GPU GPU for HP 800 G1 SFF with only 240W power (GT 1030 passive from Amazon would be good??)

Bangforbuck

Chassis Packer
Original poster
May 30, 2019
14
1
I have HP 801 G1 SFF ELITEDESK with only 240W PSU and I am looking for info about Graphics card upgrade.
https://support.hp.com/us-en/document/c03832938 more info here.

I am looking at whether this card would be good to buy or not. As far as I know GT 1030 LOW PROFILE
would be the best one I could fit into SFF case with only 240W PSU-is this correct info?
Gigabyte GeForce GT 1030 GV N1030SL 2GL Silent Low Profile – 2 G
Amazon productI am looking at this used one that only uses 10W of power and has very good price for used one.
Would it be risky to just buy used one?
 

Mitchie23

Minimal Tinkerer
Apr 11, 2019
4
1
GT 1030 falls in the Tier 5 GPU, which is Budget-Friendly Starter Kit. You get to save a lot of money but don’t get much out of performance. You might be interested in this tier if you’re not really the kind of gamer who must have the latest games right away, and prefer playing the classics or the less demanding games. You can still get 60fps for 1080p but it is not as stable as your Tier 4 GPUs. Anything higher and you get a lot of lag in-game.
 
Last edited:
  • Like
Reactions: Bangforbuck

Scott

Caliper Novice
Nov 29, 2016
28
17
If you want the best that can fit that case, You should be able to use any low profile GTX 1050ti or once it goes up for sale the new low profile GTX 1650 from Zotac, both of which are rated at 75W. 240W should be enough for one of those cards paired with a typical 65W CPU - that is only 140W in total.

Also, GT 1030's are rated at 30W, or 20W for the DDR4 version (don't get a DDR4 GT1030! make sure you are getting GDDR5 VRAM). Not sure where the 10W is coming from. Looks like the one you are looking at is a GDDR5 version though.
 
  • Like
Reactions: Bangforbuck

Bangforbuck

Chassis Packer
Original poster
May 30, 2019
14
1
GT1030 GDDR5 is indeed a 30W GPU. As mentioned, make sure you get the GDDR5 version and not the DDR4 one. Difference in performance is quite huge.
2236 is passMARK for GT1030 at https://www.videocardbenchmark.net/gpu.php?gpu=GeForce+GT+1030,
which in reality makes it about 3 times faster for games than my integrated HD 4600 with PASSMARK 720, right?
But which version of GT 1030 is it benchmarked? I does not say whether it is GDDR5..

Here I also cannot see which one is it https://gpu.userbenchmark.com/Compare/Nvidia-GT-1030-vs-Intel-HD-4600-Desktop-125-GHz/m283726vs2168
 

Bangforbuck

Chassis Packer
Original poster
May 30, 2019
14
1
If you want the best that can fit that case, You should be able to use any low profile GTX 1050ti or once it goes up for sale the new low profile GTX 1650 from Zotac, both of which are rated at 75W. 240W should be enough for one of those cards paired with a typical 65W CPU - that is only 140W in total.
I do not think this would be smart, as far as I read in my few days 'research' about this.
All 75W GPUs have specification for 300W PSU or more and I have seen many of them!
HP800 G1 with 240W is made only for 35W GPUs as far as I know (and even then there is not much reserve just in case).

240W PSU does not really give you 100% efficiency on average, 240W is just the peak.
How much could I really expect my PSU to give is a question, I read that it might be about 75%,
so that could be just 180W....
i5 CPU should take 75W, SSD 10W (and I might add one more which is then 20W), 20W 4x4GB GENESIS DDR3 RAM,
Motherboard 20W, and GPU 75W (if not more?)=about 220W already!!! It's already past those 180W, and too close those 240W too!!!
Plus peaks could be higher, so there should be some reserve-JUST IN CASE and not to worry about it while enjoying a game.
So in intensive games for GPU, I could have blue screens, freezes, I could damage my components (motherboard, CPU,...)
or even burn my whole house (heheheh). I do not want to have any risk of any of these, I want 0% risk as compared to now
when I only have integrated GPU.

I read online that I could put in HP GPU 320W, this would be the max possible power I could get in HP800G1 SFF.
I do not know which model of GPU is that, but if I really added it (however probably to put this money somewhere else),
then I would be able to get in GTX 1650 and similar as you suggest.
 
Last edited:

Bangforbuck

Chassis Packer
Original poster
May 30, 2019
14
1
GT 1030 falls in the Tier 5 GPU, which is Budget-Friendly Starter Kit. You get to save a lot of money but don’t get much out of performance. You might be interested in this tier if you’re not really the kind of gamer who must have the latest games right away, and prefer playing the classics or the less demanding games. You can still get 60fps for 1080p but it is not as stable as your Tier 4 GPUs. Anything higher and you get a lot of lag in-game.
Well I would be playing games from about year 2015 and older, and I would also like to try ANDROID and iOS EMULATORS for games,
which seem demanding..I read that GT 1030 is not much better than my HD 4600, maybe just 3 times better, but it might help for that?

I now have 4x4GB KINGSTON HYPERX GENESIS 1600 DDR3 RAM and I am thinking it might be smart (would it?)
to buy bigger GENESIS RAMS 8GB (if I upgrade GPU or not), so that I would then have 24GB (and later maybe even 32GB).
In the future I could also later use GENESIS RAM for another computer for office, whereas PSU and GPU would not be useful there.
 
Last edited:
  • Like
Reactions: Mitchie23

Mitchie23

Minimal Tinkerer
Apr 11, 2019
4
1
Well I would be playing games from about year 2015 and older, and I would also like to try ANDROID and iOS EMULATORS for games,
which seem demanding..I read that GT 1030 is not much better than my HD 4600, maybe just 3 times better, but it might help for that?

I now have 4x4GB KINGSTON HYPERX GENESIS 1600 DDR3 RAM and I am thinking it might be smart (would it?)
to buy bigger GENESIS RAMS 8GB (if I upgrade GPU or not), so that I would then have 24GB (and later maybe even 32GB).
In the future I could also later use GENESIS RAM for another computer for office, whereas PSU and GPU would not be useful there.
if that's the case, then GT 1030 is okay. you may opt for the higher tiers in the future, sooner or later the price will go down. found a good read about GPU tiers though, it's entitled 2019 Graphics Card Tier List and Comparison. for your perusal.
 

Valantar

SFF Guru
Jan 20, 2018
1,353
1,144
I do not think this would be smart, as far as I read in my few days 'research' about this.
All 75W GPUs have specification for 300W PSU or more and I have seen many of them!
HP800 G1 with 240W is made only for 35W GPUs as far as I know (and even then there is not much reserve just in case).

240W PSU does not really give you 100% efficiency on average, 240W is just the peak.
How much could I really expect my PSU to give is a question, I read that it might be about 75%,
so that could be just 180W....
i5 CPU should take 75W, SSD 10W (and I might add one more which is then 20W), 20W 4x4GB GENESIS DDR3 RAM,
Motherboard 20W, and GPU 75W (if not more?)=about 220W already!!! It's already past those 180W, and too close those 240W too!!!
Plus peaks could be higher, so there should be some reserve-JUST IN CASE and not to worry about it while enjoying a game.
So in intensive games for GPU, I could have blue screens, freezes, I could damage my components (motherboard, CPU,...)
or even burn my whole house (heheheh). I do not want to have any risk of any of these, I want 0% risk as compared to now
when I only have integrated GPU.

I read online that I could put in HP GPU 320W, this would be the max possible power I could get in HP800G1 SFF.
I do not know which model of GPU is that, but if I really added it (however probably to put this money somewhere else),
then I would be able to get in GTX 1650 and similar as you suggest.
First off, do you have two separate threads on this subject? That's really confusing - don't double post unless you want to confuse the people helping you.

Secondly: there is quite a bit that's off in the quoted text above. I'll go through it point by point so that you can get a clearer picture of your needs and options. Sorry if this turns into a novella-like chunk of text - I tend to overexplain. Still, more information = better choices :)

1:
240W PSU does not really give you 100% efficiency on average, 240W is just the peak.
What you're talking about here isn't efficiency, but the rated power output (and how it can differ from the marketing name). This point really has three parts, so I'll go through each:
1.1: (Power conversion) Efficiency is the amount of waste heat produced by a power supply when converting a given amount of power, but this does not affect the output power - lower efficiency increases input power (how much it "draws from the wall") but the output power is not affected by efficiency, just how much power is wasted as heat while providing the output. In other words, a 240W PSU should always output 240W - but it'll create a lot more heat doing so if it's rated at 75% efficiency than if it's rated at 95% efficiency. The difference is that the 75% efficient unit will draw 320W from the wall when outputting 240W while the 95% efficient unit will only draw 253W. All power above 240W is waste heat that the PSU has to dissipate, so the PSU fan on the less efficient unit will have to run faster, and it'll still be likely to fail earlier.
1.2: PC PSUs are rated for continuous power, not peak. AFAIK, the ATX spec (which all standard PC PSUs must meet) requires ratings to be for continuous output power (though don't quote me on that - it is the norm, at least). Peak output power is usually higher (though not always!), but also not particularly safe (and often quite inefficient). The heatsinks/fan are also usually not designed for dissipating more heat than required at 100% load. In other words, a 500W PSU should be able to output 500W "for ever" (every PSU will eventually die, but it should take a few years), but might also be capable of handling 600W short-term loads - this depends on the design and quality of the PSU, though. Bad quality PSUs might claim a higher rating than they really have, but that is rarely a problem in established markets these days (it was quite common 10-15 years ago). Most serious PSU reviewers test overload conditions both to see how the PSU handles them and if/when overcurrent protection (OCP) kicks in.
1.3: Not all ratings are equal. PC PSUs supply several voltages, and the output of these are normally added up for the rated wattage. Modern PCs almost exclusively load the 12V rail under normal use (which is why modern high-end PSUs can often deliver their entire rated output on that one rail), which means that older design PSUs where the 12V rail is disproportionately weak (PCs used to use 3.3V and 5V a lot more back in the day) can underperform. I have an old PSU lying around that exemplifies this beautifully:

It's nominally rated at 400W, but the 12V rail is only rated at 15A, or 180W. The 400W rating is in fact rounded up from 380W when combining the rated outputs of the 5V, 3,3V and 12V rails. In other words: check the label on your PSU for its +12V output rating. That's the most important number for a modern system. I mentioned to you in another thread that I'm doing a similar project in a Dell Optiplex 990 SFF, and its 240W PSU has a maximum 12V output of 200W.

2: Your power consumption calculations are way too high. A normal SSD does not use 10W of power (the very hottest m.2 NVMe SSDs can reach 7-8 under peak power draw, most are half that or less, and under normal usage in the 1-2W range if not even lower, while enterprise-grade HHHL PCIe SSDs might exceed 10W). SATA SSDs generally use less power than NVMe drives, and rarely exceed 2W under the heaviest loads. SSDs in consumer use cases spend the vast majority of their time idle, in which case they barely use power at all. 4 DDR3 RAM sticks do not consume 20W either, and the motherboard certainly doesn't consume 20W on its own. A far more sensible calculation is 5W per SSD for NVMe drives and slightly more than 5W per two SSDs for SATA drives, and somewhere between 15 and 30W for the motherboard and RAM combined, depending on the system (higher powered CPUs cause higher VRM losses as there's more power converted from 12V to the CPU's required voltage, but VRM losses for standard, <95W CPUs are negligible). For a non-K i5 system with 4 RAM sticks I'd say 20W is a generous allotment for the motherboard and RAM. For reference, here's a server setup where they measured the difference between 4 and 16 sticks of DDR4 to be a scant 27W. DDR4 is lower power than DDR3, but not by more than 20-30%, so if 12 sticks of DDR4 consume 27W, 4 sticks of DDR3 shouldn't exceed 10W, and outside of the VRM the only "major" power draw will be any fans connected to the motherboard, and most PC fans consume very little power, many are below 1W.

3: Just like online PSU calculators, GPU manufacturers' "PSU requirement" specs are utter bollocks. They are generally based on worst-case scenarios simply because there's no way for the GPU maker to know what other components you have in your PC or what quality your PSU is - so they go way overboard just to be safe (and avoid frivolous lawsuits in regions where that's a thing). A 75W GPU might spike to 85W for a few milliseconds, but loads that short are of no importance unless your PSU is utter garbage. (Also, GPUs without external power connectors tend to control their power draw quite strictly as pulling more power through the motherboard slot might fry the motherboard.) Still, read detailed reviews for actual power consumption numbers (as they will always vary), and combine that with measurements of your other equipment. If you look around these forums, you'll find a lot of people running 75W GPUs off <200W internal DC-DC PSUs.

An example: My Optiplex peaked around 130W measured at the wall (i.e. including the losses from heat in the PSU as well as the entire rest of the system) running Prime95 on the CPU with no GPU installed. IIRC that rose to about 135W with a low-profile Radeon HD 5450 in there under the same load, and by another 10W or so if I also fired up FurMark. The CPU peaked at 75-80W according to HWMonitor, which isn't entirely reliable, but gives a decent ballpark number for internal CPU power draw. From that, I calculated that I should be safe(-ish) adding a ~100W GPU in there, which I was hoping I could undervolt my RX 570 to reach. In reality, gaming loads don't stress the CPU near as much as Prime95, and I'm able to run my RX 570 at stock clocks while staying under the 200W 12V output of the PSU. I've measured ~265W at the wall while playing Rocket League, which sounds scary high until you factor in that this (terrible!) PSU has an average efficiency of just 65%, meaning that a 265W wall power reading translates to somewhere around 175W internally. This is of course not entirely accurate (efficiency isn't linear, but varies with the output of the PSU, so in theory efficiency could be higher at this load, which would mean I'm closer to the max output than I think). Now, I'm pushing the limits of this build, and I'm very conscious of doing so - this is a hobby project, not my main PC.

Nonetheless, this demonstrates that a "95W CPU" (i5-2400) and 150W GPU (RX 570) don't necessarily translate to 245W power draw under normal usage. A 75W GPU would be well within the capabilities of a PSU like this - including yours, unless there's something really weird about how it's configured. You're right that it's smart to leave some overhead in your PSU rating (20-30% is a good number), but base that off real-world power consumption, not on-paper numbers, as those are often complete fiction.


Your main limitations here are as follows:
-Space: How big a GPU can you fit? Low-profile? Full height? Single or double slot? The manual seems to say LP only, but unless you have lots of other AICs there's plenty of room for a dual-slot card.
-Power: I assume your PSU has no PCIe power plugs, so you're limited to motherboard power. That means 75W, though many OEM SFF systems spec their PCIe slots lower. I can't find anything about this in the manual for your PC, so 75W might be fine, but there's not really any way of knowing without trying.
-Budget: How much can you/do you want to spend on this? If I were you, I'd order a GTX 1050 if you can afford it, and return it if it for some reason shouldn't work. It'll cost you slightly more than a 1030, but the performance difference is massive, and no matter what it'll be cheaper than the 1650 LP cards whenever they arrive.


Sorry for going on and on here, but hopefully this brings you a bit closer to figuring things out. A 1030 GDDR5 will likely be a perfectly fine GPU, but its value is terrible, and the 1050 is a lot better. An RX 550 might also be a good in-between solution, at ~25% faster than the 1030, but the 1050 is again >50% faster than the 550. All of them ought to work. Btw, here's a YouTube video of someone recommending a 1050 for use in that PC (at least that's the card linked in the description, the GPU they're installing is not a 1050). Whichever one you choose, it'll likely be good, and it's sure to be a lot better than the iGPU :)
 

Scott

Caliper Novice
Nov 29, 2016
28
17
Power use aside, I have a 1030 in my HTPC because there is only one PCIe slot and I was going for full fanless. I regret not going with the 1050 and either modding the case or the card to make it work and not being so obsessive about running no fans. The performance is "fine", but not really to the level I wanted. I was going for Rocket League at 60 fps 1080p and I can't quite hit that without lowering graphics settings, and Rocket League isn't exactly too demanding of a title.
 

sheepdog43

Caliper Novice
Feb 17, 2019
25
12
It's an older low end SFF HP with a non-standard older design power supply. I'd agree with you if we were talking about a newer Cosair SF psu, but not an OEM meant for an low end office system, any overhead is there to account for differences in the PSU/config and wear over time. It's built to meet a price point, not leave enough overhead for a GPU or the occasional spike in power draw.

Sata models use more power because they need an extra chip to convert to sata protocol. Crucial M4's state right on the case it's a 2amp drive (10 watts), Samsung 850 claims 1.4 amp (7 or 8 watts), while the 860 is rated at 4watts. I've seen older Intels push almost as much as the 850 and while they may spend most of their life idle we're worried about max power draw, not how little you can use. More on this in a second.

8 watts is really pushing it for a board, you're not powering just a chipset, onboard sound alone can draw several watts especially if driving headphones.

1watt per fan? While many are less, an 80mm can draw .14amp ( almost 2 watts), a 120mm can draw .33 which is 4 watts and that's when new and not accounting for dust accumulation and wear on the bearings/bushings. Like the ssds, many will state right on them what the power draw is.


Stated power draw numbers do tend to be higher than actual power draw but none of this is new and in tip top shape so it's better to go with them than to simply make guesstimates. I'm not trying to nitpick, in most situations all of it falls within a range of close enough to not matter but in this case the overhead is so slim it's necessary. This isn't a high end psu and worse it's not even relatively new, it's simply not going to take the abuse your good psu will.
 

Valantar

SFF Guru
Jan 20, 2018
1,353
1,144
It's an older low end SFF HP with a non-standard older design power supply. I'd agree with you if we were talking about a newer Cosair SF psu, but not an OEM meant for an low end office system, any overhead is there to account for differences in the PSU/config and wear over time. It's built to meet a price point, not leave enough overhead for a GPU or the occasional spike in power draw.
So you got the impression I didn't know this from the fact that I was talking about a very similar build I have done myself? Including cautioning that I knew I was pushing things more than I would recommend others to do? Not to sound harsh, but you ought to read a bit more closely. Unless the PSU in this is either of completely terrible quality or has seen very strenous use (unlikely given that it has likely not run a GPU before) and assuming it has a 12V output of at least 200W (which I asked the OP to check), it should handle 150W output just fine for a while more. And realistically, with a GTX 1050 the power draw while gaming will be closer to 100W than 150. Of course it could be a good idea to open up the PSU and look for bad caps, signs of overheating, or other obvious signs of wear/damage, but I'm wary of recommending opening up a PSU to someone unless they seem comfortable doing this in the first place. That was not the case here.

Sata models use more power because they need an extra chip to convert to sata protocol. Crucial M4's state right on the case it's a 2amp drive (10 watts), Samsung 850 claims 1.4 amp (7 or 8 watts), while the 860 is rated at 4watts. I've seen older Intels push almost as much as the 850 and while they may spend most of their life idle we're worried about max power draw, not how little you can use. More on this in a second.
On-label ratings for SSDs have no relation to real-world power draw. Look at these results from AnandTech, for example. Note that writes are where SSDs use the most power except for heavy mixed workloads, which are essentially nonexistent in consumer use cases. Sure, some drives reach around 4W, but those are relatively old and power-hungry designs - newer SATA drives are more efficient, and consumer workloads are generally more read- than write-heavy, which has significantly lower power usage (particularly with TLC and QLC). And as you say yourself, drive loads are bursty, which means that overdoing the needs per drive becomes increasingly wrong for each drive added, as 1 SSD might average around 1W and peak at 4, 2 will average around 2W and on paper peak at 8 - but this will never happen in reality as pounding both drives with writes simultaneously over any amount of time is extremely unlikely. As for SATA drives needing another chip, I think you're thinking of external SATA-based drives using USB-to-SATA bridges - internal SATA drives have no "extra" circuitry compared to other SSDs. They have the same setup as everything else of the controller chip, DRAM and NAND (plus power delivery circuitry and other stuff, of course). And SATA controllers draw a lot less power than NVMe controllers due to their lower performance - NVMe controllers have more and higher speed processing cores to cope with the higher data throughput, leading to higher power consumption.

8 watts is really pushing it for a board, you're not powering just a chipset, onboard sound alone can draw several watts especially if driving headphones.
A non-Z chipset has a TDP of a handful of watts, and often uses less when it's not actually stressed by I/O. The VRM of a standard-wattage CPU is in the same range. Other controllers (USB, SATA, Ethernet, etc.) are normally in the mW range. The integrated audio on an OEM board does not have a powerful enough amp to consume that kind of power - enthusiast-grade audio on a retail board is another thing, obviously. Most OEM PCs have low-end audio similar to what you find on laptops, which definitely doesn't put out several watts of power. 20W for the motherboard+RAM is perfectly sensible.

1watt per fan? While many are less, an 80mm can draw .14amp ( almost 2 watts), a 120mm can draw .33 which is 4 watts and that's when new and not accounting for dust accumulation and wear on the bearings/bushings. Like the ssds, many will state right on them what the power draw is.
And again, these ratings usually account for what you say, and are significantly higher than actual power draw for a new fan. 1W for a normally sized fan that isn't designed for very high static pressure or high RPM (those generally have higher power motors) is perfectly reasonable. Going through my stack of spare fans at the moment:
NZXT 120mm (from my H200i): .16A/1,92W
Nidec Gentle Typhoon 120mm 1850rpm: .083A/1W
Fractal Silent Series R2 92mm: .18A/2.16W
Nexus 120mm (this is really old! At least a decade): .3A/3.6W
Noctua NF-A14 FLX: .08A/.96W
Fractal Silent Series R3 80mm: .06A/.72W

In other words: yes, fans can draw more than 1W, but most are around that number, or at least not more than 2-3W. An increase like that is negligible even when we're talking a setup like this, so counting 1W per fan - or 2W if you want to be safe - is perfectly fine. There are of course more power-hungry fans - radial fans draw a lot of power (often more than 1A), as do high-speed fans - the Foxconn 80x20mm stock intake fan in my previously mentioned Optiplex needs 0,36A/4,32W, but that's a high-speed (and really noisy!) fan of a very different design than most case fans.

All in all, your remarks boil down to recommending the OP to add a handful of watts of headroom. With a 240W PSU, a 95W CPU (that will never need 95W while gaming), and a 75W GPU, the OP should be perfectly fine. As I said, my i5-2400+RX 570+4 sticks of DDR3 and an SSD consume somewhere around 175W internally (after PSU losses) while gaming. That's with the GPU at 100% load. It stands to reason that a GPU with half the TDP will lead to a significant reduction in power draw from that number, so at the very least below 150W for the whole system - which should be perfectly safe. I wouldn't run Furmark+Prime95 for hours on it, but what's the point of that?