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

[electrosoft]

25 minutes ago, Papusan said:

 

And @Falkentyne's "fantastic" chips is SP93 (P: SP95 - E: SP 103). He can barely do an 2% overclock on the P-cores. Sugi's is P88 and E109. So high SP can offer trash bin. A real lottery and all weight is now put on the baby cores. An 9% IPC increase for the P-cores and a loss of +23% with HT just killed 8 core benching on the bot. Nice.

 

Might come down to cooling and skill too. Is Falk's delidded?

 

As always, gear, silicon lottery always in play but what we DO know this early on is 50k is achievable on a non chiller nor DD depending on chip quality.

 

Imagine these CPUs fully opened up delidded, DD, chilled and with continual incoming BIOS refinements.

 

 

 

 

[Papusan]

1 hour ago, electrosoft said:

 

Might come down to cooling and skill too. Is Falk's delidded?

 

As always, gear, silicon lottery always in play but what we DO know this early on is 50k is achievable on a non chiller nor DD depending on chip quality.

 

Imagine these CPUs fully opened up delidded, DD, chilled and with continual incoming BIOS refinements.

 

 

 

 

 

Nope. But on paper he have much better binned P-cores. Sugi does 5.8 delidded while Falk can barely do 5.6 on on P-cores. Most of the 50.000 score is from better E-cores. And Sugis chips have of course better babies. And the chips is binned. Probalbly for ambient temps. And remember Intel traded HT for better power efficiency and better E-cores. The P-cores doesn't do the most work as for 12th, 13th and 14th gen. And I can't undertand why Intel couldn't do more than sub 10% IPC increase for the P-cores.

 

6 hours ago, Mr. Fox said:

It seems like new product releases challenged Intel as much as it did AMD this time around. These guys need to stop counting on fanboys that believe newer is always better. While the compute performance of the 285K is truly amazing, it is underwhelming and underbuilt at everything else. It's 8 threads short of a full load and the 25% reduction in thread count makes it a failure on face value

 

Intel hit their goal without HT. Thanks to faster E-cores. So thei goal was low from the beginning. 

 

“It’s a combination of a couple things, actually,” Hallock told reporters. “First, we knew that we can actually save the wattage for hyperthreading by not including it on the product, and you see that we’re still coming out net ahead by roughly 15, 20 percent in [multicore performance] without it. So we’re able to bump up efficiency and still hit our goals in overall compute performance.

 

“The other thing that I would say is, you know, these are the same designs as leveraged from Lunar Lake,” Hallock added. “We took those cores, those designs, and were able to immediately integrate them because of [Intel’s] Foveros [technology]. So that’s the kind of one-two punch that influenced our decision: speed to market and maximizing performance per watt.”

 

https://www.pcworld.com/article/2480487/hyperthreading-is-dead-in-intels-new-core-ultra-pc-chips.html

 

See bolded text in my post below...

 

6 hours ago, Papusan said:

 

I expect we had seen 285K chips eat 400W with HT. So the power efficiency they have talked so much about would go down the drain. Reason for removing HT is for the mobile junk. And we all know Intel will use same mess for coming gaming Jokebooks. And not so sure TSMC silicon can handle 400W long-term. 

 

[Talon]

1 hour ago, electrosoft said:

Looking at Sugi's numbers, 285k is hitting 50k in CB23 already basically out of the block with some tweaks. Delidded but not even on DD yet and using a Mora 420 not even a chiller. Not to shabby! That single core score is pretty beastly too!

 

 

Unfortunately it's gaming prowess is going to be a game to game thing and things are not looking good atm overall.

 

For me, I think I'm going to sit back a smidge (same as I did for 13th and 14th) and watch and wait before I do anything.....why?

 

I spent a week+ swapping back and forth between my 7950X3D rig and 14900KS rig playing a crap ton of WoW and the 14900KS @ 5.9 all core, 8200 tuned w/ APO was just superior in every way especially with the lows and smooth game play. Any and all occasional chunk I would hit (I'll begrudgingly admit it might be a bit of "dip") with the 7950X3D was just gone with the end game 14900KS setup. Having logged 30hrs+ on each setup running the same raids/questing/PvP, the 14900KS once tuned across the board is the superior answer.

 

I would swap my 4090 back and forth with a clean install of 24H2 running the same raids and TWW outdoor content and the fluidity and game play was just superior on the 14900KS. That is why I sold off the 7950X3D and X670E Carbon. I can't envision the 9800X3D or 9950X3D providing this as smooth of a gaming experience in WoW but I'm always open to change. 🙂

 

I am to a point now, playing WoW, where I say back last night and just went, "I don't see why I need a 5090 or a CPU upgrade," and that has never happened in tandem. If I do buy a 5090, it is just pure leisurely fun.

 

After I finish selling off the rest of some of this gear (we're in the homestretch!), if I pick up a 285 setup, again, it will be for pure tweaking and optimizing fun. If somehow it can bring the pain for WoW,  I would as mentioned before just swap it into the wife's system and run the 285 as my main.

 

 

I agree. Seeing Sugi's numbers this early on with what is a good CPU (His SP=91, for example Debauer was using an SP=75), I think this is going to be a tweakers paradise chip especially with DLVR disabled and switching more to a normal approach for overclocking. Memory is coming out of the gate hitting nicely too. It is Rocketlake 2.0 in what I feel is an ok way.

 

 

 

 

I'm actually having a blast juicing this chip. And I have the baby 265K for now. MC had ZERO chips this morning for 285K. I snagged a Newegg and it was packaging, but I decided to wait. I'm not sure what I'm going to do. So far I'm not even mad at the 265K and I know if I get similar 285K silicon to this beast, I might beat out my 14900KS. 

 

I just ordered that G.Skill 9600 CUDIMMs on Newegg. They became available and I got an order in, although it's a preorder for Nov 1. 

 

https://browser.geekbench.com/v6/cpu/8473838 -- Beat my 14900KS best result. 

 

I'm running 57x to 55x on the P cores, and 51.33x on all E cores. The E cores are absolute units on these chips. 

 

 

 

 

 

 

[Talon]

Forgot the Cinebench run at same clocks. 

[electrosoft]

1 hour ago, Papusan said:

 

Nope. But on paper he have much better binned P-cores. Sugi does 5.8 delidded while Falk can barely do 5.6 on on P-cores. Most of the 50.000 score is from better E-cores. And Sugis chips have of course better babies. And the chips is binned. Probalbly for ambient temps. And remember Intel traded HT for better power efficiency and better E-cores. The P-cores doesn't do the most work as for 12th, 13th and 14th gen. And I can't undertand why Intel couldn't do more than sub 10% IPC increase for the P-cores.

 

 

It doesn't really matter in the quest for Multithreaded scores/benchmarking. A score is a score is a score. If 1 P-core and 23 E-cores swoop in and take the record of fastest CPU for multithreaded consumer level workloads, so be it. 🙂

 

Falk also uses an AIO. So AIO, not delidded vs Mora + delid. Bin will only get you so far depending on what you seek.

 

slightly sub ~10% uplift on P-cores is nothing to scoff at.

 

I personally think eliminating HT will pay dividends long term but that remains to be seen.

 

I do agree E-cores got the majority of the love this time around vs P-cores.

 

It is a tinkerer's chip IMHO. Again, Rocketlake 2.0. For those questing for monster multi scores, there is a lot to like with it and room to grow. For gamers?  I'm cautiously optimistic considering how 14th with refinements and APO was able to pull ahead of X3D quite comfortably in lows and smoothness of play.

 

4 minutes ago, Talon said:

 

I'm actually having a blast juicing this chip. And I have the baby 265K for now. MC had ZERO chips this morning for 285K. I snagged a Newegg and it was packaging, but I decided to wait. I'm not sure what I'm going to do. So far I'm not even mad at the 265K and I know if I get similar 285K silicon to this beast, I might beat out my 14900KS. 

 

I just ordered that G.Skill 9600 CUDIMMs on Newegg. They became available and I got an order in, although it's a preorder for Nov 1. 

 

https://browser.geekbench.com/v6/cpu/8473838 -- Beat my 14900KS best result. 

 

I'm running 57x to 55x on the P cores, and 51.33x on all E cores. The E cores are absolute units on these chips. 

 

 

 

 

 

 

 

Now we're starting to see some good stuff.

 

This is why I didn't just dump all over it and call it a wrap. I wanted to and will continue to watch the OC'ers start to learn their way around it and start extracting real results as BIOS/Windows updates/MC/Drivers slowly work their magic on what is a brand new chip architecture. It can only go up from here.

 

I hope I don't pull an 11th and talk about how I'm gonna wait and see then suddenly a few days later show up on this thread like over on NBR with an MSI z590 board and a shiny new 11700k to play with. 🤣

 

 

 

[Papusan]

13 hours ago, electrosoft said:

Falk also uses an AIO. So AIO, not delidded vs Mora + delid. Bin will only get you so far depending on what you seek.

 

slightly sub ~10% uplift on P-cores is nothing to scoff at.

 

It seems we now have come to the days you can only extract 200MHz on flagship processors. Exactly as in the old days with Haswell low end mobile cpu's where you got 1 or 2 bin extra. And this on so called new gen pover efficient chips that can run colder than before.  

 

Regarding 10% IPC uplift nothing to scoff at. Nice. The P-cores will perform less than previous gen with HT fused into the chips for certain multi threaded tasks. Give the p-cores 10% more gas then take back +23% of the performance with the other hand. Yup.... it is about give and take in life😁

 

 

Summary and Conclusion igorslab.de

Now it gets difficult if you want to remain objective. On the one hand, Intel has created three really interesting and efficient CPUs, but on the other hand, there are serious problems with performance in quite a few applications, especially those with fewer threads and, of course, especially in the all-important gaming. The fact that the new flagship can only compete with an older Core i5-14600K or a cheap Ryzen 5 9600X in some cases is almost depressing.

 

Quote

Why you should buy Core Ultra processors? Higher numbers the better.... Paying premium for gear 4 for gaming is amusing. How many gamers will use Gear 4 when they see high numbers? Remember they want value for they money. You just can't underclock your expensive hardware. Or can or/and will you?

 

Whats the point with this below? Will gear 4 help Intel to improve gaming? And will the gamers now switch from gear 2 to gear 4 because the higher ram speed? Because higher numbers is always better. 

 

Corsair teases DDR5-10000 CUDIMMs ahead of early November release

 

There are two major catches with DDR5-10000 CUDIMMs. Firstly, this level requires Gear 4 mode. 

[Mr. Fox]

44 minutes ago, Papusan said:

 

It seems we now have come to the days you can only extract 200MHz on flagship processors. Exactly as in the old days with Haswell low end mobile cpu's where you got 1 or 2 bin extra. And this on so called new gen pover efficient chips that can run colder than before.  

 

Regarding 10% IPC uplift nothing to scoff at. Nice. The P-cores will perform less than previous gen with HT fused into the chips for certain multi threaded tasks. Give the p-cores 10% more gas then take back +23% of the performance with the other hand. Yup.... it is about give and take in life😁

 

 

Whats the point with this below? Will gear 4 help Intel to improve gaming? And will the gamers now switch from gear 2 to gear 4 because the higher ram speed? Because higher numbers is always better. 

 

Corsair teases DDR5-10000 CUDIMMs ahead of early November release

 

There are two major catches with DDR5-10000 CUDIMMs. Firstly, this level requires Gear 4 mode. 

The computer technology space is so overrun by idiots in positions involving decision-making that it is a wonder things are not more screwed up than they already are. It's not going to get better. They are not getting any smarter. In fact, quite the opposite is true... dumb and dumber, with more dumbness on the way. 

[electrosoft]

Hmmmm, surprised to see the 285k top even the X3D in Starfield. X3D is a monster in Fallout 76 and Starfield uses the much improved engine version used in FO76. I'm very curious to see how the 285k performs in FO76 since the 4090 is perpetually bottlenecked by the CPU in there and the 7950X3D dominates in that game.

 

[tps3443]

15 hours ago, electrosoft said:

 

It doesn't really matter in the quest for Multithreaded scores/benchmarking. A score is a score is a score. If 1 P-core and 23 E-cores swoop in and take the record of fastest CPU for multithreaded consumer level workloads, so be it. 🙂

 

 

 

 

After my recent experience this week I think it matters how the chip is made up, the first thing I noticed is system fluidity on my 3175X. Zero delays, zero skips when working, just brute multitasking power like nothing is happening. I faintly remember this on my 7980XE@5.0Ghz and DDR4 4000, and I remember it being able to play two games at once 100% smooth, one in VR, and one on flatscreen. But I have been using the big/little chips for a few years now, so I have gotten use to the quirks, and it was just a normal thing. But I always thought the P/E chips while absolutely brutal in multithreaded, get a little lost in translation sometimes when the load gets high, and the apps multiply, and the windows services start stacking up. Use the OS for a few hours for work, close things out, and these high R23 scores have depleted, only a restart will replenish this. I can imagine this sounds hilarious, but it was the first thing I noticed on this absolute dinosaur of a system. I had to mention how smooth a true high core count CPU is. I can only assume the 9950X and Threadripper chips are like this as well 🧐

 

This makes me want to transition to W790 or the new W890 platform in the future and use only Monolithic CPU’s with only P-Cores. Because it is a big difference in smoothness. When we forget about what CPU we’re using, and we’re just working and doing things on the desktop running virtual machines, maybe minimizing games as well, the smoothness is definitely there on high core count Monolithic CPU’s that only have P-Cores available. 
 

After this experience, I truly wish Intel would go back to all P-Cores and monolithic chips. If they released a 10/20 chip with some crazy good IPC and massive cache size I would absolutely own one. 

[electrosoft]

19 minutes ago, tps3443 said:

After my recent experience this week I think it matters how the chip is made up, the first thing I noticed is system fluidity on my 3175X. Zero delays, zero skips when working, just brute multitasking power like nothing is happening. I faintly remember this on my 7980XE@5.0Ghz and DDR4 4000, and I remember it being able to play two games at once 100% smooth, one in VR, and one on flatscreen. But I have been using the big/little chips for a few years now, so I have gotten use to the quirks, and it was just a normal thing. But I always thought the P/E chips while absolutely brutal in multithreaded, get a little lost in translation sometimes when the load gets high, and the apps multiply, and the windows services start stacking up. Use the OS for a few hours for work, close things out, and these high R23 scores have depleted, only a restart will replenish this. I can imagine this sounds hilarious, but it was the first thing I noticed on this absolute dinosaur of a system. I had to mention how smooth a true high core count CPU is. I can only assume the 9950X and Threadripper chips are like this as well 🧐

 

This makes me want to transition to W790 or the new W890 platform in the future and use only Monolithic CPU’s with only P-Cores. Because it is a big difference in smoothness. When we forget about what CPU we’re using, and we’re just working and doing things on the desktop running virtual machines, maybe minimizing games as well, the smoothness is definitely there on high core count Monolithic CPU’s that only have P-Cores available. 
 

After this experience, I truly wish Intel would go back to all P-Cores and monolithic chips. If they released a 10/20 chip with some crazy good IPC and massive cache size I would absolutely own one. 

 

This is a strawman or at the very least moving the goalposts. 🙂

 

I'm talking about brute power overall.

 

You're talking about latency, responsiveness, smoothness and OS snappiness.

 

With that out of the way, you aren't the first to note these new tile/stacked and/or hybrid designs seem to inject a bit of hesitancy  and responsiveness degradation into D2D operations vs older monolithic designs. I will say when I've had to use older laptops or desktops, the desktop itself seems to be a bit snappier and smoother even if the chip itself is much weaker versus newer designs in computational performance.

 

I prefer monolithic designs too but I understand limitations and embrace change and try to look at the big picture.

 

AMD and ARM embraced a chiplet/tiled approach because of limitations. AMD even extended it across their GPUs. Intel was up against a wall with limitations. Ultra 200 is the first step in overcoming those limitations and they are just going to get better from here on out. Even Nvidia is rumored to be eventually going chiplet based. It is the future of computing and it is what it is. All we can do is hope refinements allow us to have our cake and eat it too.

 

 

[Mr. Fox]

8 minutes ago, electrosoft said:

 

This is a strawman or at the very least moving the goalposts. 🙂

 

I'm talking about brute power overall.

 

You're talking about latency, responsiveness, smoothness and OS snappiness.

 

With that out of the way, you aren't the first to note these new tile/stacked and/or hybrid designs seem to inject a bit of hesitancy  and responsiveness degradation into D2D operations vs older monolithic designs. I will say when I've had to use older laptops or desktops, the desktop itself seems to be a bit snappier and smoother even if the chip itself is much weaker versus newer designs in computational performance.

 

I prefer monolithic designs too but I understand limitations and embrace change and try to look at the big picture.

 

AMD and ARM embraced a chiplet/tiled approach because of limitations. AMD even extended it across their GPUs. Intel was up against a wall with limitations. Ultra 200 is the first step in overcoming those limitations and they are just going to get better from here on out. Even Nvidia is rumored to be eventually going chiplet based. It is the future of computing and it is what it is. All we can do is hope refinements allow us to have our cake and eat it too.

 

 

The sad part is some of the limitations are unnecessary and self-imposed based on preconceived ideas and the potential that ideas and initiatives that fail will be harmful to the careers of key empire-building members of one or more self-anointed brain trusts.

This is what made turdbooks objects of contempt and worthy of ridicule. A segment of consumers and industry decision-makers decided that the rest of the world needed to submit and fall in line with the notion that thinner and lighter and long battery life are the only things that matter, to the exclusion of everything else. No other options are necessary, and the needs, preferences and desires of those who differ are irrelevant.

[Papusan]

58 minutes ago, Mr. Fox said:

The sad part is some of the limitations are unnecessary and self-imposed

 

Why make it so complicated/difficult for themself? And why force all new modern tech from Jokebook hardware into desktop hardware?

The Core Ultra 200S are the worst launch of 2024 and in Intel's history, not only because of performance, but because of its general problems

 

The concept of AutoGV, or CCF Auto GV as some manufacturers call it, was introduced in Meteor Lake because it has a Tiles structure similar, if not identical, to Arrow Lake, or vice versa.

 

 

[Mr. Fox]

1 hour ago, Papusan said:

Why make it so complicated/difficult for themself? And why force all new modern tech from Jokebook hardware into desktop hardware?

Bad leaders erroneously equate the implementation of change as being a measurement of success. Good and effective leaders recognize that blocking change for the sake of change is one of the reasons their job exists and they should only be an agent of change when the virtues are easily quantified, the need for it is validated and the reasons communicated clearly to everyone affected. Before implementing change, they also proactively identify any resistance to change, embrace the expressions of resistance and seek to understand them, and pursue solutions whenever impasse is encountered. The stupidity problem originates at the highest levels of leadership and the metastasis begins at the head.

[Papusan]

7 hours ago, Papusan said:

Regarding 10% IPC uplift nothing to scoff at. Nice. The P-cores will perform less than previous gen with HT fused into the chips for certain multi threaded tasks. Give the p-cores 10% more gas then take back +23% of the performance with the other hand. Yup.... it is about give and take in life😁

 

Here's more correct numbers. You lose 18% performance bro @Mr. Fox Intel give you 9% better peerformance then with the other hand take back the double. Great deal? Intel followed Apple to try make an more efficcient processors removing HT. They even follow Apple with phone cores. If AMD can give you an 15% IPC increase on P-cores then you should think Intel was able to do the same. But nope. You get faster baby cores as replacement for better P-cores. 

 

 

Intel Core Ultra 200S in the test: computerbase.de

 

The fact that even the strongest Core Ultra 200S only delivers the performance of a Core i5-14600K in games out of the box is actually surprisingly weak. On average, the three K processors are always almost double-digit percentages behind their respective predecessors. 

 

 

[Mr. Fox]

28 minutes ago, Papusan said:

 

Here's more correct numbers. You lose 18% performance bro @Mr. Fox Intel give you 9% better peerformance then with the other hand take back the double. Great deal? Intel followed Apple to try make an more efficcient processors removing HT. They even follow Apple with phone cores. If AMD can give you an 15% IPC increase on P-cores then you should think Intel was able to do the same. But nope. You get faster baby cores as replacement for better P-cores. 

 

 

Intel Core Ultra 200S in the test: computerbase.de

 

The fact that even the strongest Core Ultra 200S only delivers the performance of a Core i5-14600K in games out of the box is actually surprisingly weak. On average, the three K processors are always almost double-digit percentages behind their respective predecessors. 

 

 

Saying that the 200-series P cores perform better than 14900K and Ryzen when HT/SMT is disabled is specious and totally ignores the reality that those CPUs include a valuable feature that Arrow Lake does not. Most users do not disable features that enhance performance, so the comparison is illogical.

 

Here is a metaphorical example of similar logic:

 

If you disconnect two of the spark plug wires on your V8 motor my V6 will beat it in a drag race. Plus, my V6 gets better gas mileage, so that makes my car faster and more efficient than your car.

 

 

 

[Papusan]

2 hours ago, Mr. Fox said:

Saying that the 200-series P cores perform better than 14900K and Ryzen when HT/SMT is disabled is specious and totally ignores the reality that those CPUs include a valuable feature that Arrow Lake does not. Most users do not disable features that enhance performance, so the comparison is illogical.

 

Here is a metaphorical example of similar logic:

 

If you disconnect two of the spark plug wires on your V8 motor my V6 will beat it in a drag race. Plus, my V6 gets better gas mileage, so that makes my car faster and more efficient than your car.

 

 

 

 

There is more.... At 4.0GHz for Raptor and Arrow Lake the difference is 18% better performance with 8 P-cores. The same with 5.7GHz. But most Core Ultra will top out at 5.7GHz on the P-cores even if you disable the baby cores. With 14900K/KS you can go on further and easly do 6.0GHz. 6.0 / 5.7 = 5%

 

So you can extract 5% better performance from higher clock speed onolder Raptor Lake (Refresh). So it match very well the 23% performance gain I posted above. 8 cores bench fun with Intel is now dead and buried for hwbot. To do that you'll need older Intel chips. Is newer always better bro @ryan ?

 

And I wonder how happy the Jokebook jockey's will  be when they get their hands on Arrow Lake Mobile. They will probably kiss Pats toes when they see/or can feel better battery runtime in games with their brand new gaming Jokebooks. 

 

Pay more get less.

 

 

Fantastic. The well known former overclocker 8-Pack managed to overclock the P-cores from 5.4 to 5.5GHz with an AIO. Yup, that's fantastic 1.8% overclock. And 5.6GHz with an custom water loop. Not bad😁

 

 

And can anyone try explain why Asus engineers had to do this stupid changes for this board? ASUS Z890 Hero motherboard feeds four of the CPU VRM phases from the 24-pin ATX connector, instead of all delivering all needed power from the 8-pin EPS connectors. What's the point with this idiotic MB design?  Did they smoke something while the  engineered this board? @Mr. Fox what the hell going on?

[Mr. Fox]

1 hour ago, Papusan said:

 

There is more.... At 4.0GHz for Raptor and Arrow Lake the difference is 18% better performance with 8 P-cores. The same with 5.7GHz. But most Core Ultra will top out at 5.7GHz on the P-cores even if you disable the baby cores. With 14900K/KS you can go on further and easly do 6.0GHz. 6.0 / 5.7 = So you can extract 5% more performance. So it match  the 23% performance gain I posted above. 8 cores bench fun with Intel is now dead and buried for hwbot. To do that you'll need older Intel chips. Is newer always better bro @ryan ?

 

And I wonder how happy the Jokebook jockey's will  be when they get their hands on Arrow Lake Mobile. They will probably kiss Pats toes when they see or canfeel better battery runtime in games with their brand new gaming Jokebooks. 

 

Pay more get less.

Fantastic. The well known former overclocker 8-Pack managed to overclock the P-cores from 5.4 to 5.5GHz with an AIO. Yup, that's fantastic 1.8% overclock. And 5.6GHz with an custom water loop. Not bad😁

And can anyone try explain why Asus engineers had to do this stupid changes for this board? ASUS Z890 Hero motherboard feeds four of the CPU VRM phases from the 24-pin ATX connector, instead of all delivering all needed power from the 8-pin EPS connectors. What's the point with this idiotic MB design? 

ASSzeus is becoming famous for being stupid. They do many things wrong and many things that make little sense. The only Z890 board I would buy if I were buying one is the ASRock Taichi OC Formula. It is also comical that so many companies are making motherboards that are grossly overbuilt compared to Z790 for CPUs that are crippled and incapable of taking advantage of the power delivery design. I guess they think they will come out ahead by offering VRMs and phases that would be needed for a CPU that pulls two or three times the power that Arrow Lake is capable of. A wasteful excuse that justifies charging inflated prices for stuff not utilized.

[Papusan]

 

LOOL. Core Ultra 9 285K aka Arrow lake feels more like an mobil part. Spot on😆 From 16:29

 

8 hours ago, electrosoft said:

Hmmmm, surprised to see the 285k top even the X3D in Starfield. X3D is a monster in Fallout 76 and Starfield uses the much improved engine version used in FO76. I'm very curious to see how the 285k performs in FO76 since the 4090 is perpetually bottlenecked by the CPU in there and the 7950X3D dominates in that game.

 

 

Starfield, uses as many cores as possible, even 12 or even 24 if necessary. No wonder you see such results. I'm sure the new faster babies help boost the performance for this game. Again saved by the babies, LOOL

[Mr. Fox]

1 hour ago, Papusan said:

Starfield, uses as many cores as possible, even 12 or even 24 if necessary. No wonder you see such results. I'm sure the new faster babies help boost the performance for this game. Again saved by the babies, LOOL

I almost think the CPU would have been better with 32 Atom cores on a monolithic die and no P-cores since you only get 8 with no HT and they have no overclocking abilities to speak of and the baby smartphone cores overclock better. Then there wouldn't be any glued together tile  chiplet crap. Huge mistake for them to involve themselves with TSMC as well. 

[Papusan]

15 minutes ago, Mr. Fox said:

I almost think the CPU would have been better with 32 Atom cores on a monolithic die and no P-cores since you only get 8 with no HT and they have no overclocking abilities to speak of and the baby smartphone cores overclock better. Then there wouldn't be any glued together tile  chiplet crap. Huge mistake for them to involve themselves with TSMC as well. 

 

[tps3443]

This thing throws really well. 
 

Great bandwidth and good latency. 
 

 

[Mr. Fox]

10 hours ago, tps3443 said:

This thing throws really well. 
 

Great bandwidth and good latency. 
 

 

That is really good.

 

You need to buy an AIDA64 key, bro. You get updates for however long you purchase support but the key lasts forever and it works forever after the period of support for included updates expires. Updates don't matter unless you purchase hardware manufactured after the version release date. And, unless something has changed you can manually download a newer zip version and extract it over an old version with the support period expired and not lose your license. Just no automatic updates after the support period ends.

 

I always only use the zipped version. It's totally portable. I do not use the installer version of any software like this (AIDA64, HWiNFO64, CPU-Z, GPU-Z, etc.). Portable apps are always best. And, most portable apps are buy once (if not totally free) and use it forever.

 

https://www.aida64.com/downloads/MGY1ZmE4MGU=

[tps3443]

6 hours ago, Mr. Fox said:

That is really good.

 

You need to buy an AIDA64 key, bro. You get updates for however long you purchase support but the key lasts forever and it works forever after the period of support for included updates expires. Updates don't matter unless you purchase hardware manufactured after the version release date. And, unless something has changed you can manually download a newer zip version and extract it over an old version with the support period expired and not lose your license. Just no automatic updates after the support period ends.

 

I always only use the zipped version. It's totally portable. I do not use the installer version of any software like this (AIDA64, HWiNFO64, CPU-Z, GPU-Z, etc.). Portable apps are always best. And, most portable apps are buy once (if not totally free) and use it forever.

 

https://www.aida64.com/downloads/MGY1ZmE4MGU=

My Apex had a free Aida copy, I lost it after hooking up the new motherboard/cpu. I did buy a key before the Apex from a janky site, but it was suspended for some reason. I’ll buy one though. 

 

The sub timings and main timings on LGA3647 can be reduced to some crazy low numbers lol. I mean, I could not do this on Z490/Z590 or even X299. I’m running CL13 currently for DDR4 3800, but I can probably bench CL12 pretty easily with enough voltage lol.  Considering it’s 96GB of Dual rank, it’s kind of amazing. I can’t imagine what can be done if my ram was water chilled like my cpu. 
 

I’m currently running 3800 cl13-14-14-34-260. 

[electrosoft]

16 hours ago, Papusan said:

Starfield, uses as many cores as possible, even 12 or even 24 if necessary. No wonder you see such results. I'm sure the new faster babies help boost the performance for this game. Again saved by the babies, LOOL

 

You say this like it's a bad thing lol.......

 

 

[tps3443]

House update! It’s almost DONE!!! Only one thing left is final inspection, and those steps are not mine, they have to build the real ones which are going to be 5ftx5ft landing. The steps on the home now are just temporary/portable wood steps.

 

Pics of Inside the home, I do not have a key yet lol. They’ll change the locks before they hand us new keys. 
 

The cabinets are all real hardwood throughout the entire home. 

 

upload pic