[OFFICIAL] ASRock z97 OC Formula Thread / Tips / Discussion / BIOS / Tools | Page 3
JavaScript is disabled. For a better experience, please enable JavaScript in your browser before proceeding.
***Offical ASRock z97 OC Formula Thread — Tips / Discussion / BIOS / Tools***
This round ASRock has decided to create just a normal ATX form factor board, so there will just be the large OCF not an M.
The board will have the same coating as the z87 which im super happy about.
Its held up very well for me in the past even after heavy frozen memory benching countless times.
The price point in the USA will be $216-$225 for a fully featured flag ship board is :up:
Courtesy HKEPC
The z97 board pics up right where the z87 left off. Its clocks all IC’s of memory just as well and doesnt have the hiccups that the z87 board had. ( Delayed post hanging on 9D or A3 for a couple seconds before booting normal ) Efficiency is top notch as usual
Thanks to L0ud for showing off what the Samsung and MFR can do.
Samsung
Hynix
Here is some examples with classic ICs PSC and BBSE
PSC
BBSE
BIOS
**Coming Soon
Tools
OC tools 0512
Last Update
5/18/2014
See less
See more
6
Reply
Save
Like
Reactions:
1
41 — 52 of 52 Posts
Quote:
Originally Posted by spdaimon
I’ll look to see if that is an option. I don’t remember seeing that.
Click to expand…
Thanks, it would be great!
Reply
Save
Like
Quote:
Originally Posted by davidkoleda
Thanks, it would be great!
Click to expand…
Nothing that I can see. Just the usual fan modes — Turbo, Performance, Standard, Silent. I know what you are talking about. One of my boards has an option for PWM or Voltage control that you select. Not sure which one it was, maybe one of the Gigabytes I have/had. Looks like you may just need to get something like Zalman’s FanMate 2 to control fan speed for single fan or get a multi-channel model. I have a Zalman and a couple Lamptron on a couple of my rigs. They both work well. The Lamptrons are a bit ‘flashier’, if you like that sort of thing. I don’t use the temp wires so I got the FC-6 so it will display voltages instead, for what its worth.
EDIT: I was thinking about it, and I guess there is some sort of voltage or PWM fan control. There are 6 or 7 fan headers. The BIOS has them grouped into twos for control. There is 3 pin and 4 pin headers. Majority is 3 pin, so those would need to be voltage controlled to work. So the answer to your question would be yes. You don’t have finite control over them but do have the preset fan levels. They seem to work like setting the fans to 50%, 55%, 60%, etc, rather than a fan curve or Cool N’ Quiet (which I think is Asus’s thing). I’ve set them to silent. ..got 7 fans in my Corsair 750D… half are running off a fan hub and gets a bit loud otherwise.
See less
See more
Reply
Save
Like
Still getting the random turn on issue. It doesnt happen consistantly so not sure if swapping out the board to another case would help trouble shoot. I think someone said this was a bug in Z97. I cant find whats triggering it otherwise.
Reply
Save
Like
Hi there, I have the Problem that my G1 980 TI runs with PCI-E 3.0 x8(shown via GPU-Z) in Slot 1 and not with X16 Speed. It shows that with Heavy Load. Without Load Card runs as PCI-E 1.1 X8. In Mainboard BIOS I selected GEN 3 for all Slots. Any Idea why?
Reply
Save
Like
Quote:
Originally Posted by Atzenkeeper500
Hi there, I have the Problem that my G1 980 TI runs with PCI-E 3.0 x8(shown via GPU-Z) in Slot 1 and not with X16 Speed. It shows that with Heavy Load. Without Load Card runs as PCI-E 1.1 X8. In Mainboard BIOS I selected GEN 3 for all Slots.
Any Idea why?
Click to expand…
I’m assuming you don’t have anything in any of the other slots?
Reply
Save
Like
Where is CPU strap in BIOS of Z97 OC Formula? My 4980HQ miracle chinese can only 4.6 GHz.
I found this topic, but don`t work method from here for my case.
Reply
Save
Like
Reactions:
1
Kashtan said:
Where is CPU strap in BIOS of Z97 OC Formula? My 4980HQ miracle chinese can only 4.6 GHz.
I found this topic, but don`t work method from here for my case.Click to expand…
Can you show your 4.6G OC of 4980HQ?
Reply
Save
Like
YoungChris said:
Can you show your 4.6G OC of 4980HQ?
Click to expand…
4.8 Z87 Deluxe
Asus Z87-DELUXE Performance Results — UserBenchmark
4.7 Z87 Deluxe
4.6 Z87 Deluxe
4.75 Z87 Deluxe (CPU Strap is allowed)
4. 8 Z87 Deluxe (result is bad because windows is broke on moment of bench)
4.9 CPU-Z
4.7 Passmark
My OCF Z97 worked less then 4 weeks. Is dead. Now i have 3 OCF Z97 which is not worked. I purchase 4-th but haven’t turned it on yet.
See less
See more
Reply
Save
Like
Kashtan said:
4.8 Z87 Deluxe
Asus Z87-DELUXE Performance Results — UserBenchmark
4.7 Z87 Deluxe4.6 Z87 Deluxe
4.75 Z87 Deluxe (CPU Strap is allowed)
4.8 Z87 Deluxe (result is bad because windows is broke on moment of bench)
4.9 CPU-Z
4.7 Passmark
My OCF Z97 worked less then 4 weeks. Is dead. Now i have 3 OCF Z97 which is not worked. I purchase 4-th but haven’t turned it on yet.
Click to expand…
Impressive OC results! Did you OC in bios or in OS? Looks like a pretty good chip.
Reply
Save
Like
YoungChris said:
Impressive OC results! Did you OC in bios or in OS? Looks like a pretty good chip.
![]()
Click to expand…
Thank you.
Both. BIOS and ThrottleStop .
This chip i`m hunting 2 years. This chip change a 5 owners before go to me. Now my Z87 Deluxe also have problems, I can not hand over for repair, due to problems with logistics. I still have a lot of motherboards in need of repair. There are 3 in total now.
I had many more benchmarks, I’ll look for the results.
Reply
Save
Like
https://corona-renderer.com/benchmark/results/cpu/4980
Reply
Save
Like
4980hq — Geekbench Search — Geekbench
4980hq — Geekbench Search — Geekbench
4980hq — Geekbench Search — Geekbench
4980hq — Geekbench Search — Geekbench
geekbench 5
4980HQ 12,783 results found
best single core 4980hq 1254/4421
best multi core 4980hq 1133/4621
my 1227/4800 4.75 1.49 4.75 1.39
See less
See more
Reply
Save
Like
41 — 52 of 52 Posts
- This is an older thread, you may not receive a response, and could
be reviving an old thread.Please consider creating a new thread.
Top
Intel Broadwell-C (5675C/5775C) Ownership Club | Page 213
JavaScript is disabled. For a better experience, please enable JavaScript in your browser before proceeding.
Broadwell C Club for discussion on the 5675c and 5775c.
Reply
Save
Like
Reactions:
2
4241 — 4260 of 4497 Posts
Sad, no DIY conversion available.
Reply
Save
Like
Kashtan said:
I finding few video — 4980HQ (also with EDRAM 128mb) better than Coffee.
https://youtu.be/H8eA-3kXjQ8?t=263
https://youtu.be/IOiRM5IdDj0?t=369https://www.ixbt.com/cpu/results/result-2017.xls
The advantage of level 4 cache over Coffey is undisputed.
Are 4980 and 5950 really better than 5775C? They accelerate to 4.6-5.0 as reported — this is amazing if true.Click to expand…
That isn’t true. The 5950hq is multiplier capped at 43x for all core loads and the 4980hq is capped at 44x for all core. The guy in that video must be using bclk of 125 since his 1600 edram is running at 2000. The 5950hq is basically a binned for low power 5775c and the 4980hq is slower in terms of performance in basically every way by a little. They aren’t crazy fast or anything and the magic reform adapter is relatively fragile to heavy power use. I either toasted my 5950hq or one of the solder beads in the extra layer stopped connecting. The adapted versions are also harder to overclock since a lot of bios stuff just doesn’t to seem to work quite right. Also neither my Z87 Deluxe or Z97 Hero will accept either as a valid cpu. But my h81t takes the 4980hq fine. And EVGA z97 take both.
The 4980hq is good if you want edram in a 8 series chipset and don’t mind having a slightly watered down 5775c. My 4980hq doesn’t have a problem doing 4.4 ghz at 1.35v, but I’m running it at 4.2ghz at 1.2v so it doesn’t meet the same fate as my 5950hq. I would rather go with a 5775c over either. Also somebody is selling 5775c’s for $150 from China on Ebay right now. I already have enough though.
Edit: 4.4ghz at 1.3v
See less
See more
Reply
Save
Like
Reactions:
1
Kashtan said:
I finding few video — 4980HQ (also with EDRAM 128mb) better than Coffee.
https://youtu.be/H8eA-3kXjQ8?t=263
https://youtu.be/IOiRM5IdDj0?t=369https://www.ixbt.com/cpu/results/result-2017.xls
The advantage of level 4 cache over Coffey is undisputed.
Are 4980 and 5950 really better than 5775C? They accelerate to 4.6-5.0 as reported — this is amazing if true.Click to expand…
https://techreport.com/review/34205/checking-in-on-intels-core-i7-5775c-for-gaming-in-2018/
This review says otherwise where even Zen+ is beating it in most titles.
Reply
Save
Like
TwilightRavens said:
https://techreport.
com/review/34205/checking-in-on-intels-core-i7-5775c-for-gaming-in-2018/
This review says otherwise where even Zen+ is beating it in most titles.
Click to expand…
You can get significantly better frametimes by disabling spectre mitigations
Reply
Save
Like
TwilightRavens said:
https://techreport.com/review/34205/checking-in-on-intels-core-i7-5775c-for-gaming-in-2018/
This review says otherwise where even Zen+ is beating it in most titles.
Click to expand…
I see this review.
Results is a very strange. 2600X better then 5775C? On Deus Ex: Mankind Divided? Guys from this site maybe still on other side this planet from Poland — purepc and oclab from Poland, which is a very highest scores take to Broadwell.
Reply
Save
Like
rluker5 said:
You can get significantly better frametimes by disabling spectre mitigations
Click to expand…
Yeah, those mitigations on broadwell are pretty harsh. I’ve seen ~10% speed up just from disabling them in most synthetics.
Reply
Save
Like
TwilightRavens said:
https://techreport.com/review/34205/checking-in-on-intels-core-i7-5775c-for-gaming-in-2018/
This review says otherwise where even Zen+ is beating it in most titles.
Click to expand…
Well, tbh Zen+ with 3200 cl14 is barely beating the 5775c stock with 1866 cl9 with the monster of a 2080ti. Put that 5775c to 4.1-4.2GHz, slam 2400 cl10 and Zen+ sees the floor.
Reply
Save
Like
Ansau said:
Well, tbh Zen+ with 3200 cl14 is barely beating the 5775c stock with 1866 cl9 with the monster of a 2080ti. Put that 5775c to 4.1-4.2GHz, slam 2400 cl10 and Zen+ sees the floor.
Click to expand…
Yes but Zen2 is +10-15% on top of that on a bad day.
Reply
Save
Like
Should I get a Z97 MPower or give it a rest? I’d lose 40USD on that change.
Reply
Save
Like
TwilightRavens said:
Yes but Zen2 is +10-15% on top of that on a bad day.
Click to expand…
And what, it just shows you how good for gaming still are these 4c/8t intel cpus. You need to go twice the cores + fast ddr4 to have more than a marginal improvement in AMD. Intel 6-10 cores are a different story.
Reply
Save
Like
Seems like the prices are dropping for a while: https://www.aliexpress.com/wholesale?catId=0&initiative_id=SB_20200720095507&SearchText=5775c
Is what it is. They are still good for gaming, especially with xbox announcing at least a couple of years of compatibility with the older jaguar versions.
Reply
Save
Like
So, I got myself Z97 MPower, because why not and Gigabytes atrocious dual bios made me lose my temper at last. Quick test surfaced an issue that in order to use iGPU you need to have bios version rather more recent (did not check which one is required) than the one from Broadwell release, despite it’s being a Z97 motherboard and NVMe support require that as well (lol). Also, you can still disable L4 by accident without disabling it, which is extremely annoying and happened at both Maximus VII Hero (manually settinf eDRAM i.e.) and Z97X Gaming X (disabling C6 state).
The board lets you set 2MHz (1MHz is default) switching frequency for FIVR, which is nuts .
Reply
Save
Like
Very interesting comments under this video.
https://www.youtube.com/watch?v=H8eA-3kXjQ
I translate from russian:
«
tig dolob
At 5775s, for example 17 ucode gives + 15-20% in games compared to FF03. Asassin has 150 fps instead of 120 fps at 4500/4200/2600 with 4×4 2933cl11 memory (avexir). And that’s faster than 9900k @ 5.1 @ 4400cl17 for a second!
5775s on Avito only one and they want 30k. I took the top 2% of this amount 3 years ago. 4.5 to 1.34 in prime passes
«
Better then CFL-R! Exactly! Holy grail!
See less
See more
Reply
Save
Like
3 things matter:
1) Better RAM from possible — high freq low latency
2) Binning is work — 4. 5 is possible
3) Microcode — is VERY important. +20% FPS! No kidding.
Reply
Save
Like
Yeah, no, uC doesn’t play that much of a role. If he got +20% performance uplift just by changing uC then he fkes up something terribly. Most likely had L4 disabled.
Reply
Save
Like
Kashtan said:
3 things matter:
1) Better RAM from possible — high freq low latency
2) Binning is work — 4.5 is possible
3) Microcode — is VERY important. +20% FPS! No kidding.Click to expand…
- Unobtainium
- Unobtainium
- User error
BroadPwns said:
Yeah, no, uC doesn’t play that much of a role. If he got +20% performance uplift just by changing uC then he fkes up something terribly. Most likely had L4 disabled.
Click to expand…
Quite likely.
These CPUs weren’t as awful in certain tasks but their high price and low sales made them a rarity along with borderline «criminal» support from mobo makers.
No one sells them even 2nd hand almost at all since no one almost bought them in first place, can find a few shady Chinese sellers at ridiculous prices, that’s all.
Finding one locally outside US for sale is a miracle for ages now.
See less
See more
Reply
Save
Like
I finding huge quantiity CPU with EDRAM on ali. ebay and sure tao. So this no problem. And real cheap. 66 USD for 5775C. Sure ship and other — i guess price reise to one hundred USD, but availability here.
Reply
Save
Like
Also, that vid you’ve linked? I can’t even open it.
EDIT: Tingled a bit with mah new MPower — L4 can be disabled with early iGPU frequency change, . By early I mean before you dial in literally every other setting you want to set, I didn’t bother to check which option change made iGPU frequency setting disable eDRAM. I’ve found that it do that while doing checker box to write off «safe» options that did not disable it. After I’ve left iGPU frequency at the very last it now does work and I’m surfing on a 1300MHz Iris Pro 6200 with 2000MHz of a working L4 cache. Minus side? There is no freaking power draw reading.
I’m deffinitely keeping that little, ugly yellow boi.
See less
See more
Reply
Save
Like
I also can’t open the video.
But concerning the microcodes, F03 and 17 are a horse apiece for me. With F03 I get about 10% better 4k random reads and writes with cdm, 5% worse ram bandwidth, same edram results (last 2 from trial version of AIDA), slightly worse for some productivity benchmarks and slightly smoother gaming with slightly higher averages (consistent, but still margin of error scale) in some, but not all games I tested with the rest of games being the same -vs 17.
I prefer F03 because it makes my expensive Optane perform better and it seems like it brings out more of what makes a 5775c different from other cpus, but the difference is really minimal overall in games.
I believe F03 did not come out with any motherboards from the dates the others were released, never having seen a microcode nearly as old in a bios, and that it is not recognized as existing per either cpu info in the bios or hwinfo64.
So I think the only way you can get it in there is with ubu, or manually inserting it yourself.
Not really anything for anyone to have to worry about, especially with windows overriding the motherboard microcode unless you stop it from doing so.
And the only way I could get performance to drop by double digits in any game is to disable edram. Not even enabling mitigations is that bad. Also I top out about 100fps average with cpu limited, 4.2, 3.8,2100, 2400cas10 in AC Odyssey (with just a 1080ti, but still set to cpu limited) and 50% more than that is pretty unbelievable for about a 10% overclock over that.
And congrats BroadPwns on getting a top notch mobo that works great with your cpu. It is good to know that one is a winner.
See less
See more
Reply
Save
Like
This mobo has an option to set FIVR frequency from 300kHz to… 2MHz. No way to measure it ofc. Putting the uC into a bios is trivial with UBU, you only need AMI Tool v4.5 or 5. 0.0.7 (I have both), then remove intelgenuineuc or something like that from System32. F03 proved to be terrible on Z97X Gaming 5 too, MPower has 20, which doesn’t seem to differ from 17.
Reply
Save
Like
4241 — 4260 of 4497 Posts
Top
CrystalWell processors: big FAQ
This article was written by a visitor and has been rewarded.
DESCRIPTION:
Lots of processors of the Crystal Well family, adapted by the Chinese for socket 1150, appeared on sale in China. Most of the lots on Taobao, there is also a seller on Aliexpress, and you can also find 4980hq with a small margin on sale in Russian Avito.
recommendations
Processors — similar to the Broadwell 1150 line, but made on the 22nm process technology. 4 hyperthreaded cores, 6MB main cache instead of Haswell’s 8MB, 128MB L4 cache, equipped with a nice Intel Iris. They are an excellent choice for upgrading systems with i3 and i5 on lower chipsets, because, again, thanks to the L4 cache, they do not need fast RAM
WARNING: These processors do not work on most Gigabyte motherboards try out different versions of Bios — in particular, on my Z87X-UD4H, the processor runs with F9 bios, but does not work with bios F10. There are nuances in the work with F9 — overclocking with a strap does not work)
3.5GHz
4850hq : 3.7GHz
4870hq : 3.9GHz
4960hq : 4.2GHz
4980hq : 4.4GHz
PERFORMANCE:
performance in modern games. For example, CrystalWell 4870hq, which has a maximum frequency set by a multiplier without a strap — 3.9, can be compared with a 4790k processor at 4.8GHz.
At the same time, after overcoming the frequency of 4.2 GHz, there is no big increase in gaming performance, which, in fact, makes the 4960hq processor a good option for a budget upgrade of a Low-end system from i3-i5. However, the younger 4870hq with a frequency of 3.9 also performs well:
0006 are absolutely unpretentious to memory, which again makes them just an ideal option for upgrading systems based on budget chipsets without the possibility of overclocking RAM, and cheap non-overclockable memory:
customized solution for games, CW looks more interesting than zeons V1 and V2:
In computing tasks, the performance of CW absolutely repeats the performance of Haswell processors at a similar frequency:
There is an increase in some specific tasks, such as archiving with Winrar, or, as one of the users noted, VP9 encoding (tablet):
PROCESSOR SELLERS:
At the moment, you can find processors with different versions of the substrate-adapter and cover. Everything described below is based on the experience of processors available to users, but the entire range has not yet been considered. There were three main types of covers:
1) Milling plate
details in chapter IRON NUANCES ) . Thermal paste is applied under the cover, high temperatures in the drain. But improvement is possible. Easily scalped. Apparently, the cover is tinned (not sure). Lots of processors with such a cover are found both on taobao and, more recently, on aliexpress, and are much cheaper than lots with a «beautiful cover». In general, if you have straight arms, this is a good option for a cost-saving purchase. The processor with this cover that one of the users had proved to be quite good — the straps function, there are no problems with the processor.
To apply ZhM, the coating is grinded down to copper — it is believed that the cover is covered with tin, but I believe that this is not so, because on the processor that I had in use I did not grind the coating from the inside — but on the outside the cover turned out to be bent and required surface leveling. Absolutely all contact pads (of which there are a lot on SW processors) and SMD elements are insulated with varnish, the cover is seated on the sealant around, be sure to seal the windows from the ends of the cover with sealant so that the ZhM does not leak into the socket, and also make sure that between the clamping areas of the cover and a sealant substrate-adapter is also available, this is necessary to unload the BGA soldering and distribute the clamping load.
2) Milled lid with glossy finish and sharp corners.
I’ll tell you about the next lot:
Not the best version of the cover from this trio. I don’t recommend buying lots similar to this one (although, as I already said, there is still little information on CW variations, and you can only judge by the models on hand. A similar externally processor may differ from this particular instance in performance). Under the cover is thermal paste. The drain temperatures are extremely high. One of the users has a processor in this design, and it proved to be extremely bad. Firstly, this instance refused to work with two memory channels on half of the tested motherboards — for example, it refused to start when installing memory modules in the slots of the second channel on: MSI H97-PC Mate, Gigabyte Z87X-UD4H, while all 4 slots worked without problems on MSI Z97 GAMING5, a pair of motherboards from Asus and Asrock. Judging by the fact that once we managed to launch both channels and see the coveted 16GB of RAM on the Gigabyte Z87X-UD4H, it’s a matter of wiring and positioning the contacts on the adapter’s textolite. Secondly, this processor flatly refused to work with strap 125. In addition, the height of the lid above the thrust legs in this case is very small, and on a gigabyte, where the socket mandrel is made of slightly thicker metal, the mandrel turned out to be a level higher than the surface of the lid, which made it impossible CO installation. Among other things, this specimen is difficult to scalp.
The cover is seated on a rigid compound, like the one on which the Fermi family GPU covers were seated. For scalping, it is necessary to clamp the base in a vise, and, heating with a hairdryer, pry off the lid from the edges with a knife, without otherwise damaging the sandwich, do not remove the lid
The shiny coating is thin, easily scratched to copper. What is very unpleasant — the lid does not reach the crystal. To shrink on the FM, the lid must be filed over the area of the stop in the crystal substrate, and, if necessary, the paws resting on the adapter substrate, as well. It is necessary to make sure that there is a complete spot of contact between the crystal and the lid:
Further standard: contact pads and SMD elements are insulated with varnish, the cover is seated on the sealant around, between the clamping pads of the cover and the adapter substrate, the sealant is also necessary to unload the BGA soldering and distribute the pressure load.
After installing the cover on the LCD, the cover starts to perform its heat-distributing functions perfectly (the processor voltage on the screen is below 1.25v):
0033
The most successful version of the cover, which I would recommend for purchase. You can find it on sale on the Russian Avito, as well as on taobao, from the seller cpu_wholesaler (There are models in the assortment: 4980hq, 4870hq, 4750hq). The processor adapter substrate in the case of the existing instance is black. This copy proved to be the best, worked on all tested motherboards, including the Gigabyte, on which the copy with the cover described above refused to work with the second memory channel. Overclocking with straps works without problems.
The height of the lid fits perfectly into the socket with any mandrel. The cover from the factory is seated on a soft black sealant. Easily scalped by prying with a knife from the edges. Liquid metal is applied under the cover, and the space around is filled with thermal paste. The LM is applied terribly, so the need for scalping does not disappear, however, a specimen may come across where the LM is applied, more or less qualitatively:
There is no gap between the lid and the crystal, there is a contact spot. To reapply ZhM, it is necessary to wash off the Chinese varnish and re-insulate the SMD components, clean off the remnants of the old sealant, apply ZhM and seat the cover on the sealant in a circle. It is also necessary to make sure that the gap between the thrust tabs and the adapter substrate is also filled with sealant, which is required to distribute the clamping load on the adapter textolite and relieve the load from the BGA soldering. This embodiment is good in that no manipulations with the lid itself are required.
After shrinking this cover on the LCD, it perfectly performs its heat distributing functions:
4) Processors from Ibey
You can also find it on sale from a Chinese in Russian Avito. The processor adapter substrate is usually green. The lid is clumsy, there are photos of processors from three different batches — they all differ in appearance. There were no problems with the processor. Overclocking with straps works without problems.
A small nice thing, the processor came to me in this package:
These processors are very hard to scalp. Sergey with Avito suggested pre-cutting the sealant in accessible places under the cover — in fact, that’s what I did. To begin with, I cut the sealant with a thin probe from the car kit, then heated the cover with a hairdryer and began to separate it by inserting a thin knife into the spacer of the resulting gap between the cover itself and the processor (not the adapter). Under the cover is thermal paste. The lid is made clumsily — it has a rough finish on the outside, there is a gap between the crystal and the lid from the inside, that is, it needs some work:
Y:
To install CrystalWell processors adapted by the Chinese into the socket, it is necessary to unscrew the lower bolt by half a millimeter-millimeter fixing the socket mandrel — which is the guide for the processor gate. It is not necessary to put a washer — just unscrew it.
It is also necessary to put a gasket from 1 mm under the top 2 bolts (A gasket is needed, because otherwise the platform under the bolts will start to press on the textolite with tracks with an edge). If this is not done, the thin textolite of the adapter may be deformed:
0005
Example of spacer washers:
In general, the point of fitting a socket is to provide the minimum amount of pressure that the processor will work with. At the same time, if the processor refuses to work, it is advisable to leave the upper bolts unscrewed, and adjust the clamp with the lower one!
As a result, when clamping the socket, the gate of the processor clamp must be visually parallel to the processor cover. It is advisable to make sure that the processor cover is a level higher than the socket gate:
Since these processors are higher than standard ones, it is not worth tightening the CO fastening screws all the way — it is necessary to leave a gap of a millimeter. There may also be problems with CO variants that have clip-on fasteners to the board, the height may not be enough, however, I had experience installing one such tower — and in my case the problem did not appear.
Since the textolite of the adapter itself is quite thin, like in Skylake and Kabylake processors, massive towers can be additionally fixed to the case.
Another iron point worth knowing about: in some implementations of the adapter, the Chinese save on gold-plated contacts, as a result of which the contacts can quickly oxidize or become dirty from contact with hands, for example. In case of any problems, it is necessary to wipe the processor contacts with an eraser.
OVERCLOCKING:
These processors can be overclocked in three ways:
articles). The frequency is set by the main multiplier, while turbo boost is turned off. The voltage for frequencies above the factory turbo boost frequencies will most likely have to be manually selected.
Overclocking example:
— Turbo boost case (and only in this case) the operating frequency of the processor is set by turbo boost — the turbo boost multiplier for one core is set to +1 to the maximum, for 2, 3 and 4 cores — the maximum multiplier. So for 4980hq, turbo boost overclocking has the following form — a multiplier of 45 is set for 4 cores, a multiplier of 44 is set for 2, 3 and 4 cores. manually.
— Overclocking with bug and throttlestop — available on any motherboard. Details:
http://forum.notebookreview.com/threads/4ghz-overclock-i7-4xxxxmq-h-to-i7-extreme-conversion-intel-haswell-cpu-microcode-bug-hack.7
/
In the case of CW, you do not need to remove microcodes from Bios, because. they are not there anyway. In the rest — I did not study the details, but you can find a discussion in the Haswell topic on the forum. Maybe I’ll add to this section later.
— Overclocking strap 125MHz — available on Z87 and Z97 chipsets. Allows you to reach frequencies higher than those available when overclocking simply by a multiplier. Straps are unlocked on the entire line of CW processors, however, no one will guarantee that overclocking will work in the case of your bundle. It depends both on the implementation of the motherboard and its Bios, and on the implementation of the processor (or its adapter). May be available on the entire line of CrystalWell processors, however, it is believed that selected crystals go to the top model 4980HQ, which is more likely to take higher frequencies.
Strap will not work in 166MHz, because it is tied to the cache frequency of level 4. The L4 cache has a multiplier of 16, at 100 straps it has a frequency of 1.6GHz, at 125 straps it has a frequency of 2GHz. 2656MHz, which he would have received at 166 straps, is impossible for him. There is no multiplier adjustment for L4 cache in this case.
Before overclocking with straps, you need to test the capabilities of your processor — what frequencies are stable and at what voltage, whatever it is, what to dance from.
When overclocking with a strap, make sure that all other multipliers are set correctly. For the first run with a strap, I recommend setting: bus 125 / strap 1. 25, main multiplier 28, Uncore / Cache multiplier — 28, memory frequency — 1066 (when using the strap, it starts as 1333), turbo boost is OFF, you can set fixed voltages for the core and cache . In case of a successful launch, increase the multipliers with voltages in turn.
Overclocking example:
General:
with intermediate checks! On average, CW processors take frequencies from 4.2 to 4.5GHz! Cash takes 3.8-4 (not uncommon with a voltage increase) to 4.2-4.4 on the most successful specimens! Pay close attention to cache overclocking! As well as overclocking the processor cores, unstable cache overclocking instantly leads to «Whea uncorractable error» bluescreens!
— Some motherboards do not display (display fundamentally incorrectly) the core voltage and SA data of processors in Bios and in CPU-Z. This is the norm. Also, you will not be able to measure it on the measurments points of the motherboard. Whether the specified voltage has been set or not — you can see in HWInfo by running the program with the sensors-only checkbox, the voltage is displayed in the Vcore line, but if there is rubbish — the only option is to see what’s in the VID lines.
-Cache frequency (Uncore ratio) must be set manually for any overclocking, I would recommend setting it to 36 if you do not plan to overclock the memory. It is worth increasing this frequency to overclock the memory, so the instances that I had in my hands kept my RAM at a frequency of 2200 with a Cache multiplier of 36-38, but in order for the memory to work at a frequency of 2400, I had to set the Cache multiplier at 40. The voltage for this element and for its overclocking can also be set manually with an override — 1.150-1.2v — safe, and, probably, the maximum voltage.
— The voltages of IO Analog and IO Digital, as in the case of Haswell, can be offset by +0.1 for stable memory overclocking.
-VRIN voltage is an interesting point. When overclocking 4790k, I lowered the input voltage to 1.73v, and this action helped to decently reset the processor temperature by shifting part of the work of the internal regulator to the external VRM. In the case of the Chinese adaptation of crystalwell, such methods led me to blackscreens in the load. Apparently, the Chinese adapter does not use the entire available array of power contacts, which causes a drop in the input voltage. According to my observations, for voltages lower than 1.3v Vcore, the standard 1.8v VRIN is quite enough for itself, but for greater reliability, you can throw up to 1.85-1.9. At 1.35v my stone already needed a minimum voltage of 1.85v, at 1.39v Vcore it needed to set 1.9V VRIN. With a constant margin, I set the voltage to 2v VRIN. Feel free to upgrade.
— Vcore voltage — my opinion is , which is the maximum safe voltage, as in the case of Haswell — 1.4v. It is desirable to set it with an override, however, in the case of easy overclocking, you can also make an offset of +0.05 or +0.1.
— I check the stability of overclocking by 3-5 runs of Linx 0.6.5 AVX for 6GB. I check the stability of memory overclocking and cache using the TestMemory program (TestMem5). It is worth determining the most stable memory and cache frequencies BEFORE overclocking the cores — because. with an increase in their bandwidth, the load on the CPU will also increase.
— It is also desirable to manually set the PCI-E line generation to 3.0. Not exactly related to overclocking, but this recommendation will be here, with all the other Bios settings.
By the way, by itself, overclocking the L4 cache from 1.6GHz to 2GHz, using the 125 strap, according to my observations, does not really give anything: a much more successful memory controller that does not cut throughput at frequencies above 2200, and could well fit into suitable high-frequency modules, for example 2600:
However, unfortunately, this makes no sense due to the low dependence of performance on the memory frequency due to the presence of the L4 cache.
BIOS MODIFICATION:
Anyone can install CrystalWell native microcodes in the BIOS of their motherboards, in some cases it can help solve problems that have arisen, for example, with overclocking. So, for example, I myself encountered the following points — without native crystalwell microcodes, MSI Z9 motherboard7 GAMING5 refused to overclock the memory, GIGABYTE Z87X-UD4H behaved even more inadequately than with them.
WARNING! You can get a brick by flashing a modified BIOS to your motherboard, so please take the time to back up the contents of your motherboard’s Bios chip before modifying. In the absence of a programmer or crooked hands, you can do it quite cheaply at the nearest service center.
Ready-made kitchen with microcode 40661 v17 for MSI and GIGABYTE (tested on MSI and GIGABYTE, bios obtained at the output of this kitchen did not start in the Asrock motherboard) based on the latest version of UBU.
Instruction:
1) put the firmware file in the folder with UBU
2) run via cmd [ubu firmware_file]
3) select «CPU Microcode» in the UBU main menu
4) select C — create FFS with MicroCodes
5 ) confirm — U
6) in the plate that appears, make sure that the list of microcodes contains 306C2, 306C3, 40661
7) press 0 — exit to main menu
8) select Exit
9) select 1 — Rename to mod_хххххх
10) you can make sure that the output file is correct by dropping it into the MC directory of MCE
Ready-made kitchen with four microcodes I have (F, 10, 12, 17) for Asus and Asrock motherboards (tested on one Asrock motherboard, in theory it should work with absolutely all brands, but it needs to be checked), based on the old version of UBU .
Instructions:
1) put the firmware file in the UBU folder
2) run via cmd [ubu firmware_file]
3) in the UBU main menu select «Update intel CPU Microcode»
4) select 1. Update CPU Microcode Haswel and/or Broadwell
5) in the case of a motherboard on an 8x chipset — replace the Haswell microcode with the CW microcode, I use 17. (In this case, most likely, the motherboard will not start with Haswell after flashing the modified BIOS), in the case of a motherboard on a 9x chipset — choose Haswell microcode for Haswell, and then change the Broadwell microcode to CW microcode.
6) Press any key, you get to the main menu, select 0 — Exit.
7) Choose 1 — Rename to mod_xxxxxxxxxx
8) you can make sure that the output file is correct by running it through MCE
FINALLY:
A fairly common question is whether the Chinese gluing will fall apart over time? Answer: I don’t know. No one knows. There are usually no problems with large BGA soldering without mechanical impacts, and there are factory models of processors with a similar layout on the market — 7980xe for example.
As for mechanical effects — there are already processors with problems after clamping in the socket, such processors work without problems where they were originally located, but when you try to install in another motherboard, one of the channels stops working due to differences in clamping and socket shapes memory:
At the same time, as a result of such deformation, nickels are pulled out of the adapter textolite, repair by heating or reballing is impossible:
, and look for individual adapters on sale (search has not yet been successful).
This material will be edited and updated as new information becomes available.
This material was written by a site visitor and has been rewarded.
My experience of upgrading from Core i7-4960hq (2013) to Ryzen 5 3600 (2019) without changing graphics card (GeForce GTX 1080Ti)
This article was written by a website visitor and has been rewarded.
About a month ago, when the dollar jumped so sharply that the prices in the store behind it had not yet caught up, I hastily purchased a kit on Ryzen 5 3600, which included mother msi b450-a pro max , proc ruzen 5 3600 oem , the cheapest memory kit goodram GR2400D464L17S/16GDC for 16 gigs and cooler deepcool gammaxx 300 900 09 .
I put up my old kit for sale at about half the price of the purchased kit for Ryzen 5 3600. It included an i7 4960hq processor, a msi b85-g41 pc mate motherboard, a memory 12 gigs ddr3 1333 (2×2 +2×4 ) and cooler zalman cnps80f .
recommendations
I didn’t feel much growth in games, but it appeared in work tasks. This is video encoding, rendering on a physical engine in cinema4d and the increase is almost twofold everywhere.
I wanted to evaluate the increase from Ruzen in more detail, but I was too lazy to disassemble the computer and put the old set back in. My kit for 12k was lying on the board for a month and did not find a buyer until today. In this regard, while I have both sets for a short time on my hands, I decided to quickly test them.
On the left there will always be ruzen 5 3600 and on the right i7-4960.
Ruzen in stock, offset slightly lowered by -0. 0375v, memory overclocked to 3800 from 2400, timings selected maximum-minimum and factory bus in sync with memory 1: 1, its frequency is 1900.
i7 4960hq — forced frequency at 3900 for all streams, the voltage is slightly raised. The memory is at the nominal frequency of 1333, but the timings are at the lowest possible and this is 1 point each and the command rate is reduced from 2 to 1. As a result, both processors operate at close frequencies with a maximum difference of 200 MHz, depending on the boost on amd.
As a result, both processors have memory with the same latency of approximately 70ns.
CPU-Z test, almost 2x increase in multithreading.
AIDA GPGPU test ids, increase approximately 2-fold
Bluebench parrots, increase approximately 2-fold.
Crystalmark zpumark, increase is close to twofold.
Crystalmark memorymark, growth is too small.
Chess benchmark, the increase is impressive and more than the difference in flows.
X264 benchmark, the increase is almost twofold, which is confirmed in practice.
Same with X265.
AMD scored almost twice as many points in this pseudo-gaming benchmark.
A similar benchmark of a newer generation, here the difference is impressive, but it cannot be called double.
Winrar test, double gain in single thread, but in multi thread a little more than one and a half.
Now let’s move on to gaming tests. The graphics card is a 1080Ti turbine.
Having played a little with the curve, we managed to reach the frequency of 1800+ on the chip in any game, it practically did not sag lower. The video memory was not overclocked.
The operating system and video driver are the same, namely 10×64 1909 and 442.59 . I just connected the old kit to the same system as the new one.
World of tanks encore rt, ultra settings, resolution 1920×1080, RT disabled. There is an increase in rare events, but the final score is the same, which is not encouraging.
Need for speed heat. 1920×1080 ULTRA. There is no benchmark, but I just stupidly pressed forward from the save point for a minute, after waiting before that until everything was loaded for half a minute. The car went forward destroying everything on the way, as a result, it fell into the river 2 times and returned, so the scene is the same on both platforms. Formally, there is no difference, but I noticed looking closely that on Intel some details and textures are not loaded at all even after a while, while on AMD everything is loaded, although with some delay at the beginning. You can consider this a good increase.
Shadow of the tomb raider. 1920×1080 DX12, the settings are just high , at the highest there is a strong emphasis on the GPU, even on Intel. There is an increase, but judging by the numbers, it does not bring much joy, especially considering that this is one of the processor-dependent games. I expected to see more difference here due to the fact that the Intel processor is already 7 years old.
Resident evil 3 remake. The settings are set so that there are no orange warning items and an emphasis on the GPU, respectively. Resolution 1920×1080. There is no benchmark in the game, but one of the cut scenes on the engine was used, then a static pose. The difference is not visible, except for 0.1%, and according to other FPS indicators on Intel, even a little higher. Again, no luck.
Gears of war 5 Bencmark. 1920×1080 Ultra. Growth is not visible again, the advantage of AMD is at the level of error, probably the focus is on the GPU and it was necessary to test it at high.
Forza horizon 4. 1920×1080 Ultra benchmark. The situation almost repeats the previous one, probably the ultra-preset causes the 1080Ti to hit the GPU even in full HD and it was necessary to lower the settings.
Battlefield V 1920×1080 Ultra DX11. A landing cutscene in Norway was used, then a static pose for regular intervals. On Intel, the FPS is even slightly higher, however, at the level of error. On the other hand, again the focus on the video card is to blame and ultras are also inappropriate here.
Assassin’s creed odyssey. 1920×1080 High. Another one of the most CPU-dependent games and there is no difference here, Ruzen even lost on the minimum wage. There was no emphasis on the GPU, since the settings were not ultra and the GPU load in places fell below 90%. Not a fan of this game, just downloaded the benchmark for the sake of knowing that this is a popular test for processors.
Crysis 3. Very high. 1920×1080 FXAA. Used a long cutscene with a tower in the second mission. Here is a completely different picture and the best gaming gain. You can observe an increase of up to 40%, which is a lot, and this is also felt during the game with the naked eye, it clearly goes smoother on the Ruzen.
From everything we can conclude that AMD has shown itself very well as a processor for work applications relative to the older Intel and the increase is even higher than the threads. My costs in this case are proportional to the increase.
But if we take gaming applications, then only 1 out of 9 games has a distinct increase.
So we can conclude that Intel cores are better suited for games, and AMD cores for work tasks. But new generation consoles are coming on the same zen2 architecture as my ruzen and the picture can change dramatically towards amd, and games will start to be better sharpened already for zen2 cores and not intel and ruzens of intel will win not only in work tasks.