Discussion Raptor Lake Build Thread

[Carfax83]

New MSI BIOS out that allows my MB to run my DDR5 7200 at full speed. Appreciate it MSI engineers!

Post some Aida64 scores!

 

[Rigg]

CPU: Intel Core i7-13700K 3.4 GHz 16-Core Processor
CPU Cooler: Scythe Fuma 2 Rev.B 39.44 CFM CPU Cooler
Motherboard: Asus TUF GAMING Z690-PLUS WIFI ATX LGA1700 Motherboard
Memory: G.Skill Ripjaws S5 32 GB (2 x 16 GB) DDR5-6000 CL30 Memory
Storage: Samsung 980 Pro 2 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive
Video Card: Asus TUF GAMING OC Radeon RX 6900 XT 16 GB Video Card
Case: Fractal Design Meshify C ATX Mid Tower Case
Power Supply: Fractal Design Ion+ 760P 760 W 80+ Platinum Certified Fully Modular
Case Fan: be quiet! Silent Wings 3 73.33 CFM 120 mm Fan
Case Fan: be quiet! Silent Wings 3 73.33 CFM 120 mm Fan
Case Fan: Noctua S12A PWM chromax.black.swap 63.27 CFM 120 mm Fan
Case Fan: Noctua S12A PWM chromax.black.swap 63.27 CFM 120 mm Fan
Case Fan: Noctua F12 PWM chromax.black.swap 54.97 CFM 120 mm Fan
Case Fan: Noctua A12x15 PWM chromax.black.swap 55.44 CFM 120 mm Fan
Custom: Thermalright CPU Contact Frame for LGA 1700

Stock 6400 MT/s (EXPO timings):



Tune 6400 MT/s (EXPO timings), -75 mV offset (L3 LLC), 85C temp limit, MCE on, +2 TVB profile (voltage optimizations disabled)

 

[Rigg]

Core utilization OC 55/55/54/54/54/54/54/54. with -50 mV offset (L3 LLC), 85C temp limit, MCE on, +2 TVB profile (voltage optimizations disabled)

 

[Carfax83]

@Rigg, does your motherboard tell you the SP rating for your CPU? Seems like a good chip. I would recommend turning off MCE though.

 

[Rigg]

@Rigg, does your motherboard tell you the SP rating for your CPU? Seems like a good chip. I would recommend turning off MCE though.

I don't see SP rating anywhere. The BIOS seems very similar to the ROG UEFI I'm used to. Where is it located in yours? Buried somewhere in a sub menu?

MCE doesn't really do anything with the temp limit on. I'm just using the temp limit to limit power instead of PL1/PL2. It's on by default so I just left it alone.

*EDIT
I don't think its a particularly good piece of silicon. The TVB +2 in combination with less offset voltage and lowish LLC yields constant 200mhz boost when I'm below my temp limit. I'm basically trading optimal heavy multi core for 2-300 MHZ boost in everything else. With more voltage and cooling this concept could be taken further. I can get up to 5.8 single core to hold (slightly below 1.4v under load) with this method but it absolutely tanks multi core performance. The vcore is too high to overcome the power limit need for thermals to be in check.

 

[Carfax83]

I don't see SP rating anywhere. The BIOS seems very similar to the ROG UEFI I'm used to. Where is it located in yours? Buried somewhere in a sub menu?

It's in the all the sections on the bottom right hand side, under predictions. Mine is 109, which is apparently a really good score. I think it's accurate because I've been able to undervolt my CPU down to -135mV with perfect stability at 5.3ghz on the P cores and 4.3ghz on the E cores.

MCE doesn't really do anything with the temp limit on. I'm just using the temp limit to limit power instead of PL1/PL2. It's on by default so I just left it alone.

Interesting, I may have to try that method one day. I keep my PL1 and PL2 at 235w, which is enough to hit and sustain the operation of the CPU @5.3ghz/4.3ghz under heavy loads with low temps for an air cooled setup.

I don't think its a particularly good piece of silicon. The TVB +2 in combination with less offset voltage and lowish LLC yields constant 200mhz boost when I'm below my temp limit. I'm basically trading optimal heavy multi core for 2-300 MHZ boost in everything else. With more voltage and cooling this concept could be taken further. I can get up to 5.8 single core to hold (slightly below 1.4v under load) with this method but it absolutely tanks multi core performance. The vcore is too high to overcome the power limit need for thermals to be in check.

I guess it depends on what kind of workloads you're primarily doing. For me, all core clocks is way more important than anything else because I use my machine mostly for gaming and encoding. Honestly, Raptor Lake to me is already blisteringly fast at 5.3ghz that everything is really snappy so I see no reason to boost single core performance even more; especially if its at the expense of the all core performance.

 

[Rigg]

It's in the all the sections on the bottom right hand side, under predictions. Mine is 109, which is apparently a really good score. I think it's accurate because I've been able to undervolt my CPU down to -135mV with perfect stability at 5.3ghz on the P cores and 4.3ghz on the E cores.

I'll take a look later. Maybe I've been too focused on the settings I'm looking at and didn't notice it down in the corner.

I'm stress testing the above OC with some adjustments for 5.8 SC. Mine is stable at 52/42 with -90mv with an 85 C limit and no boost shenanigans. I need 90's to hold 53/43 stable under load with that offset. I can't do much better than 30500 in CB23 (10 min) using the arbitrary 85c limit I like. I'm really not sacrificing much MT or power efficiency at this thermal limit. I think its worth the trade off to get steady 5.6 in games and 5.8 ST.

Interesting, I may have to try that method one day. I keep my PL1 and PL2 at 235w, which is enough to hit and sustain the operation of the CPU @5.3ghz/4.3ghz under heavy loads with low temps for an air cooled setup.

Works for me. I don't care how much power it uses as long as its not past the limits of my cooling. Either way works. I experimented with both. I thought the temp limit was easier since that's ultimately what I'm trying to keep stable anyway.

I guess it depends on what kind of workloads you're primarily doing. For me, all core clocks is way more important than anything else because I use my machine mostly for gaming and encoding. Honestly, Raptor Lake to me is already blisteringly fast at 5.3ghz that everything is really snappy so I see no reason to boost single core performance even more; especially if its at the expense of the all core performance.

I hear ya. I'm trying to optimize for ST/gaming mostly. As long as multi-thread is comparable to stock I'm good. It's doesn't look like I can cool it well enough for it to make much difference in MT anyway.

Turning on TVB +2 and disabling TVB power optimization will give you an extra 200 Mhz boost in lightly threaded stuff/games with little or no trade off. I'd play around with it if I were you. I only needed an extra 15 mV to keep it stable. Real bench will tell you in hurry if it isn't stable. 15 minutes of that with 1/2 of your total system RAM is a a much better quick real world test than 10 min CBR23 in my experience. Once i've settled on an OC 8 hrs of Real Bench overnight is a really good stress test.

 

[Rigg]

Core utilization OC 58/58/56/56/56/56/56/56
TVB set to downclock -200 Mhz when 8 cores are loaded and 75c/85c temp limits A/B are met.
-50 mV global auto voltage offset (L3 LLC), V/F offset #11 +50 mV #10 + 40 mV #9 + 40 mV #7 - 25 mV #6 -25 mV
85C global temp limit with MCE on





This was a complete pain to get stable. You need to use offset voltage to effectively throttle the CPU under heavy load. The problem is the CPU is extremely overvolted out of the box. When you increase clocks the auto voltage increases too. That's fine and all but you cant get the voltage down under heavy load with the reduced offset needed to keep the higher clocks stable. The CPU is super easy to get stable in single core and lighter multi-thread workloads with reasonable voltage at 56/58. I obviously can't cool it at 5.6 all core in heavy workloads with a dual 120 air cooler. Just like at stock it wants to hammer up against the thermal limit at 1.3+ V only trying to boost to 5.6 instead of 5.3. The result is terrible performance for heavy loads.

After a bunch of trial and error I found that with a global -50 mV offset I could then add/subtract at different points in V/F curve to get stability and closer to optimum voltages along the frequency curve. The V/F curve adjustments are wonky/imprecise and I had much better luck using them with the global offset and low LLC as apposed to leaving voltage on auto (with higher LLC) and using V/F offsets exclusively. YMMV.

The key was manually setting the 1-200 Mhz down clock @ 75c/85c when all 8 cores are loaded. Telling the CPU to go 5.4 at 75c under heavy all core load in combination with offsets (and some V droop) nets much lower voltage and higher boost when running up against the 85c thermal limit.

 

[Carfax83]

I hear ya. I'm trying to optimize for ST/gaming mostly. As long as multi-thread is comparable to stock I'm good. It's doesn't look like I can cool it well enough for it to make much difference in MT anyway.

I'm genuinely curious as to what kind of games you're playing that emphasizes ST performance to such a degree.

Turning on TVB +2 and disabling TVB power optimization will give you an extra 200 Mhz boost in lightly threaded stuff/games with little or no trade off. I'd play around with it if I were you. I only needed an extra 15 mV to keep it stable. Real bench will tell you in hurry if it isn't stable. 15 minutes of that with 1/2 of your total system RAM is a a much better quick real world test than 10 min CBR23 in my experience. Once i've settled on an OC 8 hrs of Real Bench overnight is a really good stress test.

My monitor is 4K so I'm effectively GPU limited anyway. That was a major reason behind why I decided to undervolt and underclock my CPU, because I knew that I would be GPU limited in the vast majority of the games I play. Not all of them though.

Witcher 3 Next gen is an abomination in terms of its CPU performance because it uses a DX12 wrapper instead of a native DX12 renderer, so it effectively only uses two threads for rendering. This is one of the rare instances where having DLSS 3 makes a huge difference in performance, practically doubling my FPS.

 

[Sulaco]

Finally got the last few pieces delivered and ready to build! This will be almost entirely a gaming PC. Coming from an 8700K and a 1080. Looking forward to the results.

 

[Rigg]

I'm genuinely curious as to what kind of games you're playing that emphasizes ST performance to such a degree.

Honestly, I just enjoy tuning/overclocking new hardware. Not sure I'll see any practical benefit from the effort. I'd be just fine with a temp (power) limit and a global undervolt much like you are running yours. I'll probably will end up daily driving with a less aggressive overclock. I'm just having fun seeing how much can be squeezed out of it on air. It could just be confirmation bias, but it seems noticeably snappier boosting up to 5.8 when just using the machine for day to day stuff.

This is my living room PC. I run at 1440p 120 with freesync on my 3 year old Samsung QLED 82". I can't even really distinguish a significant difference at 4k from my view distance and the TV won't do over 60hz in 4k. I usually cap at 120 to save energy if I don't need all of the horsepower in a particular game. I don't even play shooters and competitive multiplayer stuff anymore where al that FPS is needed. Extra gaming performance is mostly just an excuse for me to justify my OCD CPU tuning.....LOL. I'll probably tune the hell out of the RAM at some point because reasons.

 

[Carfax83]

I'll probably tune the hell out of the RAM at some point because reasons.

If you have any good tips about RAM tuning, let me know. Memory tuning can become extremely technical if you want to mess around with secondary and tertiary timings. Most people (including myself) just mess with primary timings as that's a lot easier.

But if you can tune secondary and tertiary timings, then you can really increase performance substantially. In fact, I would say that properly tuned DDR5 6600 would probably be faster than XMP DDR5 7600 in most scenarios.

 

[igor_kavinski]

I got the Panzer EVO case, mainly coz it has carry handles. I see now that it includes an RGB controller for the four included case fans. Anyone have experience with it? Anything I should know before trying to build my Z790/12700K in it? I will not be putting in a GPU for the time being. At least, don't plan to but let's see.

PANZER EVO RGB - The RGB Crystalline Titan - COUGAR

PANZER EVO RGB The RGB Crystalline Titan Panzer EVO RGB is the perfect combination of four massive tempered glass panels and four powerful ARGB fans in an exceptional case. With the included fan controller, you will be able to control all the power of each Fan’s 18 LEDs and enjoy more than 100...

cougargaming.com

 

[Harry_Wild]

Intel 13500 65W is slowly getting back in stock but there is a markup now! It is a premium of around $50-$90! MSRP is: $242.00! Online retail prices at $297-$350!

 

[Harry_Wild]

Still waiting patiently for Intel to release the 13600 rumor to be happening in February! If need be, I will pay the markup to buy it just to get over with!

 

[igor_kavinski]

PANZER EVO RGB - The RGB Crystalline Titan - COUGAR

PANZER EVO RGB The RGB Crystalline Titan Panzer EVO RGB is the perfect combination of four massive tempered glass panels and four powerful ARGB fans in an exceptional case. With the included fan controller, you will be able to control all the power of each Fan’s 18 LEDs and enjoy more than 100...

cougargaming.com

DO NOT get this case unless you prefer a frustrating build experience.

It's huge but actually quite cramped on the inside. The screws that came with the case had trouble bolting the mobo properly. One or two corners of the mobo were coming loose because the screw was getting pulled out of the screwhole. Extremely annoying. The PSU is also from Cougar. 1050W. But the supplied cables were barely long. Did not enjoy stretching them at all but at least the system is up and running. More here: https://forums.anandtech.com/threads/help-me-decide-i7-12700k-or-13700k.2610235/post-40940702

 

[Rigg]

I tuned my RAM up to 6800 MT/s. I tuned the primaries, tWR, tFAW, tRFC2, tRFCsb, and tREFI. It went really smooth once I gave up trying to get 7000 to work. Passes OCCT, memtest86, the dram calc memtest 1000% per thread, and y-cruncher Pi-2.5b. Not bad for a fairly cheap Hynix M-die memory kit intended for AM5.

F5-6000J3038F16GX2-FX5 - G.SKILL International Enterprise Co., Ltd.

Flare X5 DDR5-6000 CL30-38-38-96 1.35V 32GB (2x16GB) AMD EXPO Designed for DDR5-enabled AMD AM5 platforms, Flare X5 series DDR5 memory is built for high performance in a low-profile design. Featuring AMD EXPO overclocking profiles for easy memory overclocking, users can easily enable overclocked...

www.gskill.com

 

[Rigg]

Just an FYI to the thread followers IMC and SA voltages are complete overkill on my Asus board. I'm stable at 1.225 on both. With auto the board was pushing 1.4 v to the IMC at higher speeds . This seems to be a tradition for Asus. My Maximus XI hero did the same thing. Probably a good recipe for burned up memory controllers and unstable overclocks.

 

[Harry_Wild]

Intel non K 65W CPU except the 13900 is still constrained until March-April! I decided not to buy my 13600 until Fall, at which time; the 13600, NvMe, 6200 RAM and motherboards for it should be more consumer friendly price!😁 Also, the next generation Intel will be promoted at that time, my buy should be selling at a 20% discount by then instead of a 10% premium at the moment!

$800 X .80 is: $640.00! Save $160.00.

 

[coercitiv]

Just an FYI to the thread followers IMC and SA voltages are complete overkill on my Asus board. I'm stable at 1.225 on both. With auto the board was pushing 1.4 v to the IMC at higher speeds . This seems to be a tradition for Asus. My Maximus XI hero did the same thing. Probably a good recipe for burned up memory controllers and unstable overclocks.

This happens to other brands as well, activating high memory speeds through XMP can push voltages up. I assume they do it to ensure easy "one click" overclocks. It seems to scale with memory speeds, at least it did on my MSI Z370 board. I don't remember if I had a similar situation on the MSI Z690 board /w 3600MT/s memory, but I suspect I never allowed this board to boot with SA voltage on Auto after enabling XMP.

 

[Carfax83]

I tuned my RAM up to 6800 MT/s. I tuned the primaries, tWR, tFAW, tRFC2, tRFCsb, and tREFI. It went really smooth once I gave up trying to get 7000 to work. Passes OCCT, memtest86, the dram calc memtest 1000% per thread, and y-cruncher Pi-2.5b. Not bad for a fairly cheap Hynix M-die memory kit intended for AM5.

Wow, you got into the 40s for latency! The Hynix A-die memory seems to be really strong in the frequency department and delivering bandwidth, but when it comes to timings, the M-die is better, or so I hear.

Is the tune going to be your daily driver, or is it just for benchmarking?

 

[Rigg]

Is the tune going to be your daily driver, or is it just for benchmarking?

Assuming I don't run into any stability problems in games I'll daily drive it. It seems really stable so far and passes memory stress tests. I really didn't have to push the voltages to get there. 1.35 v on the memory and 1.225 on the IMC/SA. My only concern is thermals causing instability with the tRFC and tREFI being pushed pretty hard. Only saw peaks of 51c on the sensors during OCCT tests. Average was below 50c. Hopefully that's good enough to not cause any issues when the GPU is dumping hot air into the case.

 

[BoomerD]

Nothing so fancy as many of youse...

i5-13600K
MSI Mag Z690 Tomahawk DDR4
32GB Corsair Vengance LPX DDR4-3600 RAM
Arctic Cooling Liquid Freezer II 280mm
EVGA RTX3070 FTW3
Western Digital SN770 2TB NVMe
HP EX950 1TB NVMe
Western Digital Black 2TB HDD
Crucial MX500 1TB SSD
Seasonic Focus+ Gold PSU
be quiet! Dark Base Pro 900 Rev.2 case
3X140 mm Dark Wings fans PWM

 

[Carfax83]

Assuming I don't run into any stability problems in games I'll daily drive it. It seems really stable so far and passes memory stress tests. I really didn't have to push the voltages to get there. 1.35 v on the memory and 1.225 on the IMC/SA. My only concern is thermals causing instability with the tRFC and tREFI being pushed pretty hard. Only saw peaks of 51c on the sensors during OCCT tests. Average was below 50c. Hopefully that's good enough to not cause any issues when the GPU is dumping hot air into the case.

Personally I don't bother stress testing memory, because stress testing typically isolates the RAM itself rather than stressing how the RAM interacts with the rest of the system. To me, games are much better for that.

Case in point, my DDR5 7800 tune was unstable at 1.475v and I had to bump it up to 1.5v to be stable in some of the games I play. To me it wasn't worth it though, so I just dropped down to DDR5 7600 at 1.45v with less aggressive timings.

Games like the Dead Space remake will find any unstable memory overclocks or timings REAL quick!

 

[Harry_Wild]

ASUS Prime Z790M-Plus LGA 1700(Intel 12th&13th Gen) microATX motherboard (PCIe 5.0, 3xM.2 slots, 10+1 DrMOS, DDR5,1 Gb LAN, DP, USB 3.2 Gen 2x2 Type-C®,front USB 3.2 Gen 1 Type-C, Thunderbolt™ (USB4) support)

Intel 13700 CPU 65W

Crucial P5 Plus 2TB PCIe Gen4 3D NAND NVMe M.2 Gaming SSD, up to 6600MB/s - CT2000P5PSSD8

Crucial 32GB (2 x 16GB) DDR5 5600 (PC5 44800) Desktop Memory Model CT2K16G56C46U5

To upgrade my Lenovo SFF Desktop slim tower.

Update:

Purchased Seagate 530 FireCuda instead of the Crucial P5 Plus!