intent to get a masters degree. I will try as hard as possible to make this happen, and if it doesn't then I will know that my education has failed me or I have failed myself.

Thought I would share this with all of you to give a little bit of hope during a time where things seem so bleak. It might be a while, but I'll eventually figure it out 🙂

[Papusan]

Bro @Mr. Fox.... Here in Norway we are now come to 27. January so happy birthday my good old friend❤️🎂🥳

 

 

 

[tps3443]

5 hours ago, Mr. Fox said:

Well sure you would. So would I. But I don't think I'd be happy with everybody that owned a 5090 being capped at the same maximum memory overclock. That totally defeats the purpose of owning a product you bought to overclock. Super stupid. The silicon quality should determine the max overclock not the NVIDIOT Nazis.

 

That's one of the Ryzen gotchas, too. You can overclock as much as you want to as long as you're happy raising BCLK to bypass the stupid low fused limits on the CCD core ratios.

 

Gamer sheeple ruin everything for overclocking enthusiasts and the idiots that make the hardware seem to think that gamers and overclockers are the same thing. It must really suck for them to be that stupid, but they probably don't even have an awareness of it.

---

I need a better 9950X so I can run the FCLK at 2200+ but the memory seems to be doing well enough.

Well, at least it’s better than me hitting 34Gbps, and you only hitting 31Gbps due to dud GDDR7. (WE ALL HIT 34Gbps!!!)) lol. 😂 

[Mr. Fox]

1 hour ago, Papusan said:

Bro @Mr. Fox.... Here in Norway we are now come to 27. January so happy birthday my good old friend❤️🎂🥳

 

 

 

Thank you, Brother. Next year my age and year of birth will be the same. 

1 hour ago, tps3443 said:

Well, at least it’s better than me hitting 34Gbps, and you only hitting 31Gbps due to dud GDDR7. (WE ALL HIT 34Gbps!!!)) lol. 😂 

In order for winners to exist there has to be losers. It totally sucks to lose, but it sucks worse when winners are the same as losers. When everyone is a winner then everyone becomes a loser. Losers are what makes being a winner so special.😝

  

5 hours ago, jaybee83 said:

tru, the perf bump would be REAL, and that at half price 😛 guessing time: would 16 P cores be enough to beat a 9950X? 

Yeah, it would beat it pretty soundly. My 13900KS and 14900KS beat my 9950X at many things already. They lose to the 9950X in a few things like Cinebench because smartphone cores without hyperthreading suck.

[Papusan]

Here's where Jensens products lands via their partners.... It seems no one can stop Jensen and nvidia. Probably only the small top of the big iceberg. And only god know how many real 4090 chips and GDDR6X packages with 48GB vram has been pushed into China. We will see the same with Blackwell chips. Not the castrated 5090D die's. The real one's.

 

Chinese AI firm DeepSeek has 50,000 NVIDIA H100 AI GPUs says CEO, even with US restrictions

Chinese AI lab DeepSeek has access to tens of thousands of NVIDIA H100 AI GPUs for training, according to DeepSeek CEO.

 

[Mr. Fox]

33 minutes ago, Papusan said:

Here's where Jensens products lands via their partners.... It seems no one can stop Jensen and nvidia. Probably only the small top of the big iceberg. And only god know how many real 4090 chips and GDDR6X packages with 48GB vram has been pushed into China. We will see the same with Blackwell chips. Not the castrated 5090D die's. The real one's.

 

Chinese AI firm DeepSeek has 50,000 NVIDIA H100 AI GPUs says CEO, even with US restrictions

Chinese AI lab DeepSeek has access to tens of thousands of NVIDIA H100 AI GPUs for training, according to DeepSeek CEO.

China should not be allowed to have anything except for their dirty air and dirty water that nobody else wants.

On 1/25/2025 at 11:04 AM, Raiderman said:

It's stupid and deliberate manipulation. I'm getting fed up with it!

Just the fact that they did something that made it incompatible with OSes before Winduhz 11 shows a pretty sinister intent. There should not be any such thing as "Windows 7" or "Windows 10" or "Windows 11" drivers. There should be drivers for Windows x64. Period. Whatever it is that makes it Windows 11 dependent and incompatible with older versions probably isn't doing Linux users any favors either.

[Papusan]

I had Gigabyte on the list with nope. See my previous post. Not first time I see similar from this brand. 

 

https://www.overclock.net/posts/29419721/

[Mr. Fox]

14 hours ago, Tenoroon said:

Thought I would share this with all of you to give a little bit of hope during a time where things seem so bleak. It might be a while, but I'll eventually figure it out 🙂

This is a good time for all of us to sit back, do nothing, spend nothing, and reflect on the things that actually matter in life. Sometimes we get wound up like an 8-day clock about things that don't matter all that much in the grand scheme of things. The only way they will listen is if the cash stops flowing. Budweiser knows how that works now. There are lots of companies, in many industries, that need to learn hard lessons the same way. But, there may not be enough people that are smart enough to teach those lessons. This is the wildcard. Stupid is dangerously common.

 

From 9:48 (the meat of the video) - this is in line with what I have been saying. 

 

[Papusan]

4 hours ago, Mr. Fox said:

This is a good time for all of us to sit back, do nothing, spend nothing, and reflect on the things that actually matter in life. Sometimes we get wound up like an 8-day clock about things that don't matter all that much in the grand scheme of things. The only way they will listen is if the cash stops flowing. Budweiser knows how that works now. There are lots of companies, in many industries, that need to learn hard lessons the same way. But, there may not be enough people that are smart enough to teach those lessons. This is the wildcard. Stupid is dangerously common.

 

From 9:48 (the meat of the video) - this is in line with what I have been saying. 

 

 

This isn't the end. It's the start.... 

 

DeepSeek R1: The Chinese AI model that has made NVIDIA lose 384 billion overnight

NVIDIA Corporation has seen its stock market value lose more than $384 billion overnight, and this is due to the Chinese artificial intelligence startup DeepSeek.

 

Exactly what Nvidia  need...

Chinese AI company DeepSeek is making some serious waves in the world of AI, with its R1 model being trained for 90-95% less money... making the industry re-consider all those AI GPU purchases... and seeing NVIDIA stock dropping 11% in pre-market trading, bleeding out $384 billion. Ouch.

 

 

And China/budget-conscious creators/prosumers want that 24GB framebuffer. 

Leak: RTX 5080 performs worse than RTX 4090

The last-generation flagship is up to 23 percent faster than the RTX 5080.

[Mr. Fox]

Will the new Astral GPUs feature extra rust for the extra premium?
 

 

[Papusan]

50 minutes ago, Mr. Fox said:

Will the new Astral GPUs feature extra rust for the extra premium?
 

 

 

Hmmm. Maybe the reason Asus this time claim they added Protective PCB Coating (against moisture and debris) for their flagship cards.  So probably a well known problem with previous gen cards. But you pay extra premium for it.

[Mr. Fox]

6 minutes ago, Papusan said:

Hmmm. Maybe the reason Asus this time claim they added Protective PCB Coating (against moisture and debris) for their flagship cards.  So probably a well known problem with previous gen cards. But you pay extra premium for it.

Maybe it has something to do with the robotic soldering they brag about or the cleaning method they use. Perhaps whatever chemical they use to wash them after soldering. It's odd that their brand is the only one where it is seen, so it has to be something they do differently in manufacturing. Perhaps it is easier for them to spray some kind of crap coating on the GPU than to clean it properly.

[win32asmguy]

19 hours ago, Mr. Fox said:

What I really want is an LGA1700 15900KX 16 P-core hyperthreaded no E-core 6.0GHz AVX-512 CPU. Then I wouldn't care that my 9950X can't do 2200+ FCLK.

 

That is why I built a sapphire rapids desktop. Pure P-core chip without artificial limitations. The RDIMM requirement does increase memory latency but it is not enough to make gaming less enjoyable. I guess 9950X is much less expensive from a cost per P-core stance but then you have to deal with issues created by a non-monolithic die and the fact that it is sold with different varying bin CCDs.

[Mr. Fox]

55 minutes ago, win32asmguy said:

 

That is why I built a sapphire rapids desktop. Pure P-core chip without artificial limitations. The RDIMM requirement does increase memory latency but it is not enough to make gaming less enjoyable. I guess 9950X is much less expensive from a cost per P-core stance but then you have to deal with issues created by a non-monolithic die and the fact that it is sold with different varying bin CCDs.

My next build, if and when there is one, will likely be a HEDT with 3090 Ti SLI. New and shiny is looking less pretty all the time.

 

Very valid points about why making a 5090 FE waterblock is going to be an unattractive proposition. I suspect this will be a side benefit for NVIDIA.

 

[Papusan]

49 minutes ago, Mr. Fox said:

Very valid points about why making a 5090 FE waterblock is going to be an unattractive proposition. I suspect this will be a side benefit for NVIDIA.

 

One thing for sure... The repair shops have good times forwards. Just as NorthridgeFix🙂 Nvidia made a huge business replace melted 12VHPWR connectors. Now they need to clean up LM spills, +  replace mosfets, shunts, caps and inductors as well. 

 

[electrosoft]

6 hours ago, Mr. Fox said:

This is a good time for all of us to sit back, do nothing, spend nothing, and reflect on the things that actually matter in life. Sometimes we get wound up like an 8-day clock about things that don't matter all that much in the grand scheme of things. The only way they will listen is if the cash stops flowing. Budweiser knows how that works now. There are lots of companies, in many industries, that need to learn hard lessons the same way. But, there may not be enough people that are smart enough to teach those lessons. This is the wildcard. Stupid is dangerously common.

 

From 9:48 (the meat of the video) - this is in line with what I have been saying. 

 

 

3 hours ago, Papusan said:

 

Leak: RTX 5080 performs worse than RTX 4090

The last-generation flagship is up to 23 percent faster than the RTX 5080.

 

The pros of the 5080 that even Nvidia recognized is that it is marginally faster than the 4080 Super. They also kept the price the same instead of raising it even in the face of inflation and past trends. I think we keep forgetting that.

 

You are getting a newer card, with newer features that is 10-15% faster for the same price.

 

$999.99 in 2022 is $1059 in today's money

Unlike classic Nvidia they didn't raise the price of the card

You are getting more performance than before

 

I think Nvidia knows this is a soft cycle and they need a newer architecture and a node shrink for next time. In Intel speak, in the land of Tick-Tock, this is definitely Tick-Tick land.

 

This is also why they lowered the price of the 5070ti and kept the 5070 the same as last gen.

 

Last gen launch vs this gen launch:

 

3070($499)->4070($549) = 10% price hike. 4070-5070= same price

3070ti($599)->4070ti($799) = ~33% price hike. 4070ti($799)->5070ti($749) = ~7% price reduction

3080($699)->4080($1199) =  ~33% price hike. 4080($1199) ->5080 ($999) = ~20% price reduction

3090($1499)->4090($1599) = ~7% hike. 4090($1599) ->5090 ($1999) = ~25% price HIKE

 

5090 is where the real shanking is taking place. Everywhere else you're getting more for less actually even the 5070 with inflation calculated in.

 

 

 

 

 

 

[jaybee83]

17 hours ago, Mr. Fox said:

Yeah, it would beat it pretty soundly. My 13900KS and 14900KS beat my 9950X at many things already. They lose to the 9950X in a few things like Cinebench because smartphone cores without hyperthreading suck.

 

would it though? iirc the baby cores were at about 0.67x IPC of the respective P cores. at 16 E-cores, that would equate to 16*0.67 = 10.7 P-Cores

 

gotta consider the frequency differences too, so lets say 4.3 Ghz max. for E cores and 6.0 Ghz max. for P cores, thats a factor of 1.4 in advantage for P cores.

 

10.7 P-Cores / 1.4 = 16 E-Cores equate to 7.4 P Cores when both IPC and frequency are considered.

 

so current configs at 8 P-Cores + 16 E-Cores would already equate to 8 + 7.4 P-Cores = 15.4 P Cores.

 

A 16 P-Core config would then just be only +4% faster overall at full utilization...would that be such a huge difference then?

 

3 hours ago, Papusan said:

 

This isn't the end. It's the start.... 

 

DeepSeek R1: The Chinese AI model that has made NVIDIA lose 384 billion overnight

NVIDIA Corporation has seen its stock market value lose more than $384 billion overnight, and this is due to the Chinese artificial intelligence startup DeepSeek.

 

Exactly what Nvidia  need...

Chinese AI company DeepSeek is making some serious waves in the world of AI, with its R1 model being trained for 90-95% less money... making the industry re-consider all those AI GPU purchases... and seeing NVIDIA stock dropping 11% in pre-market trading, bleeding out $384 billion. Ouch.

 

 

And China/budget-conscious creators/prosumers want that 24GB framebuffer. 

Leak: RTX 5080 performs worse than RTX 4090

The last-generation flagship is up to 23 percent faster than the RTX 5080.

 

oooh me likey....Nvidia stock go CRASH, AI demand go DOWN, more supply of gpus for US with better prices 😛 (one can dream lulz)

[jaybee83]

On 1/24/2025 at 10:31 PM, jaybee83 said:

 

yes yes and yes on all points.

 

the suprim offers better cooling, especially on the vram, more heatpipes overall and that at 400 USD LESS (= 500€ LESS) than the astral. plus the fact that scalpers will focus on the FE and Astral first and foremost.

 

the two extra power stages on the astral doesnt make up for all that (24 vs. 22 on the suprim for the core, and an equal 7 vs. 7 for the vRAM). wont make any difference for the amount of OC possible anyways...

 

i got my card picked, triggerfinger is ready 😛 

 

btw guys, i just realized:

 

The 24+7 power stages on the Astral are 80A rated, so that would be 24*80=1920A total.

 

The Suprim SOC "only" sports 22 phases, but theyre likely 90A rated (tbc, but likely so far), thus = 22*90=1980A.

 

Thus, even LESSS reason to pay the Astral premium *lulz*

[electrosoft]

2 hours ago, jaybee83 said:

 

btw guys, i just realized:

 

The 24+7 power stages on the Astral are 80A rated, so that would be 24*80=1920A total.

 

The Suprim SOC "only" sports 22 phases, but theyre likely 90A rated (tbc, but likely so far), thus = 22*90=1980A.

 

Thus, even LESSS reason to pay the Astral premium *lulz*

 

Plus I've just had such good luck with MSI overall the last several years from their products to their warranty services.

 

Even this MSI Ventus 3x 4070 Super is so quiet under load and no coil whine (but again, that might be my PSU helping out I realize).

 

Suprim is definitely #1 on my list.

 

EVGA will always be #1 on my list....in my heart 💔

 

30th is right around the corner. Still debating if I want one and if I want to venture out to Microcenter.

[Mr. Fox]

9 hours ago, Papusan said:

 

One thing for sure... The repair shops have good times forwards. Just as NorthridgeFix🙂 Nvidia made a huge business replace melted 12VHPWR connectors. Now they need to clean up LM spills, +  replace mosfets, shunts, caps and inductors as well. 

 

All of the blocks and restrictions the Green Goblin has put in place have done a lot to create opportunities for user-induced damage to video cards. If the norm were that all GPUs we're unlocked like Galax HOF and K|NGP|N cards there would be no shunt mods or volt mods or firmware mods and cross-flashing issues. The average gamerboy would behave exactly like they already do in terms of accessing the BIOS and making adjustments... Paralyzed by ignorance and fear of the unknown, and always in too much of a hurry to play games to be bothered with anything more complicated than a one-click solution. The lockdowns and crippling tactics are what drive the efforts by the  adventurous overclockers to have something awesome that should have been awesome (but wasn't) before they paid for it. Motherboard manufacturers would be financially crucified by enthusiasts and would never get away with the same kind of control freak nonsense, and I don't know why video card manufacturers are allowed to. Everyone keeps opening their wallet for overpriced crippled trash.

[Papusan]

The 5090's prices are coming in here home. GIGABYTE GeForce RTX 5090 AORUS Master ICE cost $50 USD less than the ASUS GeForce RTX 5090 ROG Astral OC. Pick your poisen. But as I have stated before... I prefer Asus cards over Gigabyte. 

 

The Asus 5090 TUF is around $200 cheaper than ASUS GeForce RTX 5090 ROG Astral OC. Incl the 25% Norwegian tax and shipping.

 

And to show prices for the 5090 LC versions. GIGABYTE GeForce RTX 5090 AORUS Xtreme WaterForce cost $44,25 USD more than ASUS GeForce RTX 5090 ROG Astral LC OC. Who would think Gigabyte was more expensive than Asus ROG cards. That's disgusting. (3,366.07 United States Dollar for the Giga LC card) incl the 25% Norwegian tax and shipping. Is Gigabyte gone nuts?

 

 

[jaybee83]

19 hours ago, electrosoft said:

30th is right around the corner. Still debating if I want one and if I want to venture out to Microcenter.

 

while ure at it....why dont u just bring me and the rest of the boys here some 5090s from MC 😄 kthxbyeeeeee

[electrosoft]

As predicted, Nvidia's pricing was a shock to AMD's system.

 

Initial pricing was to be ~$899 for the 9070XT which is INSANE but it looks like they were trying to keep the launch pricing of the 7900XT. They also were going to go for $749 for the non XT even.

 

If I had to guess, I think they thought Nvidia was going to come in $100 higher (or caught wind it might) on the 5070ti vs $50 lower. I honestly don't know what they were thinking with the $749 on the non XT.

 

Someone needs to drag AMD into an alleyway, put them up on the wall and say, "This ain't yer CPU division bub....you ain't Nvidia! Fix your $&*%!"

 

https://videocardz.com/newz/retailer-confirms-powercolor-radeon-rx-9070-xt-red-devil-limited-edition-is-in-stock-talks-amd-pricing-strategy

 

7900xtx prices have dropped and continue to drop. Sapphire Nitro+ = $919. Pulse = $829 because they can't compete with the 5080 anywhere near their former prices and had a rough go even against the 4080 Super.

 

I'm watching and if the Nitro drops below 900 beans I might have to snatch one up.

 

I've been purposely running WoW at 4k Ultra 10 + RT on the 4070 Super the last week to really get a taste of that lower fps life and logging gpu usage to really get an idea of the sweet spot before I return it Friday and a 7900xtx is ~50% more powerful than the 4070 Super with the newest drivers.

 

Obvious a 5090 would immediately introduce the mother of all CPU bottlenecks....

 

 

[jaybee83]

11 minutes ago, electrosoft said:

Obvious a 5090 would immediately introduce the mother of all CPU bottlenecks....

 

see? thats how it goes> get 5090 = CPU bottleneck = get X3D = RAM bottleneck = get faster RAM = SSD botleneck....etc

 

keeps the industry going baby! 😄