PCTuning fórum

Sobo píše:V tom update videu Roman taky rikal, ze s defaultnim nastavenim biosu (mysleno jeste mimo nasobic CPU a voltaz) bude v soucasnosti kazda X299 deska throtlovat, jenom to uzivatele zkratka nejsou schopni zachytit, protoze se to docela tezko poznava.

Treba u ASUSu by se melo vypnout SVID a nastavit Digi+ Power Control na 140%, aby to neomezovalo proudy do CPU. To je ten duvod, proc ve VRM testu od Toma OC3D nameril teploty VRM jenom kolem 70°C - throtlovalo mu to a nevedel o tom. Kdyz to prenastavil, spotreba CPU se zvedla o 100W a uz to slo do haje. Taky asi dost zalezi na tom, v cem se to testuje - ruzne mody Prime95 muzou dle videa udelat rozdil az 200W na sestave.

Blue Storm píše:nic o defaultu neříká, naopak říká, že to má člověk přenastavit, pokud chce dosáhnout co největší spotřeby procesoru.

Blue Storm píše: Jinak co se těch renderů týče, tak ještě jsem nezažil, že by nějaký při testech žral více jak Cinebench. Sám k měření spotřeby a teplot v záteži používám AVX zátěž při renderování H.265 a spotřeba je vždy menší než v Cinebenchi.

Je vidět, že maximální AVX zátěž i při relativně nízkém nastaveném napětí (1,16 V) zatíží kaskádu tak, že začne po několika minutách omezovat svůj i procesorový výkon.

v nesmyslném nastavení Prime95

elmor píše:For example, at 300W output with 90% efficiency means input power is 330W. That's 30W dissipated mainly by the MOSFETs. Using 8 phases gives 30/8 = 3.75W per MOSFET. Delta between PCB and MOSFET junction is then 5*3.75 = 18.75*C, so if you're hitting throttling when the VRM is reading 105*C you're looking at 105+18.75=123.75*C MOSFET junction temperature. All the MOSFETs mentioned in this post have recommended max junction temperature (continuous) at 125*C and absolute max ratings (short term) around 140-150*C.

Without delidding, you're looking at CPU power consumption of max 250W on 6/8/10C with a good AIO cooler when increasing Tjmax to 105*C (94*C default). After delidding, or considering higher core count CPUs, this number is increased to the 300-400W range. If you add 1+1, you can figure out why so many boards do not have better VRM cooling.

At 250W output you're looking at roughly 140A of current at 1.8V VCCIN. Most VRMs for X299 are rated at 400A or more, which is overkill. This is, of course, assuming that you can take care of the heat generated by the VRM at that output current. The kicker here is that this limit is the same no matter how many phases a board has. Going to higher phase counts doesn't change efficiency by much. And, you still have the same total power dissipation from the VRM that you have to remove within a very limited footprint. If you have the same heatsink for two VRMs which are identical except for phase count, you will still be limited to the same max continuous output current.


ASUS X299

Solutions are all ASP1405 in 7/8-phase mode using 7/8x IR3555 (60A). Rating without considering thermals is 480A output (420A with 7 phases). VR_HOT is triggered at 105*C. Switching frequency is 500 KHz. The only difference between boards are the inductors and the heatsink solution. Prime X299-A/Strix X299-E have 7 phases for VCCIN, Prime X299-Deluxe/TUF X299 Mark1 have 8 phases.

Nothing new here, same VRMs as on the X99 models. You can check der8auer's video on what to expect from the heatsinks, about 250W continuous on Prime X299-A/Strix X299-E Gaming/Prime X299-Deluxe. TUF X299 Mark1 is slightly better because the backplate helps dissipate the VRM heat (5-10*C lower at same output). That's inside a case with minimal airflow and stressing the GPU at the same time. R6A/R6E will be much better, stay tuned.


Gigabyte X299

I've only had a look at the X299 AORUS Gaming 3 which uses IR35201 in 8-phase mode using 8x IR3556 (50A), but I'd expect it to be the same on the more expensive models, perhaps with IR3555 instead. Rating without considering thermals is 400A. VR_HOT is triggered at 115*C.

Design is pretty much the same as the ASUS boards. Only thing that might cause some concern is the higher throttling limit, which also allows the CPU to pull a bit more power than the competitors. Using the example from above at 300W output you'll be at 130-135*C MOSFET junction temperature before throttling. According to der8auer's data on the board, you can output up to 290W before hitting the 115*C limit.


MSI X299

Best case first, X299 Gaming M7 ACK. This thing has IR35201 in 5-phase mode, doubled using IR3599 to 10x IR3555 (60A). Rating without considering thermals is 600A. VR_HOT is triggered at 105*C. Switching frequency is configured at 300 KHz, after doubling this is effectively 150 KHz.

Quite a nice VRM setup, but the LED infused heatsink is not great. As described above, there's no point to have this many phases when the heatsink is unable to dissipate enough heat for it to make a difference. From my own testing, about 230W continuous output is what you're looking at without direct airflow before reaching 105*C. That's on an open test bed.

Going down in price, things are getting interesting with MSI. X299 Tomahawk Arctic is also using an IR35201 controller, but this time in 4-phase mode. This is then doubled to 8-phase and driven using IR3598. 4-phases are on top and 4-phases at the bottom. Each phase consists of 2x NIKOS PK616BA for high-side and 2x NIKOS PK632BA for low-side. It's a bit more complicated to calculate theoretical max for this configuration, but without considering the heatsink a simple estimate is the high-side continuous current at 25*C ambient temperature. This gives 2x13A = 26A per phase, total 208A (same method for IR3555 yields a derating from 60A to 56A).

Not sure what to say about this board, I'll let my test data speak for itself. The heatsink is a metal plate with another piece screwed on top. Going into the OS fully idle at desktop with a 10C CPU, the VRM thermistor is reading 60*C. Power output at this point is roughly 10W. After Prime95 testing, I had to settle for 175W output before reaching the 105*C thermal limit of the VRM. As with the other MSI board, this was on an open test bench without direct airflow on the VRM.


ASRock X299

The board I tested was the Fatal1ty X299 Professional Gaming i9. A little bit of a surprise here, they went with an Intersil ISL69138. It's configured in 6-phase mode with doubling to 12 phases using ISL6617. For output they use ISL99227 power stages which are rated to 60A. Rating without considering thermals is 720A. Thermal throttling occurs when then power stage reaches 140*C (using internal sense).

ASRock has a very nice design here capable of very high theoretical output. However, the power dissipation issue still remains and the heatsink is just as underpowered as the others. Because of the higher throttling limit (measured 132*C at the center of the inductors close to power stages), it's able to accommodate up to 320W output power before throttling. A cause of concern here is that at this point the power stage will be close to 140*C, while the recommended continuous temperature is the same as the others at 125*C.

Sobo píše:Dalsi recenze snad se vsemi SKX: https://nl.hardware.info/reviews/7443/3 ... n-premiere

racecar píše:

Sobo píše:Dalsi recenze snad se vsemi SKX: https://nl.hardware.info/reviews/7443/3 ... n-premiere

Tie hry su desne.

Sobo píše:Nah, pretaktujes mesh z 2400 na 3200 a melo by to byt v podstate stejne jako SKL nebo BWE.

racecar píše:Fakt som chcel investovat ale Intel robi vsetko preto aby som "setril".